Data and Al are an integral part of the reality we all live in – but

how can companies use these technologies to drive revenue

growth? What commercial models and monetisation strategies
are available to create data and Al-enabled assets with the
capability to deliver powerful future gains?

Existing literature frequently claims that data and Al can

facilitate “new business models”; but it rarely explains what
these models will look like.

Growth Reinvented addresses the “what” and the “how” by

exploring a detailed playbook of Al and data-enabled business
models. Drawing on dozens of real-life case studies and
existing company examples - including a brand new body of
original research - it provides a practical step-by-step guide for
companies looking to realise the value of and monetise their data
and Al offerings.

Rather than just focusing on current offerings with data and

AI, Growth Reinvented shows how new technology can be
designed as a standalone opportunity to drive forward systemic
change. Business leaders are urged to consider using data and
Al-enabled service and revenue models to enable a more radical
transformation: one that disrupts industries and carves out
exciting new avenues for growth.

In a world of unrelenting turbulence, Growth Reinvented’s advice

on how to implement agile, scalable data and Al-based business
models, will give businesses the insights and momentum they
need to achieve crucial competitive advantage.
Mika Ruokonen is an expert in Al driven
business transformation at digital
innovation and engineering consultancy,
Futurice. His 15 years’ experience in digital
technolo- gies and software solutions
have included various roles such as P&L
responsibility, business development,
research, sales and consulting. During that
period he has come to realise that while
businesses quite easily solve tech-related
challenges and invent great new products
and services, they com- monly struggle with
commercializing them.
Growth
Reinvented
Turn your data and artificial intelligence into money

Mika Ruokonen
1 Introduction:
the importance
of data and
AI-enabled
business models
There have been mind-blowing advances in digital
technologies and computing in recent decades. Today, there
are very few companies that haven’t, in some respect, joined

added value unlocked by digitalisation.

Over the past ten years, companies have an industrial economy born in the late 19th
placed particular emphasis on the impor- century to a new knowledge-based economy
tance of “big data”. How to collect relevant centred on abundant data.1 Slowly but
data from their own operations (or from surely, data has become “cool” and being a
third parties), how to analyse and use data, data scientist is a dream job for many.
how to turn data into new product/service
offerings, and how to add more value to More recently, data has had to share
the business through data have all become centre-stage with another hugely disruptive
critical priorities. This preoccupation has dynamic: artificial intelligence (AI). While AI
underpinned the seismic transition from is not an entirely new technology, the past
few years have seen it make its way out of
academic discourse and trade press opinion
pieces and into boardroom decision-making.
Terms such as machine learning, deep learn-
ing, neural networks, predictive analytics,
voice/image recognition and algorithms
have become familiar to many business
leaders. Global players including Facebook,
Amazon, Google, Airbnb, Ford, IBM, Apple,
Tencent, Baidu and Alibaba are investing
billions of dollars in AI capabilities, and
some of them have already described their
focus as “AI first”.2/3 At the same time, new
AI-powered companies are launching on a
regular basis, all looking to disrupt different
sectors with their game-changing ideas.

In this new industrial paradigm, many of

the world’s most valuable corporations are
essentially businesses built around sophis-
ticated, powerful and lucrative algorithms.4
At Google, that algorithm is PageRank
for ranking websites. At Facebook, the
algorithm puts people in social contact
with each other. Amazon has developed a
commerce algorithm to help people buy
and receive goods. Uber, meanwhile has a
navigation algorithm that enables people to
book a driver nearby, and to find the best
route between origin and destination. There
are many other similar examples.
1.1 Why data and AI-enabled
business models, and why
now?
To date, the conversation about data and AI has been
dominated by the technical perspective (i.e. how algorithms

what is technically possible when computers increasingly

take advantage of data and new algorithms); and the societal
perspective (i.e. what does the future of humankind, work
and society look like, and what ethical issues does the
adoption of AI raise?).

By comparison, while there has been some there are barely any descriptions of what
discussion around the commercial aspect of these new business models could be. How
AI, it has generally been vague, one-sided or can any company plan and implement new
incomplete. The role of AI in driving busi- data and AI-enabled business models when
ness process improvement and automating no one has made the effort to describe
operations has been quite well covered, what they are, how they work, and what the
but AI’s contribution to new business change towards them would look like? To
creation has hardly been touched on. How sum up, while there are many pieces of work
can companies grow and generate more envisioning the data and AI technologies
revenues with AI? What are the relevant of the future, there are far fewer that
commercial models and monetisation describe the resulting data and AI-related
tactics when talking about AI as a corporate businesses of the future.
asset? Answers to questions like these are
badly needed and in short supply. What sets this book apart is that its key
objective is to explore data and AI-enabled
Business books, articles and presentations business models. While current literature
often claim that data and AI can enable focuses on how companies can use data
completely “new business models”, yet and AI to amplify their existing product and
services, we raise the bar by discussing reduce their own risks and costs, increase
the commercial opportunities related to sustainability and safety, wish to benefit
using data and AI to create new products, from the latest technology and ultimately
services and business models, so high- find more optimal ways to run their daily life
lighting this technology as a standalone or business.
opportunity. We believe that companies
should approach data and AI as tools to While these variables were in place before
expand their businesses – by creating new the Covid-19 crisis, they have now further
products, services, innovations and winning accelerated as various industries expe-
more market share – rather than just seeing rience severe headwinds and companies
them as a way of enabling cost reductions.5 need to redefine their formula for building
successful and stable businesses in the
Given the recent Covid-19 pandemic and the long-term.
resulting changes in the global business
climate, it’s appropriate to ask why data In a world of unrelenting turbulence, imple-
and AI-enabled business models should be menting new data and AI-enabled business
on the agenda for companies right now. The models, combined with a commitment to
answer is that even before Covid-related transition towards technology-enabled
disruption, companies were coming under scalable new product/service offerings and
pressure from: revenue models, could be key to building
• Rapid transformation in their business resilience in business. Through these activ-
environment. ities, a company can reduce its dependence
• Changing attitudes and expectations on legacy businesses and assets, stay rele-
among their customers. vant to its customers, open new avenues for
• Tougher global competition. growth, differentiate from the competition
• Commoditisation of existing product/ and build an ability to bounce back from
service offerings. adversity. As we grapple with Covid -19,
• Redundancy or erosion of old business these activities have become more critical
models. and urgent than ever. Building new data and
• Rapid emergence of new business AI business models is arguably an important
models, product/service offerings and step in protecting revenues and margins
partnerships. that are now under increased pressure.
• New data and AI technologies emerging
for companies to harness. Even if disruption of the marketplace isn’t
imminent, it makes sense for business
The critical takeaway for companies leaders to follow the latest developments in
is that their current product/service data and AI and anticipate potential changes
offerings and business models might not in their sectors. Virtually all companies need
be relevant to their target market in the to be on a constant transformation journey,
future. Customers might want services finding new business opportunities and
instead of products (or vice-versa), look to reconsidering their business models.6/7
It’s important to note that data and AI – like companies can, and should, master how to
many other company assets – do not have extract business value from data and AI.
inherent economic value. Instead they are
tools to help businesses generate more Like the steam engine, electricity and
money for shareholders. Companies wishing internal combustion engine, AI belongs to
to harness data and AI technologies should an important category of innovations that
strive to understand how their assets can economists call general-purpose technolo-
be turned into tangible business value, gies – those that impact various industries
providing them with a competitive edge.8/9 and catalyse waves of complementary inno-
vations and opportunities.10 Assuming data
As with all rapidly evolving technologies and AI become ubiquitous, their impact on
and new business models, there are plenty businesses will be dramatic.11 But in terms
of uncertainties related to data and AI. of the technology itself and how data and
These include: AI-enabled commercial models will develop,
• Who is the customer? it’s still early days.
• Is there a market and what exactly is
that market?
• Is the technology ready to be
commercialised?
• What final product/service is being
sold?
• How do businesses create a pricing
structure and revenue model for data
and AI-enabled products and services?
• How should any data and AI-enabled
new business be positioned and
structured compared with the existing
business?

While discussions about digital business

models date back to the advent of the inter-
net in the 1990s (as we see in detail in Chap-
ter 3), to date there hasn’t been any serious
analysis of data and AI-enabled business
models. In part, this may be because
businesses feel like they are ill-equipped
to do so. But it’s important that they don’t
fall into the trap of seeing the commercial
aspects of data and AI as a mysterious or
complex “black box” – something that they
just can’t crack. The opposite is true: all
1.2
scope of this book
When talking about how businesses can revenue models that can be used for selling
harness data and AI, at least three per- data and AI-related assets.
spectives come to mind, as illustrated in
Figure 1: Why do we discuss data and AI? It’s import-
• How do data and AI change what ant to note that AI runs on data – there
businesses sell? is no AI without data. What’s more, now-
• How do data and AI change how adays the value of data can be amplified
businesses make money? significantly when it is bundled together
• How do data and AI change the actual with smart algorithms. As enablers of new
work done within businesses? business, data and AI are interlinked and
inseparable.
The focus and scope of this book is in
tackling the first two questions. In other The final question – how do data and AI
words, we are investigating how data change the actual work done within busi-
and AI can transform the way compa- nesses? – is not within the scope of this
nies create, deliver and capture value, book. We will not discuss process rede-
through changing what they sell (prod- sign using data and AI, the way in which
ucts and services), and how they make AI and data will change how and what
money (revenue models). In the following people work on, or how to run intelligent
chapters we discuss how data and AI can operations. There is a plethora of books,
be turned into commercial products and articles and blogs that explore the future
services (or commercial components of impact of data and AI on the role of hu-
larger solutions) and we also review the mans in companies, society and everyday
 Customers &
partners
Operations & Financials &
processes metrics

What is How money

being sold is made
In scope

In scope
How work is
Competence &
done Data &
know-how
technology

Vision &
Culture &
strategy
leadership

Figure 1: The scope and focus of this book (source: Mika Ruokonen)

life.11/12/13/14/15/16/17 Consequently, there is no technologies and business models stands.

need to repeat these discussions here. Another thing to note is that this book
treats data and AI as software as opposed
Neither is this book a “deep dive” on either to hardware innovations although hardware
data and AI technologies or business model plays a role. For instance, companies might
literature. There are many other books that need sensors for collecting data, servers
deal with these topics in great detail (on busi- and computing power to store and process
ness models see Johnson,9/18 Osterwalder their data or to run their algorithms. They
and Pigneur,19 Rissanen20 and on AI from a might also need equipment and devices such
tech perspective, Chollet and Allaire,21 Gray,22 as robots, machinery, cars, laptops, phones,
Raschka and Mirjalili23). Our guiding keywords drones, cameras, speakers and implants
in this book are “adequate” and “purposeful”: geared to the output of the algorithms and/
we review existing literature and thoughts or which will be bundled in the final product
to give the reader a meaningful picture of or service.
where current thinking around data and AI
But the intelligence relevant to the new step forward, insufficient when it comes to
product/service offerings and revenue helping firms thrive in turbulent times? As
models described in this book is typically AI is currently developing quickly, and few
sophisticated, state-of-the-art software companies have fully commercialised their
(no hardware can “get smart” without it). data yet, the honest answer is that we don’t
This deliberate choice of focus sets certain know. However, this book will explore possi-
parameters on the discussion in this book. ble responses to these questions.
For instance, there’s not much mention of
“robot” or “drone”. Instead, I talk about
“software systems”, “software engineering”
and “software development” in relation to
data and AI. Once again, there are other
excellent books for those interested in the
hardware side of things, covering everything
from the computing power required for data
and AI through to the business opportunities
related to robotics.

A caveat: while recent advances in AI

technologies have been remarkable, there
is a consensus among scientists and practi-
tioners that the technology is still far from
fully developed. For example, it may take
decades before computer-generated “gen-
eral intelligence” is possible.10 Consequently,
this book extrapolates to some extent, in
terms of predicting what the data and AI
technologies of the future will look like, and
surmises what the shape of consequent
business opportunities will be (especially the
new products, services and revenue models).

The cover of this book uses a strong word:

“reinventing”. Are we seeing a data and
AI-enabled revolution in the ways that
companies are making money? After all the
challenges caused by the Covid-19 pandemic,
are these assets helping companies to
completely reinvent their formula for stable
revenue generation and business growth?
Or is this just an evolution – a small logical
1.3
used

that are being used in this book. We commonly refer to the

following words and concepts when exploring data and
AI-enabled business models:

• Data: The digital footprint either of cognitive and creative processes,

people (private individuals and em- e.g. problem-solving or learning. The
ployees) or of machines. The essence, essence, applications and business ben-
origins and business benefits of data efits of AI will be discussed in Chapter
will be discussed later in Chapter 2.1 and 2.2. and in Appendix 2.
in Appendix 1. • Business model: A resilient and
• Big data: Data in such large quantities comprehensive business system for
that new tools and methods are needed creating, delivering and capturing value
to analyse it. This term is not used much for customers. In our view, a business
in this book as it is difficult to define the model is the overarching foundational
precise amount of data needed to make concept that encompasses the prod-
it “big”. Instead, we talk about data in ucts/services and revenue models that
general. are the focus of this book. Chapter 3
• Artificial intelligence (AI): A machine and Appendix 3 will delve deeper into
doing something that in the past was the business model frameworks.
believed to have required human
• Offering: The combination of products
and services that companies offer for
sale. In this book we use offering and
products/services interchangeably.
• Revenue model: The structure and logic
around how a company generates rev-
enue or income, including the revenue
streams that the company wishes to
implement. So, while a business model
describes how a company generates
value, a revenue model describes how
a company generates revenue from the
value it has created. The revenue model
is our answer to the question of “how
money is made”, and various data and
AI-related revenue model options are
described in Chapter 5.
• Customer: Someone who buys and
consumes data and AI-enabled offer-
ings. For the purposes of this book,
companies might have various target
groups for their data and AI-enabled
offerings and revenue models. These are
not limited to their existing customers,
but can include partners, subcontrac-
tors and suppliers, even companies
outside their own industry or value
chain. The customer’s perspective will
be discussed in Chapter 2.4.
1.4 Co-creation with
companies
This book was written at Futurice, a Euro- discussions through to four-day deep
pean digital innovation and engineering con- dive workshops (Chapter 6 refers to these
sulting agency headquartered in Finland. sessions in more detail). Finally, we also
Futurice designs and builds digital services collected data by studying dozens of compa-
for companies in various industries such as nies through online information sources.
banking, retail, media, logistics, telecoms,
healthcare, manufacturing and the public Figure 2 features observations about data
sector. With a total workforce of 600-plus, collection from talking to senior people
Futurice has offices in Helsinki, Stockholm, within firms. Many of them felt that they
London, Oslo, Berlin, Munich, Stuttgart wanted to identify new data and AI-enabled
and Tampere. This allows a multinational growth avenues for their firms, but their
approach to studying data and AI-enabled data and AI harnessing plans weren’t ready,
business models; it also provides a robust and they were curious to learn more.
framework for us to work with companies
that have these business models under
consideration or in the design process.
Futurice’s mission is to “build future-capable
companies”, which is also a nice umbrella
term for this future-oriented book. In
developing this book, we collaborated with
Lappeenranta University of Technology, a
research and educational institution located
in Eastern Finland. A research team there
has recently studied similar topics and they
provided a valuable intellectual sparring
partner during the writing of this book.

The raw material for this book included

various books, articles, presentations and
blog posts (for a full list, see the refer-
ences section at the end). In addition, we
conducted new data collection by working
with 20 companies in a co-creation mode:
the sessions varying from short 1.5-hour
“We are currently a local labour-intensive “Data and AI-enabled business models are a
service business with limited growth very topical thing for us right now. We have a
projections. But data and AI would enable us lot of highly valuable data that we would very
to build a new scalable technology licensing much like to leverage more in our business. But
business for the international markets and that
killer idea that would attract our customers
growth plans” and make new business models possible too”

— A Chief Strategy Officer — A Strategy & Development Director

“Building new data and AI-enabled businesses

is a strategic thing and very relevant to us, but “We have a strategic choice to make soon.
in order to get those new businesses built, there Either we focus completely on commercialising

in the industry overall, towards data-enabled else for free. Or we stick to the existing
leadership and ways of working. We are not business model and build data and algorithm
there yet, but it will eventually happen” commercialisation in small scale on the
side. Depending on this choice, we might
— A Chief Financial Officer

organisation, competences and culture”

“We always have the same problems: what is — A Chief Executive Officer
the business model for our data, how could we
get insights out of our data, and who would
gain value out of our data? As our industry
“We have a very large customer base. But our
is rather traditional, we really need to start
thinking outside the box”

— A Project Manager potentially add more value to our clients,

competition and build new revenue streams”

ideas related to our data and AI assets and — An Innovation Manager

they look pretty promising, but we need to take
them much further this year. And I need to get
my team to understand them and committed to “So far, our data and AI-related activities
execution” have been heavily focused on optimising the
operations in our core business. We do not yet
— A Chief Digital Officer
have any data and AI-enabled new business
concepts or even processes to identify and build
them. But we do have nascent ideas and we are
Figure 2: Comments from business leaders on curious to learn more on how to get started”
data and AI
— A Head of New Solutions
1.5 A sneak peek at the main

Figure 3 provides a “sneak peek” of the customers can discover new insights or
book’s key conclusions and recommenda- improve their operational performance
tions. e.g. flight planning.25
• Partnerships, e.g. between IBM and
The playbook outlined here consists of six Maersk, which have joined forces to
product/service offerings (the columns) create blockchain-enabled shipping
and seven revenue models (the rows) that data platform TradeLens, whose goal
can be used for data and AI when they are is to promote efficient, predictable and
sold as standalone assets. These various secure exchange of shipping and cargo
options, as well as their benefits and limita- information, enabling collaboration and
tions, are discussed in detail in Chapter 5. trust across global supply chains.26

What is the purpose of this playbook? You The purpose of our playbook is to make
may already have heard about companies these examples more accessible, explaining
commercialising their data and AI assets: which products/service offerings and
These include: revenue models could be created, and
• CB Insights which aggregates and analy- illustrating them with real-life examples.
ses massive amounts of market data by The playbook also aims to describe the
using algorithms, then delivers analyses combinations of product/service offerings
and reports to executives and investors and revenue models so that teams can start
through its digital platform.24 creating a comprehensive, attractive and
• Boeing which has started to open usable data and AI commercialisation plan
access to its equipment data so that its within their businesses.
 “What is being sold”

“How money EXISTING

DATA, DATA AND OUTCOMES
is made” PRODUCT/
ANALYSES
DATA
AI ASSETS AI-BASED OR
SERVICE PLATFORMS
AND REPORTS SERVICES CAPACITIES
OFFERING

depends
on existing
FREE,
product/
FREEMIUM
service
offering

depends
PRODUCT/ on existing e.g. a data
one-off use
TRANSACTION product/ dump or one one-off
of capacity
REVENUE service report
offering

depends e.g. how

on existing much work
PROJECT one-off per hour/
product/ is needed
REVENUE project day
service to deliver
offering assets

depends
on existing
ADVERTISING
product/
REVENUE
service
offering

depends continuous
on existing data,
LICENSING, continuous continuous e.g. a fixed
product/ analysis
SUBSCRIPTION use retainer part of fee
service or reports
offering provision

e.g.
depends e.g.
MARKETPLACE, attached to attached perfor-
on existing attached
PLATFORM, the size of to the im- mance
product/ to number
BROKERAGE, the revenue provements against
service of transac-
PERFORMANCE generation gained defined KPIs
offering tions
opportunity

BUSINESS sales of
MODEL additional
OVERTURN services

Likely model A model worth testing

Figure 3: Playbook of data and AI-enabled product/

service offerings and revenue models (source: Mika Ruokonen)
 Sell customers their Collect data on your customers → Enrich and analyse it to create
own data insights, observations and recommendations → Turn the recom-
mendations into a sellable product → Apply a monthly licence fee
for access to the recommendations

Build a data and Identify unused assets → Find ways to match supply and demand
AI-optimised platform for them → Create an algorithm that does the matching work
to match supply automatically → Build a marketplace that delivers assests on
and demand demand → Apply a commission feematically → Build a marketplace
that delivers assets on demand → Apply a commission fee

Build data and Collect data on the use of your product → Identify ways to provide
AI-optimised service the same value without selling products anymore → Provide access
concepts that replace to the products and optimise provision with algorithms → Invoice
ownership with customers based on outcomes or capacity.to the products and
access

Turn your core Understand what you know/do the best in the world → Think how
competence into an to build rules for how that knowledge/activity works → Create an
algorithm algorithm that does the job → Turn the algorithm into a sellable and
payable product → Apply a monthly licence fee

Figure 4: The killer concepts

Figure 4 presents another “sneak peek”,

this time of four “killer concepts” that
are discussed in Chapter 7. Whereas the
playbook presents the various options that
companies have – some more lucrative than
the others – here we focus on a few select
ones that companies are already imple-
menting, that seem to work well and deliver
good financial returns.
 Your culture Your Customers
• Innovation-focused culture • Problems worth solving
• Data and AI-focused culture • Jobs to be done
• Change management • Willingness to pay
• Core vs. new business • Customer selection

6 1
Your execution Your maturity
• Prioritised roadmap Six steps towards • Strategy and positioning
• Portfolio of initiatives data and AI- • Commercial models
• Development teams
5 enabled business 2
and business acumen
models
• Iterative development • Processes and ways
• Acceleration and scaling of working
• Data and technology
Go/NoGo 4 3 • Organisations and
competencies

Your business case Your ambition

• Value propositions
• Common goals
• Investments
• Proactiveness vs. reactiveness
• Revenue model selection
• Executive commitment and sponsorship
• Financial impact
• Product/service offering selection

Figure 5: A recommended six-step path for business leaders (source: Mika Ruokonen)

Figure 5 presents a third and final “sneak While these recommendations are a “path”
peek”: Our recommended six-step path for that companies can follow, they also serve
business leaders wanting to explore data as a guide to some of the topics that we
and AI-enabled product/service offerings are about to cover in the next chapters. In
and revenue models within their own com- Chapter 8, we return to these recommenda-
panies. We argue that teams should explore tions in more detail.
data and AI opportunities around their cus-
tomers (existing and potential ones), assess
their own data and AI maturity, as well as
setting their ambition level and creating a
business case for their data and AI-enabled
business models. If the opportunities seem
lucrative, they can move to implementation
mode.
1.6 How to get value out of
this book
This book aims to add value in three ways. First, we collect
and structure existing information around data and AI-
enabled business models: we bring technology thinking
and business thinking together to serve as a synthesis and
foundation for the reader to review and relate to. Second, we
generate new information on data and AI-enabled business
models, for instance how they work and how to implement
them, and consolidate this as a playbook for the reader to
use. Finally, using our case studies, we aim to highlight
the current opportunities and challenges companies are
experiencing around data and AI.

There are several groups of people that ogists closer together and help them to
will get value out of this book. If you jointly envision new business opportunities
are a business leader, the book will help related to data and AI.
you become familiar with new data and
AI-enabled business models. It will help you If you are a venture capitalist or financial
structure your thoughts on the topics, then analyst, the following pages will provide you
provide guidance on how to implement new with guidance and frameworks on whether
business models in your firm. If you are an or not to invest in a certain company that
R&D or product development professional strives to (or claims to) harness data and AI
working on data and AI, the book will help commercially. If you are a student or a re-
you understand the business potential of cent graduate, the book will offer tools and
your inventions, and help you to speak the frameworks that help you start your career
same language as the business people in in firms that harness data and AI in their
your firm. The aim in this book is therefore business. Finally, if you are a policy maker,
to bring commercial teams and technol- the ambition is that this book will help you
 TRADITIONAL COMPANIES IN DIGITAL INHERENTLY DATA AND
COMPANIES TRANSFORMATION AI-NATIVE COMPANIES

Strict focus on the traditional Quite a lot done with regards to Company culture and operations
(physical) core business. digitalisation in the firm already. built on top of data and AI by
However, also signs that the E.g. a new business portfolio in definition. Data seen as a core
business environment is changing, place, automation and optimisation valuable asset of the firm.
e.g. new competition emerging, of operations ongoing, digital AI development ongoing and de-
expectations of customers chang- revenue models experimented or livering results, strong data and
ing, commoditisation of existing implemented. AI capabilities and resources.
product/service offerings ongoing.

Data and AI-enabled business mod- Data and AI-enabled business mod- Data and AI-enabled business
els can be seen as a new disruptive els can be seen as an extension to models can be seen as the core
strategic opportunity to escape the existing digital strategy, and a commercial strategy in the
from the limitations of the current new opportunity to expand to new company, something that needs
business. markets and new revenue sources. to be 100% mastered in order to
succeed in business.

Figure 6: Three types of companies categorised according to their starting positions

(source: Mika Ruokonen)

understand the business potential of data development projects of the firm have
and AI so that you can create governmental been geared towards (note that these
support programmes that help companies firms might be considering other similar
which adopt these technologies to succeed. breakout scenarios on their plate that
have nothing to do with data or AI). Prime
From a strategy perspective, different examples of these companies can be
companies will be in very different starting found in traditional industrial sectors,
positions when it comes to data and AI- where the core businesses are often
enabled business models. Figure 6 illus- asset-heavy, but firms are simultaneously
trates three different starting positions. trying to identify new digitalisation
-related opportunities.
Depending on which category your company • For companies that have already worked
is in, the strategic importance of data and AI on digital transformation for quite some
business models is likely to be very different. time, data and AI-enabled business mod-
• For large traditional companies, it’s els might be a powerful new extension
common for data and AI-enabled busi- of the existing digital strategy, and be
ness models to be seen as a new dis- regarded as an opportunity to generate
ruptive revenue-generation opportunity more revenues.5 Companies like these
that differs significantly from the core are often found in the retail, banking,
business, which typically all the strategic IT and telecommunications industries,
 where vast experience of digitalisation
often already exists.
• Finally, for companies that are inherently As a summary of the rest of this book,
native with data and AI, the business here are our master conclusions and key
models presented in this book are takeaways on data and AI-enabled business
probably already the cornerstone of models. The chapter where each of the
their commercial strategy – fundamental topics is discussed is numbered in brackets.
to making the business work and gen-
erating value for shareholders. These The business opportunities enabled by
companies could be digital business data and AI (2) as well as business model
ventures, digital media companies, innovation need to be seen as a source of
website providers or online marketplaces growth (3). By building a solid foundation
and platforms whose business is often for data and AI business models (4) and
fully dependent on the success of the applying a coherent playbook of product/
chosen digital operations. Typically, they service offerings and revenue models (5),
have very few physical operations or businesses can start building new, stable,
assets in place. scalable and recurring revenue streams
(6). These efforts may result in a new data
The word strategy applies to all the three and AI-enabled company culture, greater
categories, but in three different ways. At business resilience and transformative
the same time, the creation of additional systemic change to the firm (7-8).
shareholder value using data and AI is hap-
pening through new revenue generation in But let’s now move ahead to Chapter 2,
all three types of company (see also Lewis where we delve deeper into data and AI to
and McKone27 and McKinsey28). find out what these disruptive technologies
entail, and how they are powering various
business operations today.
2 Data and AI
as enablers of
business today
In this chapter we set out to achieve three objectives. First,
we study how data and AI are used for business purposes in
general. Second, we work to understand how companies can
identify important problems worth solving (belonging to data
and AI customers), and how to solve those problems using
their data and AI assets. Finally, we review the perspective of
data and AI customers, to understand why and how they buy
these assets.

A key conclusion of this chapter is that,

while there are still obstacles that prevent
companies from taking full benefit from
their data and AI assets, data and AI-enabled
businesses are expected to proliferate and
play a bigger strategic role across a wide
range of sectors.
2.1 Why is data important in
business?
Since the start of this millennium, more it is no longer clear whether a product or
and more companies have started to talk service is valued for its intrinsic value (e.g.
about data. Many of them refer to “big its physical features) or for its ability to
data”, a term that relates to data in such process or generate data.31
large quantities that new tools and methods
are needed to analyse it. There is no lack In many industries, data sits at the centre
of publications and seminars on how to of companies’ strategic goals and activities.
collect, analyse and use data in business. Data plays a pivotal role to help companies
Indeed, firms have become borderline get closer to their customers, increase
obsessed with their data and are moving operational productivity, improve products
towards strategies that see data as a and services, and steer and optimise sales
core asset. It seems that every company and marketing operations. Data can also
is becoming a data company, whether or help improve how existing assets are used,
not it has figured out exactly what that help inform smart investment decisions, and
means.29/30 The current turbulent times have empower people across the organisation.
further accelerated these developments as
business leaders demand more visibility (in Figure 7 highlights typical benefits that
other words, data) into what’s going on in companies seek to extract from their data
their businesses so that they can determine (see also Bean32, Ruokonen33).
how to thrive in an uncertain world.
Traditionally, data has been viewed as an in-
Data, then, is increasingly becoming the ternal asset that is valuable to a firm’s inter-
core value of products and services. Often, nal operations (e.g. used in identifying cost
 Investment
decision

Ecosystems and Intelligent

integration operations

Sales and Valuable Agility and speed

marketing Data in operations

New business Efficiency and

opportunities optimisation

Customer
centricity
and user
experience

Figure 7: Typical business benefits of data (source: Mika Ruokonen)

savings or risk). Nowadays, it is increasingly considerable. To summarise, with data as

regarded as an asset that can help generate a raw material, various new types of value
new revenue streams, develop competitive creation are possible – be it within the firm
new products and services, and that can or externally.32/34/35/36
even be used as a commodity that is shared
and traded. The commercial opportunities of harnessing
data lie not only inside the company but can
Using data as an internal asset requires far also be extended to encompass the firm’s
fewer security, intellectual property and le- entire value chain. Figure 8 illustrates the
gal precautions. But the potential economic case of a manufacturing firm, such as the
gains are inevitably limited by the company’s Siemens conglomerate or elevator manu-
internal structure and situation. Using data facturer Otis, that makes products to be
as an asset outside the firm’s boundaries is sold to its industrial customers. We identify
less common and requires various methods at least 12 different opportunities that are
to distribute the data to potential customers by-products of harnessing data; some of
(something that we tackle in Chapter 5); these are consolidations of various smaller
but the commercial gains as a result may be opportunities.
 Cloud and Analytics

Suppliers Production plant Products Customers

11 7 6 5 4 3 2 1

Subcontractors
Employees
and partners

10 8 9

12

Opportunities for harnessing data

1. Data provision to end customers 7. Data-enabled inventory and warehouse

management
2. Personalised services e.g. predictive
maintenance and spare parts 8. Data-enabled product and business development

3. Data-augmented sales operations, field force 9. Optimisation of competences and workforce

optimisation productivity

4. Intelligence in manufactured products 10. Data provided and sold to subcontractors and
partners, procurement productivity
5. Optimisation of production e.g. speed,
efficiency and quality 11. Data provided and sold to suppliers, procurement
productivity
6. Predictive maintenance of production equipment
12. End-to-end demand-sensing data ecosystem

Figure 8: Examples of data-harnessing opportunities in a

company’s value chain (source: Mika Ruokonen)
In the ideal scenario, the entire value chain • Company culture does not support
could become an “end-to-end demand-sens- commercial use of data, e.g. mindsets,
ing data ecosystem” that brings together processes, ways of working, inhibit data
suppliers, subcontractors and partners to commercialisation.
meet the needs of the end-customers in • Data processing or commercialisation
full.37/38 While this scenario is often far from projects or programmes are not prior-
reality in manufacturing firms (and value itised in development roadmaps or not
chains), the vision is well-established. receiving enough managerial attention.
• Inability to form common data stan-
When assessing the commercial opportuni- dards, or data platforms among busi-
ties of harnessing data, a lot depends not ness units and/or various players in the
just on how much data a firm has, but on value chain and/or industry.
how well-equipped it is to make sense of it.39
Business leaders often complain that their In summary, for many companies, using data
companies have collected vast amounts of for commercial purposes is still new, and
unique, high-quality data, but have not yet they may be far from ready to harness all
proved capable of unlocking the business the available opportunities.
value from these good-quality data assets.
There are many challenges and obstacles A comprehensive overview of the business
for companies seeking to leverage data for opportunities around data is available in
business purposes including: Appendix 1.
• Lack of data overall, poor quality of data.
• Poor data infrastructure, data is
scattered and not integrated to form
commercially valuable packages.
• Lack of capacity to understand, process
and/or commercialise large amounts of
data, or to view data opportunities from
various perspectives (e.g. customers’,
suppliers’ and/or partners’).
• Disagreements on who owns the data
that is being commercialised, either with-
in the company or between companies.
• Legal or ethical issues related to the
data.
• Unrealistic expectations for commercial
use of data, lack of the necessary
commercial models, business case
calculations don’t support data commer-
cialisation, not enough proven impact on
the bottom line.
2.2 How can companies
harness AI in business?
AI represents the next wave of technological development
following a long period during which companies (and also
public research organisations) invested in digitalisation,
analytics, computer science and data capabilities.
There are many definitions for AI, but here analysing texts and making them readable
are two simple summaries. Wikipedia40 by AI), speech recognition (computers
states that: “Artificial Intelligence is intelli- interpreting spoken language), biometrics
gence exhibited by machines.” By compari- (computers analysing human physical and
son, John McCarthy of Stanford University emotional characteristics), machine learning
stresses the human element in creating AI, (computers learning from data), computer
saying that: “AI is the science and engineer- vision (computers “seeing” objects similar
ing of making intelligent machines, especially to the way in which humans see) and smart
intelligent computer programs.” robotics (AI-augmented robots performing
advanced tasks). AI, therefore, is not just
The term artificial intelligence is typically one thing, instead it is an umbrella of various
used when a machine mimics human cogni- intelligent technologies that can work in
tive functions associated with the human a standalone capacity or be combined to
brain and does something that in the past create the solution.41
was believed to require the cognitive and
creative processes of the human mind. Appendix 2 includes a detailed review of
Improved problem-solving and learning are what AI is, why it is such a hot topic right
among the key benefits associated with now, how it is different from traditional pro-
the emergence of AI.11 Today, computers do gramming and analytics, and how machine
many things faster, more efficiently, with learning happens. Appendix 2 also briefly
greater precision and at a different order of touches on the ethics of AI, as well as the
magnitude compared with humans.10/15 impact of AI on countries’ GDPs.

When defining AI, observers typically Business leaders are currently shifting their
identify several key elements, for example: thinking from a basic “What is AI?” ques-
natural language processing (computers tion, e.g. how machines can be intelligent
interpreting and creating written natural and what types of AI there are, towards a
human language), text analysis (computers more advanced “How can I benefit from AI
Predict

A number, e.g. what will this A yes/no answer, e.g. will One from a set of options
stock price be tomorrow? this wheel fail in the next e.g. what department will
week? this call go to?
Personalise

Recommend content Target communication

e.g Netflix, Amazon, New e.g. send email campaign
York Times, Google Ads only to interested users

Recognise & identify input sources

Images Sound Text

e.g. face recognition e.g. song recognition e.g. chatbot

Uncover structures and identify patterns

Discover groups Discover anomalies

e.g. automatically detect e.g. manufacturing defects,
user segments fraud detection

Figure 9: Examples of applications for AI

in my own firm?” question, exploring the Instead they should focus on improv-
tangible business benefits of the emerging ing their activities around key value
technology and how to capture those drivers, then leverage AI to gain even
benefits in practice. bigger improvements.
• The list is by no means comprehensive,
There are various business applications of so a business leader might end up
artificial intelligence. Figure 9 highlights analysing their company and finding
some of them and divides them into four new areas of business where AI could
categories. be applied to deliver great commercial
value.
Among the applications listed in Figure 9, • Also, when choosing their commercial
prediction is already widespread in various approach, many companies use “both-
industries and can be highly beneficial. Here and” rather than “either-or”: they
are just a few examples: combine various tactics to gain maxi-
• Media companies such as The Financial mum commercial impact.8/35
Times can predict which digital pieces of • That said, it is not necessary for a
content are read or watched, and which company to implement all the tactics
digital ads are seen by audiences.42 in the list; implementing just one
• Financial services companies such as well-chosen business-critical tactic
PayPal can predict risks and identify might be enough to move the needle
fraudulent customers.43 commercially.
• In order to maximise their returns, • It is often down to data scientists
investors such as BlackRock can predict to turn business problems into data
which stocks will rise or fall or which science problems; so it may be a good
companies to invest in.44 idea to discuss value drivers, as well as
• Manufacturing firms such as ABB, the tactics listed here, with the data
meanwhile, can predict which equipment science team to help them understand
is likely to develop defects and plan the critical link between business
maintenance activities to proactively fix priorities and the data science models
problems before they occur. that they are about to build.
• All companies can predict future market
developments during turbulent times AI’s applicability depends on which sectors
and plan their responses accordingly. a business operates in. In their 2017 book
AI&U, Sharad Gandhi and Christian Ehl11
It is also possible to use AI to investigate a provide a comprehensive set of examples
business’s value drivers and try to under- of AI being harnessed in different industry
stand how to boost each of these.45 Figure sectors. One thing to note is that companies
108 presents 19 tactics that firms are taking in different industries may use different
– all of which are good starting points. terminologies and applications. In manu-
• A key point here is that firms should facturing, companies talk about “robotics”,
not just focus on “doing AI in general”. “internet of things” and “smart factory”;
 • Sell more intelligent products
Direct
to customers
revenues?
Can we use AI to • Sell AI capabilities, platforms
generate more and dashboards to customers
revenue? or third parties
Indirect
revenues? • Optimise sales, marketing and
customer service operations
• Optimise cross-selling and
up-selling, so increasing sales
volume
• Optimise customer experience,
customer acquisition and
retention
Extracting • Optimise pricing
business
• Innovate new products using
value from
data and AI
Artificial
Intelligence
• Optimise operations, processes and ways
of working, eliminating unnecessary manual
work
Can we use AI to
• Optimise sourcing, procurement, logistics
generate cost
and customer delivery, optimising the use of
savings?
materials
• Eliminate errors and defects in products and
operations

• Optimise the use of fixed capital e.g. property,

Can we use AI to premises, equipment and systems
improve capital • Optimise payables and receivables
usage? • Optimise warehousing and inventory

• Move towards data-driven management,

harness data and insights in the organisation
Can we use AI and workforce
to improve our • Identify strengths, weaknesses and risks in
organisation? the organisation and workforce
• Improve stakeholder and partnership
management

• Disrupt the industry through AI applications

Can we use AI to • Take an AI leadership position in the market-
re-position the place, create intangible value by envisioning
firm? the future
• Improve market analysis to create better
insights, stronger vision and provide more
value to customers
Figure 10: Tactics for harnessing AI: boosting
company value drivers (source: Mika Ruokonen)
in healthcare, companies talk about and has yet to make a breakthrough into
“telemedicine”, “AI doctors” and “mining of mainstream business thinking.
medical records”; in retail, the emphasis
is on “inventory management”, “consumer Key questions to ask include:
understanding” and “product search • Can we apply AI to this particular
ranking”; and in defence, the focus is often application, process and/or task here?
on “cyberwarfare” and “drone warfare”. All • Does it work and what is its impact?
these terms and applications involve AI in • Can it help us rethink our activities in
some way but, depending on the nature of that application, process and/or task?
the business and industry, how AI manifests • What are the other relevant applica-
itself might be very different (see also MIT tions, processes and/or tasks that can
Technology Review Insights35). To sum- be solved with AI?
marise, it is fair to say that: • Which set of AI technologies do we need
• In B2B industries such as manufac- to take full advantage of AI?
turing, machinery and construction,
companies typically lean towards using In their 2018 book Human + Machine,
AI to generate cost savings, diagnose, Paul R. Daugherty and H. James Wilson46
monitor, optimise and automate opera- state that the biggest business value
tions and to improve capital usage. from AI comes when machines and people
• Companies in the B2C sector such as work together and augment each other’s
retail and media typically harness AI to capabilities and strengths. In their view, AI
improve their consumer understanding should be used to perform repetitive tasks,
and demand prediction, create more analyse huge datasets and handle routine
personalised marketing/AI-enabled cases, while people should be focusing
offerings/pricing and sharpen inventory on resolving ambiguous information,
management and quality control. exercising judgement in difficult cases and
dealing with dissatisfied customers.47 They
It’s also possible to look at AI in terms of point out that, in order to reap maximum
how it can add value to different functions business value from AI, companies and
within an organisation. According to a their employees need many new skills such
recent Microsoft study41, AI is currently as the creation of training data models
most widely adopted by IT/technology or the ability to explain and sustain the AI
departments (in 47% of cases), followed system’s performance. If companies want
by R&D (36%) and customer service (24%). to take full advantage of their AI, they need
Interestingly, the study revealed that there to build symbiosis between their machines
are several functions that hardly use AI, and their employees, leading potentially to
including product management (9%), HR a virtuous cycle in which both these com-
(7%) and procurement (4%). This reveals ponents manage to continuously learn from
that in many companies, AI is still seen as a each other and improve their operations
“technology-driven” project or “R&D stuff”, over time.48
While many companies have taken signif- Finally, as AI uses data as a raw material
icant steps in applying AI to their opera- and input, challenges related to data (as
tions, certain obstacles remain, and they discussed earlier in Chapter 2.1) can also be
are not only technology-related but very an obstacle when it comes to harnessing AI.
much also business-, people- and market-
related: Each business is free to choose its strategic
• For instance, typical AI applications are approach to AI. Will it position itself as the
not error-free and often need human leading player, kickstarting transformation
supervision, especially when making in the industry? Will it be in proactive
final decisions based on the AI output. “attack mode”, seeking new AI technology
At some point in the future, when higher to harness, new products and services to
trust in AI is achieved, human supervi- launch, new markets/customers to conquer
sion might be seen as redundant and and satisfy, and new revenues to generate?
dropped, but this is not yet possible.11 Or will it intentionally choose a more passive
• There is still fear and resistance to approach, do the minimum to meet regula-
change in companies – especially if AI tory requirements, wait and see whether
might be automating and replacing the technology or the market is mature
some manual work, thereby making enough, let others take the biggest risks
some employees redundant or forcing and make the costliest mistakes, reacting
them to change their role entirely in only when challenged by competitors? Many
the organisation.49 Also, some people business leaders are still struggling to find
might be sceptical about the business answers to these questions.
potential of AI or have ethical concerns
related to how AI is used. To summarise, AI has established its
• While many companies have integrated place on the C-suite agenda with various
AI-related pilot projects into their companies talking about harnessing AI for
operations, they sometimes struggle to business purposes. Yet AI is also excessively
create a coherent AI strategy, hire or hyped up by managers. For instance, a
train the appropriate talent and scale recent pan-European study by Microsoft41
up the AI technology across functions, revealed that while 71% of companies
geographies and divisions, therefore consider AI an important topic on the
missing its full potential.35/49/50/51 executive management level, 61% are still in
• Business-related challenges such as the planning or piloting stages, and only 4%
making a business case, getting a considered themselves advanced, actively
budget approved and measuring ROI are using AI across many processes and to
real in many companies.35/51 enable complex tasks. Another study by
• Failures of initial pilot projects (maybe McKinsey50 found that, out of the surveyed
due to overambitious plans or unrealis- companies, 63% reported increased reve-
tic expectations) can give AI a bad name nue and 44% reported cost savings when
that prevents the company from moving applying AI in their business, but scaling
ahead quickly and with confidence.49/52 up the usage of AI across multiple business
units and functions is still a challenge: only
30% of surveyed companies have so far
succeeded. There still seems to be a clear
discrepancy between talk and actions: the
management ambition does not yet trans-
late into real projects, deliver tangible value
or achieve a wide-scale impact in business.

From a business leader’s perspective, it

is sensible to think of AI as a technology
that enables automation, optimisation and
efficiencies, and which can contribute to
new business and new revenue generation.
The other way to look at it is to position
both data and AI as an asset in the com-
pany’s balance sheet and sell the entire
company to a customer who desperately
needs data and AI. In these scenarios, data
and AI assets are considered valuable,
because they can help the customer to
grow or improve their business. Data and
AI are sold as a full package, as part of the
rest of the company’s assets. Deals like
this have already been made, for instance,
Facebook paid WhatsApp $22bn in 2014
due to its data on 600 million consumers
worldwide, and Microsoft paid $2.5bn for
Minecraft. A key element in these very high
valuations was the consumer data, which
had the potential to help the acquiring
firms to target new customer segments.
In addition, it also helped them to develop
their machine learning algorithms as these
acquisitions provided valuable input and
training data.
2.3 Designing valuable data
and AI solutions
If data and AI assets are to deliver any value to customers,
the company selling them must identify real problems that
are worth solving (or jobs to be done9) so that the assets are
1/45/48

solving is not always easy (see also Johnson9). Questions to

ask include:
• Is there really a problem that has and the desired impact and value of the
strategic importance? Who exactly has models.
that problem?
• What is the root cause of the problem? We should assume that the problems
• Can data and AI assets solve the prob- worth solving are fundamental to all data
lem, if they are delivered and priced in a and AI-enabled business models. It’s only
meaningful way? once the problems worth solving with AI
• How satisfied are the customers with have been identified and described, that
current solutions that do not neces- a company can move towards building
sarily have much to do with data and new, crystal-clear data and AI-driven
AI-enabled offerings? value propositions for customers and the
• How well do the new data and AI-enabled revenue models which support them. In
offerings solve the problems of these fact, customers’ willingness to pay for the
potential customers, relative to the resulting solution is a key test of whether
other options that the customers might companies have correctly identified a
have? problem worth solving.
• And, apart from direct commercial
returns, how could the data and AI-en- Furthermore, we should assume that there
abled offerings potentially also deliver might be several target groups for the com-
benefits to society and its various pany to approach with data and AI commer-
stakeholders? cialisation in mind, and several “customer
personas” that might need to be described
Businesses selling data and AI assets need and understood. Customer selection is an
to ask questions to understand the nature essential element in both finding problems
and size of the business opportunity worth solving and targeting data and
AI-enabled offerings and revenue models There are three best practices in discov-
at them. After all, the potential customers ering data and AI-related problems and
for data and AI product/service offerings designing solutions to them, as follows:
might not just be restricted to a company’s • Being curious: The designer needs to
current customers but might also exist keep an open mind and ask relevant
elsewhere in the company’s value chain (e.g. people to explain things to her/him as
suppliers, partners, end-customers) or in part of the design process. For instance,
other industry segments altogether. The potential customers could talk through
better a business understands its potential their daily life, key activities and com-
customers, the easier it is to create and mon pain points, and a data scientist
implement rich and unique datasets or could explain the impact of data and AI
sophisticated algorithms that deliver in solving the pain points.
commercial value. • Being aware of their own background
and biases: The designer must under-
Customer case studies are useful in defining stand how he/she sees the world and
customer problems that might be solved what potential impact their attitudes,
using data and AI. Well-written case studies prejudices and world views might have
can articulate the data and AI-related on the product or service that is being
business idea, the potential target group, designed.
as well as the tangible change in the target • Being aware of potential issues: If
group’s behaviour that might happen as a there are any potential areas of conflict,
result of deploying data and AI. the best practice is to involve key stake-
holders early in the design process.
For example: “Our suppliers seem to be struggling These stakeholders could be people with
to deliver products/spare parts to us because they privacy concerns or people whose jobs
lack the necessary visibility into our operations to or tasks might be under threat once the
estimate what we need and when. If we provide them data and AI solution is implemented.
data on our operations, they will learn to predict our
needs better and schedule their deliveries better. We In addition, another best practice is to
could use our operations data as a tradable asset: we work with domain specialists: the “best
can ask suppliers to give price discounts in return for problems to be solved” and “best solutions
the valuable data they receive from us.” to be designed” are often identified through
teamwork and by combining people from
- different functions.45/48/53 It is important not
ability. If we could use our customer data to measure to underestimate the role of great product
and predict any unnecessary costs and spending, we design: it is not only the algorithms that
could potentially help them to save money and so matter but also how usable the end prod-
uct/service is.54 In order to get maximum
a new consulting business that helps our customers impact out of the design process for new
data and AI-enabled business models, a
competition in this space currently.” company should dedicate a small, focused
team55 to the work and have at least the aligning with enterprise architecture,
following roles taking part and collaborat- helping people understand what is
ing (see also Workera56): technically possible.
• Business owners: Setting the vision
and strategy, helping to generate buy-in Typically, the design process for data and
from the organisation, defining business AI-enabled business models includes a
goals and KPIs for the solutions, bring- lot of testing of hypotheses and hacking
ing in deep industry- and domain-spe- something together (e.g. early prototypes)
cific knowledge, turning emerging ideas in only a matter of days or weeks.6/7/9/31/53
and market insights into concrete value Canvases are often valuable tools to get
propositions and business models, people to talk about their problems and
defining what success looks like. challenges and to ideate new solutions
• Data scientists: Formulating the task so (we return to this topic in Chapter 3.2).
that machine learning methods can be An important rule of thumb is to “get out
applied, defining data science problems of the building”: in other words, problems
and performance metrics, doing data worth solving are hard to identify “in the
engineering and data modelling, select- laboratory”, instead they typically can be
ing correct machine learning models and revealed “in the field” by observing people
algorithms and training methods, con- and their daily work.
tinually redesigning and tuning those
models and methods, doing feature The double diamond model originally
engineering, interpreting, explaining developed by the British Design Council57
and visualising results, participating is undoubtedly one of the most well-known
in the larger data science community methods available for discovering and
to learn from other experts about the defining problems worth solving and deliv-
latest methods and tools. ering solutions to them. Figure 11 illustrates
• Designers: Studying customer personas the model (this one is a version that has
and their needs, collecting customer been created by Futurice, but it is based on
insights, identifying problems worth the Design Council original model). The four
solving, designing data and AI-enabled steps in the double diamond are as follows
solutions including their user interfaces, (see also Costa58):
pitching new product/service offering • Discover: Initially the goal is to under-
and revenue model ideas, validating the stand, rather than simply assume, what
solutions to make sure they address a the problem is. This phase is all about
real human need. data gathering and it involves speaking
• Software and IT engineers: Collecting to and spending time with people who
and consolidating data, setting up data are affected by the problem.
storage and software infrastructure, • Define: Next, the problem worth solving
building interfaces between data and and its root causes are defined and
AI models and other relevant software, illustrated, based on the insights and
deploying machine learning models, observations gathered in the first
 discovery phase: they are filtered and Designing valuable data and AI solutions
assessed in this phase. means that the company selling them is
• Develop: In the second diamond, able to find real problems worth solving for
people provide different answers to the the customers; dedicate a small, focused
problem worth solving. Inspiration is and multifunctional team to the work, stay
sought from the multidisciplinary team curious, unbiased and inclusive; and follow a
or from elsewhere in the company, or rigorous design methodology that includes
from society at large: imagination plays observations, exploration, prototyping and
a key role in this phase. validation.
• Deliver: Finally, different solutions are
prototyped and tested out at small

scale: those that don’t work are rejected

and those that do work are improved
and refined.

1 3
General Problem Solid
Discover Develop
problem worth concept
& define & deliver
definition solving vision

2 4

1. Confirming business objectives, participants 3. Co-creating service concepts to match the

and schedules for the work. Understanding customer needs and applications. Creating
and exploring end user groups, their needs initial mock-ups of service concepts and
and their business perspective. Workshop and validating concepts with end users. Co-creation
interviews. workshops and validation interviews with the
end users.
2. Mapping end user personas and customer
journeys to deepen the understanding of most 4. Summarising to create a coherent documenta-
viable applications. Analysing and prioritising tion of customers, applications and proposed
identified applications. Follow-up interviews if service concepts. Creating a plan for the next
needed. steps to set priorities and timelines for solution
development.

Figure 11: The double diamond model for defining problems and designing solutions
2.4 A customer’s perspective
So far, we have discussed data and AI from better decisions on which companies to
the perspective of businesses selling data invest in.
and AI-enabled solutions, including how • Retailers might be interested in buying
they can identify problems worth solving in data that helps them to predict future
the customer’s daily life and operations, and demand for products, helping them to
solve those with these assets. But there is stock accordingly.
also the customer’s perspective to consider.
The customer might be an existing one, or Companies wishing to do more targeted
alternatively a new one, or stakeholders marketing and sales:
who are potential targets for data and • For instance, companies aim to target
AI-enabled product/service offerings and their advertisements to specific cus-
revenue models. There are several reasons tomers or reach the most interested
why companies might be interested in and responsive customer groups for
buying data and AI assets: their products.36/59
• Retailers and brands might buy data e.g.
Companies wishing to improve their on website visits to understand consum-
decision-making: er buying intent and preferences, then
• This includes making more impactful steer their online advertising spending
decisions (e.g. measuring conditions towards those high-potential customers.
before and after the decision) more • Industrial companies, on the other
timely decisions (e.g. generating im- hand, might buy data on their large B2B
mediate reactions to real-time data) or customers to understand who the key
more granular decisions (e.g. developing decision-makers in those organisations
localised or personalised pricing). are, so that they can target their adver-
• Investors are often interested in buying tising spending to those individuals.
various data sources in order to make
Companies wishing to feed and train their viders, evaluating and comparing them
algorithms with acquired data: and trying to work out where and how
• As AI is impossible to develop without to get a data and/or AI solution that fits
large quantities of data, companies with its needs. So, the customer thinks:
need to ensure they have adequate “Where can I get these assets? Which provider
datasets in their possession if they might provide a relevant, purposeful, reliable
intend to develop sophisticated algo-
rithms that are useful. the existing data and AI assets and the ones I’m
• Algorithm developers might turn intending to buy to create something new and
towards Google-owned data science valuable?” Businesses selling data and
community Kaggle and translation AI solutions need to be able to answer
services provider Lionbridge, which questions like these.
provide datasets that are targeted to • The customer is not yet aware of the
this purpose. potential data and AI-enabled solutions
• These customers might also benefit or business opportunities that exist.
from government-related datasets such They might not yet understand what
as EU open data portal, or datasets data and algorithms are available
provided by universities. overall, how they could solve issues or
how they could create new value. This
Companies buy algorithms to speed up might be due to a lack of information,
their AI development work: know-how or imagination, or all three.
• Algorithmia Inc., a US-based startup Businesses selling data and AI solutions
provides an online marketplace that need to work hard to build awareness
connects computer science research- of these solutions, convince customers
ers’ algorithms with developers who that they are reliable and credible, and
may have uses for them.60 also introduce potential new ways to
• Going forward, customers may be able buy and pay for them.
to piece together various algorithms,
from various sources, and create From the customer’s perspective, the data
entirely new software applications from and AI assets aren’t just “nice to have,” they
these components,61 more quickly than need to serve a purpose. It is likely that the
starting from scratch. customer wishes to change their behaviour
in some way, by harnessing these assets.
So, what is the customer’s perspective on They might want to do certain things
data and AI? There are at least two plausi- smarter, faster, cheaper or more efficiently,
ble positions: or avoid doing something altogether. So, if
• The customer has identified a problem there is a meaningful change in behaviour,
and is actively looking for data and there is likely new added value as well.
AI-enabled offerings to solve it. In this
case, the customer might already be Take Uber, for instance. It provides a data
collecting information on potential pro- and AI-optimised platform that enables
people to move quickly from one place to
another, avoiding the use of more expensive
taxis or potentially even owning a car, and
it also provides a new source of income
for drivers. Given that the use of Uber has
spread quickly, the company has clearly
turned location information and route
optimisation into value that customers are
willing to pay for (and into work that drivers
are willing to do for money).

So, the customer’s perspective is to gain

value by doing certain things better or
faster or smarter. The customer might or
might not be aware of their needs or the
data and AI assets that are available to solve
their problems. The deeper the company
selling AI and data solutions can dig into the
customer’s motives, drivers, challenges and
aspirations, the better positioned it is to
provide valuable solutions.
2.5 Conclusions on data and
AI as enablers of business
As we have seen, the business opportunities related to data
and AI are huge and expected to keep on growing in the
future.11 They can be related to boosting a company’s existing
business or to creating completely new product/service
49

There are still significant obstacles that direction of various industries and business
prevent companies taking full advantage of sectors in the future.11 Currently the best
their data and AI assets: notably, the tech- algorithms in the world are already truly AI,
nology is far from ready, there may be legal however the development of algorithmic
or ethical issues, resistance to change, or technology should be seen as a continuum
inability in companies to scale new ventures in which many algorithmic systems con-
towards full-blown businesses. But we can stantly improve and become increasingly
assume that many of these obstacles will automated,39 steadily reaching a higher level
be overcome in one way or another in the of performance and sophistication.
coming years, through innovation activities
in R&D labs, facilitative regulation by gov- In order for companies to deliver tangible
ernments or by company leaders improving value as a result of their data and AI, they
their skills. It is fair to say that data and need to uncover real problems worth
AI-enabled businesses are likely to prolifer- solving and design products/services
ate and play a bigger role in the strategic and revenue models that address those
problems, then apply metrics that measure
and prove the value created. Starting small
often makes sense, as does co-creation
with target customers. Finally, successful
data and AI-related new ventures require
the creation of a multifunctional team that
includes data scientists, service designers
and business people.

After reviewing the role of data and AI

in today’s business, we now explore the
concept of a business model to see which
existing paradigms apply to data and AI.
3 Business model
innovation as a
source for growth
In this chapter, we will do two things. Firstly, we will recap
what has been written about business models and how
the concept of a business model is regarded as a tool for
companies to plan their operations and ventures. Secondly,
we will discuss the nature of business model innovation:
how companies can plan, validate and execute new business

A key conclusion of this chapter is that,

when it comes to data and AI assets, various
business model innovations and configura-
tions are possible in firms and that there is
no single business model “silver bullet”.
3.1 What is a business
model?
The business model as a concept gained The New New Thing, Michael Lewis63
momentum in the 1990s, largely as a result of referred to the business model concept
the emergence of the internet – which made as “a term of art, and like art itself, it’s one of
new digital business models possible.62 Back those things many people feel they can recognise
then (and to some degree today), internet when they see it (especially a particularly clever
startups were seen as radically new ventures ”.
that challenged incumbents in various indus- • Joan Magretta , meanwhile, emphasised
62

tries, and the question of business model business models as narratives by firms
selection became topical in nearly all firms. about their customers, money-making
One could argue that the concept of business and logic; stating that business models
models was an attempt by academics and “are stories that explain how enterprises work”.
practitioners to understand and explain new • Management guru Peter Drucker64 de-
disruptive dynamics in the business world, scribed business models as “assumptions
or to try to protect pre-existing companies about what a company gets paid for”, and in
from them altogether. cases where the assumptions are wrong,
outdated or no longer fit with reality, the
Yet, business leaders often still struggle to company might stagnate and/or hit a
speak the same language and lack a shared crisis.
vocabulary when they talk about business
models. As a result, definitions of business As we see, these definitions are quite
models are plentiful: vague. Simply put, it looks as though
• Some have seen business models as a business models typically answer questions
kind of a mystery: in his 1999 book such as “how do you plan to make money?”,
and “is your money-making plan still valid, given
your customers and your market?”. In short,
a business model is about successfully
establishing and executing a resilient
and comprehensive business system that
creates, delivers and captures value for
customers.

Appendix 3 includes a review of interesting

business models and business model inno-
vation frameworks that are also applicable
to a company’s data and AI assets.

People often conflate business models with

other concepts such as vision, strategy,
positioning, resource allocation and the
firm’s value chain, which risks distorting
and misusing the term.62 While no one is
obliged to stick with a strict or sometimes
artificial definition of the business model
concept, a commonly established vocab-
ulary and definition wouldn’t do any harm
(as Magretta states, no organisation can
afford fuzzy thinking on concepts that are
fundamental to performance). Our contri-
bution to clarifying business models comes
in Chapter 5, where we try to summarise
new business models, particularly product/
service offerings and revenue models, from
the perspective of data and AI.
3.2 The Business Model
Canvas
The Business Model Canvas by Osterwalder and Pigneur19
provides a comprehensive visualisation. In recent years, it
has become the de facto tool for managers and consultants,
as it allows any business model to be presented in one easy-
to-grasp sheet. Yet the canvas is quite broad (as one-sheet
visualisations typically are). So it is probably best seen as a
snapshot of the business model at one point in time,20 rather
than something that captures the evolution of a business
model over the course of time.

Osterwalder and Pigneur stated that a large selection (let alone all) of them would
business model is “the rationale of how this probably require a book in its own right.
business creates, delivers and captures value”. Their
Business Model Canvas, developed with 470 In the Business Model Canvas, the value
co-creators, is “an organised way to lay out your propositions box describes what makes a
assumptions about not only the key resources, key company special in the market and what
activities and key partners of your value chain, but value the company brings to its customers;
also your value proposition, customer relationships, in other words, those things that the
channels, customer segments, cost structures and customers are willing to pay for. This is
revenue streams – to see if you’ve missed anything typically a set of products and services
important and to compare your model to others”. that create value for a specific customer
segment. The revenue streams box, as its
Figure 12 (Osterwalder and Pigneur19) shows name suggests, describes how revenue is
the Business Model Canvas, which com- generated, what revenue models are being
prises nine boxes. As there are numerous used, how customers would prefer to pay,
industry segments, as well as untapped and how money flows into the company.
spaces between traditional industry bound-
aries, there are many different possible In this book, these two boxes – the value
business model configurations.65 Listing a propositions box and the revenue streams
box – are essential. Later, we will discuss
 Key Part- Key Activi- Value Propo- Customer Customer
ners ties sitions Relationships Segments

Key Re- Channels

sources

Cost Structure Revenue Streams

Figure 12: The Business Model Canvas

data and AI-enabled business models from will also play a role in delivering data and AI
the perspective of “what is being sold” solutions, provided that the people working
(products or services) and “how money is there have been taught to sell and support
made” (revenue models). When it comes data and AI-enabled product/service
to thinking about these two boxes, under- offerings.
standing the “right side” of the canvas
(the “value side”) is key: namely, who are The Business Model Canvas is a rational
the customers of the data and AI product/ systemic set of assumptions and choices.
service offerings? What relationship does Therefore, “there’s no right side without
a company have with them? And how are the left side”:
the product/service offerings delivered to • Activities, resources and partners (the
them? As we will see later, various target cost side) are fundamental to building
markets, customers and revenue models new product/service offerings and new
might apply. We expect that digital channels revenue models. Meanwhile the cost
such as websites, apps, dashboards, plat- structure sets boundaries for pricing
forms, documents and software systems and revenue model implementation.
are likely to be a common way of delivering • The more a company innovates on the
data and AI product/service offerings to left side of the canvas, the more free-
customers, while physical channels such as dom it has to implement business model
field sales or customer service departments options on the right side.
• Or, the better a company can predict tially new and better one: can the company
how its cost structure, resource alloca- be sure that the new model really delivers
tion or partnerships will develop in the better financial results in both the short
future (using data and AI, of course), the and the long-term? Thoroughly validating
more options it might have in price-set- new business models with customers is a
ting, for instance. good idea here.
• Consequently, if a company is able to
build new business models using data In this book we adopt a strict focus and
and AI, it can increase its probability of intentionally “only” tackle “what is being
success by reviewing all the nine boxes sold” (the product/service offerings) and
of the Business Model Canvas, through “how money is being made” (the revenue
the lens of data and AI, to see where models) – in the context of the firm’s data
opportunities for improvement lie. and AI assets. Throughout, we attempt to
look at these two aspects from a customer
Given the potential for wide variations value perspective and through understand-
among business models, it is important ing real customer problems worth solving.
to evaluate and compare various models It is worth noting that there is a plethora
(or configurations) to determine, once of written resources that tackle data and
implemented, which are better. An obvious AI-related business model renewal in the
criterion for comparison is the one that other areas of business model canvases
delivers the best financial results. But the (e.g. activities, resources, processes and
reality may be more complex than that: costs).
• While some models will deliver immedi-
ate revenue and profit, they might not To summarise, canvases and boxes are use-
be resilient and durable in the long- ful for generating new data and AI-enabled
term, e.g. they might be easily copied by business models. It’s easy for an innovative
competitors (in which case, timeline and team to gather around a large canvas,
vulnerability issues matter). generate ideas, use colourful Post-it notes
• What’s more, while some models might to list and categorise them, and overall to
work immediately for some customers, approach new business design in a structur-
they might be very difficult to scale up al, systemic and holistic manner. In Chapter
to cater for global markets (in which 6, we suggest and field-test a focused set of
case, scalability matters). four canvases that are designed to uncover
new business opportunities in data and
It’s important to view new business models AI product/service offerings and revenue
not in terms of what they deliver immedi- models.
ately, but instead in terms of what kind and
size of business opportunities they enable
in the future. This is especially important
if the company in question wants to switch
from an existing business model to a poten-
3.3 Business model
innovation
Business model innovation centres on a company proactively
striving to change its business model in order to reach new
market opportunities or to challenge the current competition.
As the world of technology accelerates, innovation will be the
54
An increasingly
popular way to make change happen is to work iteratively
and continuously, including plenty of trial and error. Given
the current uncertainty in the world, the capability to
innovate, validate, launch and scale new business models is a
critical core competence for all businesses.
As an example, Figure 13 represents the enough. Instead, new business models are
transformation that is currently taking needed,66 and failure to reap the benefits
place in the manufacturing and machinery of data and AI might prove fatal. The good
industry, e.g. in companies such as the news is that, for many companies, the
US-based Caterpillar, Japan-based Komatsu opportunities are greater when data and AI
and Swiss-based Liebherr. In this scenario, are included.
expertise in traditional hardware such
as machines and equipment is no longer
 • Value creation shifting in-
creasingly towards software,
Ecosystems, services & data
services and ecosystems.
• Competition intensifying
here, too.
• Failure in moving towards
software and services busi-
ness and ways of working
Connected might prove fatal.
VALUE

& intelligent
hardware

Traditional hardware

PAST FUTURE

• Constant increase of competition in traditional hardware production.

• Pressure coming from low-cost countries, pushing prices of “physical”
products and value of assets down.
• Niche and top-quality products best protected, but not safe from value shift.

Figure 13: Example of a transformation towards new business models (source: Mika Ruokonen)

Business model innovation may be sparked • Parallel (e.g. a new business that
by “technology-push”: the emergence of co-exists with the existing one without
new technology that enables the company having major links with it), or
to change how it operates its business (in • A replacement (e.g. the introduction of
our case, emergence of data and AI). Or a new disruptive business model that
alternatively it may be caused by “mar- cannibalises the existing one).
ket-pull”: changes in the customer needs
or market environment that encourage the Depending on the situation, a company
company to address the new conditions in needs to find meaningful ways to push
an innovative way. In some cases, these two ahead with the transformation. In an incre-
may happen simultaneously. mental change, the transformation might
not be huge, whereas the replacement of
Furthermore, business model innovation20 a pre-existing business model could be a
can be: major undertaking that challenges a firm’s
• Incremental (e.g. a small change in one very existence.
of the business model elements)
• An extension (e.g. a new business prac- Just 10% or less of innovation investments
tice to complement the existing one) at large companies tend to be focused
on developing completely new business recently in the music industry, complement-
models9, with the remainder of the in- ing its music streaming services with direct
vestments going to strengthen existing licensing deals with artists). These moves
business models instead. This may explain by Netflix disrupted the entire film industry,
why so many radically new business model and later other firms such as Amazon, Disney
implementation projects fail: the amount and Apple copied Netflix’s video streaming
of experimentation with them is just too business model. Nowadays, people do not
low, accounting for only a fraction of all the necessarily need to buy, own or rent DVDs:
innovation activities in many large firms. instead they can watch content online
whenever and wherever they want, against a
But radical business model renewal is fixed monthly subscription fee.
possible. Netflix, as an example, started in
1997 as a DVD rental company by mail but In other words, the relationship between
then transformed itself twice: first into an the customer and the product/service
online video streaming service in 2007 and offering (the video content) has changed
later starting its own film content creation in dramatically. In theory, Netflix could have
2012 (Spotify made a similar disruptive move remained as a DVD rental company, but it

Business model innovation

Value Value Value

creation proposition capture
innovation innovation innovation

New capabilities New product/ New cost

service offerings* structures
New technology
New customers New revenue
New partnerships models*
New channels
New processes
New customer
relationships

Figure 14: Business model innovation elements *) Focus areas in this book
realised early on that consumer behaviour Figure 14 identifies the common business
was shifting towards heavy use of content, model innovation elements that a com-
accessibility and comfort, as well as afford- pany can change and pivot as part of its
able pricing. At the same time, the emer- innovation activity.67 Here, we organise
gence of new digital technology was making the different elements – which resemble
new online viewing solutions feasible. There those in the earlier Business Model Canvas
were various important market trends – based on their impact on value creation,
that enabled this dramatic business model value proposition and value capture. The
change. focus areas of this book, i.e. “what is being
sold” and “how money is made” represent
Netflix’s transformation was not entirely only two out of 10 of the potential areas for
due to data and AI, but the addition of innovation, yet their importance for data
these two elements undoubtedly helped and AI-enabled business model change
the company to provide personalised cannot be underestimated.
viewing experiences and optimise its video
content to match the needs of audiences.
What’s more, while Netflix was not a
global corporation going through a major
transformation (it was relatively small
with 1,500 employees when it launched its
online streaming model), it nevertheless
revamped its model thoroughly and inspired
other companies to consider renewing their
business models too.

Netflix is not alone. Any company can

proactively innovate new business models
and radically change its model if it really
wants to. And, it’s important to note that
business model innovation never happens in
a vacuum. Instead it almost always relates
to important market developments such
as the emergence of technologies, legal
moves, changes in consumers’ habits, or
willingness and ability to pay. A successful
business model innovator often spots these
developments, optimises the timing of new
business model launches and then executes
iteratively and rapidly.
3.4 Business model
innovation and company
size
When it comes to business model innovation, the new models (“how could we implement this
size matters. Large companies have the new model to cover the largest part of the market
processes and resources required to drive possible?”). They often lack the necessary
business model innovation, and they are used resources, processes and ways of working
to running several business units in parallel. to push the new (often disruptive) business
However, they can face challenges when it models to market, especially if their goal is
comes to the need for entrepreneurial skills to become widely adopted and dominant
(“who will drive the business model renewal project in models in global markets. So, while business
our company?”) and organisational preference model innovation in large companies might
for the status quo (“how can we successfully be a mindset and capability issue, in smaller
drive the shift towards a new business model if the companies it is typically a resource issue.
existing model still functions pretty well?”). The
more radical a new business model, the more When it comes to where new business
difficult it is typically for large companies to models sit within organisations, very few
grasp and implement. It’s common for large small companies consider placing their
companies to only change their business disruptive new business models to the side
model if there is no alternative.68 of the core business, because in their setup,
typically “everyone needs to do everything”
Small companies, on the other hand, often and various roles mix to form dynamic teams.
have an inherent ability to continuously In addition, the startup’s business model
experiment with their business models and might be disruptive by definition, so there
prepare themselves for possible changes.68 is no legacy business to separate from. For
For these firms, business model innovation large companies, by contrast, positioning the
activity resembles the “lean startup” ap- disruptive new business models to one side
proach,69 in which they build, measure and of the main business represents more of an
learn iteratively and pivot their business opportunity. These companies might wish to
models whenever necessary. They are give their new business models autonomy,
used to building hypotheses, testing and including leadership teams, governance
validating them with customers, and building mechanisms and management, planning and
new product/service offerings and business budgeting practices that differ significantly
models as efficiently as possible. For small from the core business.6/70
companies, the challenge lies in scaling
There might be dual structures (hence the
term “ambidexterity”: Osterwalder and
Pigneur,19 Rissanen and Karhu71) in which
large companies separate legacy business
model exploitation and new business model
experimentation into different units (e.g. put-
ting the latter in R&D units or innovation labs
or separate legal entities), trying to balance
investments into the two and increase their
chances of survival in rapidly changing busi-
ness environments.7 Inertia might force the
company towards rigid thinking and actions
rather than being open to new and disruptive
ideas, making business model innovation
difficult. Also, the existing business might
“suck the oxygen out” of the new disruptive
business models when they need attention
and investment. It is not essential for a large
company to separate new business model
innovations from the existing business, but
many large companies choose to do so. The
more radical the new models, the bigger the
temptation to separate them and “set them
free”.72/73
3.5 Conclusions on business
models as a source for
growth
At this juncture, the reader might ask why it was necessary
to describe business models. A key takeaway is that a
company’s business model is the foundation on which data

Understanding business models in general helps to identify

what adds value and how to innovate new business.
Indeed, if the data and AI-enabled new of- not automatically create a new model,9 if
ferings and revenue models are to succeed, there is no real change in how the product/
they must be valuable to the customers and service offering and revenue model is
innovative enough to inspire people to work designed and executed (more on this topic
on them. Our intention in this book is not to in Chapters 4 and 5).
delve too deeply into business models. But
the central proposition should be clear to Given the current economic uncertainty and
all: business model innovation is an im- changes in the market, all companies need
portant method for companies to generate to think carefully about where the oppor-
organic growth, while simultaneously taking tunities of data and AI-enabled business
their business to a new level. models lie. They also need to consider how
to test new models in a smart way, and
Innovations within data and AI-enabled how to implement the resulting changes
business models are not significantly dif- in a real-world context. In the future, if AI
ferent from other forms of business model becomes ubiquitous, a kind of “new electric-
innovation. There is no single business ity”, then an almost infinite number of ways
model “silver bullet” that can be used with for extracting economic value will present
all data and AI assets. In fact, it is totally the themselves.11 Depending on which data and
opposite, because numerous configurations AI-enabled business model a company opts
are possible.31 In addition, it is important to for, the change can be either complicated
remember that simply adding data and AI and painful (for instance, using data and AI
components to an existing business does to enter completely new markets) or easy
and rewarding (for example, selling existing When thinking about business model inno-
data and AI assets as part of an enhanced vation, a company needs to keep its eyes
product/service offering to existing cus- open and anticipate potential changes in its
tomers). In Chapter 5, we illustrate this point business landscape – especially changes that
by cataloguing both complex and simple are enabled by data and AI. Often, instead of
AI-enabled business models. “me alone” it might be “us together”.

When discussing innovation in new data and Scaling-up of new data and AI-enabled
AI-enabled business models, it is important business models is, of course, an issue.49 A
to understand not only the business itself company might have an innovation process
but the surrounding business landscape in or accelerator programme in place that
which it operates:55 provides a structured and repeatable
• What if a customer suddenly reinvents formula and process for innovating, testing,
its business with data and AI and starts validating, iterating, launching and scaling
to serve entirely new markets or custom- up new ventures.55 In that sense, a business
ers? model innovation project that builds on data
• What if a supplier embraces the use and AI could often be treated like any other
of data and AI and starts to bypass innovation project in the firm, requiring
the company in the value chain, selling a team, an executive sponsor, adequate
directly to end-customers? funding and KPIs that are followed in order
to track progress. For the purpose of new
What are the implications for a firm’s innovative business models, Johnson9
business performance in these cases? These suggests that it might be worth identifying a
examples are not unheard of; the company first foothold market in which to test models
therefore needs to be ready to use its own safely and with low costs. For instance, a
data and AI assets to react, perhaps also small geographic region or customer group,
redefining its own role in the business preferably on familiar or otherwise friendly
landscape: territory. If the first incubation is successful,
• Which direction could a data and AI-in- the company can proceed to:
fused disruption come from, and when? • An acceleration phase in which the
• Could the firm support its customers venture becomes more formalised,
better if they already benefit from data embracing greater structure and rigour,
and AI? introducing more people into the team
• Could it support suppliers better in their and being exposed to a larger target
data and AI efforts, or perhaps make an market.
alliance with them to serve the market • A transition stage in which the venture
together? becomes “business as usual”, is often
• Finally, could the firm use data and AI to integrated in the core business, and
claim a new position in the value chain? through training and board level spon-
sorship becomes a core organisational
capability across the entire firm.
If a company takes its data and AI business
model efforts seriously, these different Next, we leave business model for a while
phases of new venture scaling should be and return to the core focus of the book:
considered right from the outset. They might how data and AI can change “what is being
then help the company to move beyond sold” and “how money is made” in com-
piloting – so unleashing data and AI’s consid- panies. We will therefore concentrate our
erable business potential. By contrast, if the thoughts on data and AI-enabled product/
data and AI business model pilot projects are service offerings and revenue models;
not structured so they can be scaled up, they exploring how they can help companies
might remain just a curiosity; not changing generate growth for their business.
the strategic scope and direction of a firm,
not impacting the bottom line enough, not
getting enough managerial attention, and
eventually failing or fading away.31
4 The foundation
of data and AI-

Given rapid advances in data and AI tech- might bring a significant competitive
nologies and the consequent emergence edge, allowing the business to position
of new technology-driven opportunities, itself as a company that can truly
companies are increasingly exploring how harness data and AI for the benefit of
they might renew their business models its customers – something that other
using data and AI. The anticipated benefits businesses might not yet be capable of
of data and AI-enabled business models doing.
include: • Implementation of new models might
• AI and data business models that are bring resiliency to the business portfolio
scalable, stable and recurring and in the wake of Covid-19.
not dependent on one-off deals or
individuals. There is a wide variety of options for a
• Many AI and data business models for a company to consider when renewing
have transformative potential, allowing its business models with the help of data
access to totally new customer groups, and AI. In order to extract the full benefit
new business opportunities, new and impact, a company will need a lot of
channels to market.A fresh thinking, an innovative approach and
• First to market: The implementation courage. As stated previously, our primary
of new AI and data business models focus in this book is around “what is being
sold” (the product/service offerings) and
“how money is made” (the revenue models).

This chapter sets the scene for data and

AI-enabled business models. It will explore
four elements that are the foundation of
the data and AI-enabled offerings and
revenue models to be discussed later. It will
investigate the fundamental issues that a
company typically needs to think about and
understand before making actual choices
on what to sell in the future, and how to
monetise it. The four elements are:
• Setting the ambition for data and AI-
enabled offerings and revenue models
(in other words, looking for a consensus
in the firm about the aims and how to
achieve them).
• Understanding the maturity of the
company when it comes to implementing
data and AI in product/service offerings
and revenue models (as the maturity
sets important constraints on the firm’s
development and implementation).
• Understanding the sector in which
the company operates (because
the business characteristics, again,
impose important constraints on the
development and implementation of the
data and AI-enabled product/service
offerings and revenue models).
• Understanding how to make the new
data and AI-enabled offerings and
revenue models unique and game-
changing within a given sector (this
is vital because the more unique the
concept is, the easier it is to create and
sustain a profitable business from).

The rest of Chapter 4 explores these four

elements in more detail:
4.1

Companies wishing to renew business models analysed and “productised” before being
with data and AI need to decide a number of offered to players outside the firm? With
things: what is our ambition level? What do the data and AI assets available, should the
we want to achieve? Are we satisfied with firm move from being a product business
“augmenting” or “boosting” our existing to a service business – or vice- versa – or is
business models and product/service the status quo a better bet? Questions like
offerings with data and AI (note: nothing wrong these set the scene for harnessing data and
with that!)? Or do we wish to explore horizons AI and have big implications when it comes to
and areas beyond our current business, e.g. via deciding strategy and business models.
new markets, new customer segments or new
product/service offerings?27/28 Figure 15 describes a “data and AI strategy
compass” for companies to consider. It
Are existing customers the target for new offers various paths that may be chosen
business models or does the firm want to start one-by-one or in combination (e.g. both a
hunting for and approaching completely new new product/service offering and a new
customers? Alternatively, does the firm want revenue model).
to attack a totally new industry segment by
harnessing its data and AI assets – which then An alternative way to review data
brings with it completely new customers and opportunities is the well-established Ansoff
competitive opportunities?7/27/28 Matrix: a two-by-two matrix that brings
together 1) markets and 2) product/service
Is the existing data valuable and ready to offerings, but doesn’t include much about
be sold, or does it need to be processed, the revenue models, which are obviously
 New product/
service offering?

New revenue model?

Current customers /
market /
value chain?
Data and
AI-enabled
business
models

Current New customers /

revenue model? market / value chain?

Current product/
service offering?

Figure 15: A strategy compass for data and AI-enabled business models (source: Mika Ruo-
konen)

one other part of the equation. Figure be revolutionary. The inevitable question,
16 illustrates this well (original Ansoff,74 however, is whether this incremental
modified by Mika Ruokonen). improvement is enough in the long-run. It
is possible that rival companies investing
Discussions with business leaders in several in data and AI might succeed with a more
companies working on their data and AI disruptive approach, and in turn threaten
assets reveal that few are currently ready risk-averse companies at some point in the
to take a radical approach. Rather, for many future.
the best and easiest approach is to start
with “obvious” opportunities, with quick If a company decides to raise its ambition
wins, or by boosting existing businesses. level by seeking totally new data and AI-
This might mean targeting one specific AI enabled product/service offerings and
technology – be it image recognition, robotic opportunities, it taps into a whole new
process automation, speech recognition universe of interesting problems worth
or such – then applying it to a certain core solving, and significantly greater business
process of the firm to gain improvements. potential.5/7/75/76 Deciding to take the more
The end game here is to generate value for ambitious route creates the opportunity to
existing customers in existing markets, via differentiate significantly from competitors,
an established product/service offering to raise competitiveness to a new level,
and revenue model. In many industries, this and claim a completely new position in the
approach is already helping management industry value chain. Ultimately, this opens
teams to make smarter decisions and to the door to a significant increase in value.
deliver competitive advantage. A data and Inevitably, the risk of failure increases
AI-enabled firm – operating within the as well: the further the company moves
existing business framework – is still a from its traditional core business, the
data and AI-enabled firm. That alone can more uncertainties there are, e.g. from
 Markets

NEW
Market development Diversification
Seeking completely new Seeking completely new
customers for existing customers for new data
data and AI-enabled and AI-enabled product/
product/service offerings service offerings and
and revenue models revenue models

Market penetration Product development

Providing data and Creating radically new
AI-enabled product/ser- data and AI-enabled
vice offerings and revenue product/service offerings
models to existing cus- and revenue models that
tomers, as part of existing are provided to existing
products and services customers
EXISTING

EXISTING NEW

Products and services

Figure 16: The Ansoff Matrix

a reputation, legal, technical or business include ArcInfo, a web-based tool for

perspective. As such, the firm must take a welding parameter data analysis; and
deliberate and well-thought-out decision WeldEye, a mobile application to enhance
because it will probably have long-standing welding quality control. Welding data is
implications – for good or bad. now a key element of Kemppi’s product/
service offerings. The company has already
Let’s look at an example of differentiation, opened a data interface that enables
value creation and re-positioning in the other firms to create applications that
industry value chain within a traditional enable next-generation welding.77 In the
equipment manufacturing industry. future, Kemppi’s core business might
Founded in 1949, Finnish company Kemppi no longer be equipment sales but the
Ltd. manufactures welding equipment. provision of data services to help partner
In recent years, the company has built companies thrive. This means Kemppi
a robust portfolio of digital offerings could potentially re-position itself from an
to complement the sales and use of its equipment manufacturer (where it faces
traditional welding equipment. These intense competition and commoditisation)
towards becoming a data services provider • If a company sells connected products
that delivers value to the entire industry (e.g. cars or machinery) and, through
segment (something very few players are data and intelligent systems, learns
capable of doing yet). to predict the usage patterns of its
products, it could potentially start
In an influential article, Bessis76 makes selling insurance for its devices.
an important distinction between “AI- Data and AI in this scenario could be
inside” offerings and “AI-first” offerings. used to calculate potential errors or
Whereas AI-inside design in his view maintenance needs in products and this
means that a few AI features have been insight would underpin the probabilities
added to an old product/service offering, needed for insurance purposes.
the AI-first design means that the “deal Consequently, the firm’s business
between what humans do and what machines profile will change (from machines/cars
do has been renegotiated”, and by using the to services/insurances, from one-off
AI-first offering humans will stop doing sales to insurance contracts), while the
something they used to do by themselves. customer group could well remain the
In AI-first offerings, there are typically same. Tesla, to cite one example, has
a handful of deep AI-centric features started to sell its own car insurance in
that remove legacy features altogether, California.78
so creating an entirely new AI-centric • If a telecoms company serves its
product experience. AI-first products will consumer customers with mobile
be highly successful if the technology is phones and related subscriptions,
mature and reliable enough and the market and collects valuable data on their
is ready for them. For instance, where location and creditworthiness, could
users are comfortable with delegating it leverage its data to build alternative
specific tasks they used to carry out to and complementary revenue streams
machines and wish to merely oversee from banking services? This is what
software-automated and algorithm-driven French telecoms giant Orange has
digital operations.52/66/76 When it comes to done: it launched its banking services
ambition-setting in a company, the AI-first in 2017 and has since become a major
approach is invariably a high-ambition challenger to incumbents such as BNP
choice – whereas AI-inside may not be a Paribas, Societe Generale and Credit
high-ambition choice: this depends on the Agricole. Under the EU’s Payment
readiness of a company’s market and how Services Directive 2, banks must make
developed and intense data and AI-related customer data available to other
competition is within it. companies, if the customers agree,
making it easier for challengers like
The ultimate disruptive move for any given Orange to operate.79/80/81
company, powered by using data and
AI, would be to enter a completely new So, given the various options that
industry segment: companies have, including the incremental
approach, expansion to new markets and • Do our capabilities, organisation,
new product/service offerings, or even leadership style and company culture
entering a new industry segment, what favour driving data and AI-enabled
ambition level should companies aspire disruption, or instead favour an
to? There is no standard answer to that incremental approach?
question. It depends on the market, the
competition, the company and its leaders. As examples seen in this chapter illustrate,
However, the following questions might help: there are endless opportunities for
• Where do we see the most lucrative data companies to renew their business with data
and AI opportunities? Within existing and AI. There is strong evidence that most
markets? Beyond them e.g. among new successful companies in the world are not
customers or product/service offerings? limited by industry boundaries or forces,
Or perhaps between industry segments? instead they often try to disrupt others.7 But
• After studying our market and even if disruption is not the end goal, the
customers, do we see an opportunity to ambition level should still be set rigorously
disrupt our own industry, using data and by management so there is consensus about
AI? How could we drive disruption and the company’s future strategic direction.
take advantage?
• Which other companies have data and There is no right or wrong approach when
algorithms that would enable them to it comes to data and AI-related ambition.
enter our industry segment and disrupt Not every company needs to become an
it? If that happens, what is the impact on AI-first company. Not every company needs
our company and what would we do? to develop completely new data and AI
• Which new markets or industry product/service offerings and revenue
segments could we potentially enter, models, or hunt for new customers, or
using our data and algorithms? attack new sectors. Ambition is always
• If we enter a new industry segment, a matter of choice. And sometimes a
should we do it alone, by harnessing prudent incremental approach might be
our own data and AI capabilities? Or a smarter choice than a radical and risky
with partners, e.g. by selling our data or one. Whatever the chosen ambition, it’s still
algorithms to someone who runs the new important for companies to seize data and
business on our behalf? AI-enabled market opportunities and to
• Is there another company that has follow the latest developments.
relevant data and AI capabilities that
would help us to enter a new sector?
If so, could/should we collaborate with
them?
• Given the market, the competition and
the opportunities we see, is it ok to take
an incremental approach, perhaps follow
others, and still stay in the game?
4.2

• Do we have adequate understanding data science problems so that we can

of the commercial opportunities and start building solutions to them?
models related to data and AI? • Does our technology stack and data
• Are our data and AI-related activities infrastructure enable us to meet the
linked to our overall company strategy ambition that we have set, and does it
and plans, and are we addressing these also enable fast experimentation of new
at a strategic level? data and AI-enabled business models?
• Do we have what it takes to renew our • Do we have the necessary data lakes,
business models in practice, e.g. enough data pipelines, interfaces, analytics
entrepreneurial spirit, business acumen tools, governance models and
and change management capabilities? architectural clarity in place that enable
• Are we able to turn our business us to start building intelligent solutions
problems quickly and efficiently into on top of them?45/66
• Can we attract and keep the necessary a servicing business (e.g. maintenance
data and AI talent, and build strong offerings complementing equipment
leadership for our data and AI-related sales), then to guaranteeing outcomes to
activities? customers,83 and finally to autonomous
operations that optimise everything. In
Figure 17 represents a simplified “data addition, while at the beginning the focus
and AI maturity model” for companies to may be increasing efficiency in their own
consider. The closer the firms are to the operations, later companies can pivot
“champion” position (stage 5), the more towards bringing more value to their
data and AI-related activities are prompting customers (notice the term “pull economy”)
significant changes in core functions28 and the new business models also kick
and the more opportunities there are for in. Figure 18 by GP. Bullhound82 describes
business model renewal. By contrast, if this anticipated transformation in the
companies are still starting their journey manufacturing sector.
(stage 1) or in the exploration phase (stage
2), some new data and AI-enabled business Without data and AI-enabled operations,
models might not be realistic yet. a shift towards a more advanced state
would not be feasible in manufacturing
In this book, we have assumed that a firms. At first, the data typically illustrates
company has the necessary legal rights for what is happening e.g. in a production
storing, analysing and using its data, and site or in the logistics chain. But later, the
that there are no legal or ethical matters company can move towards predicting
that would prevent commercialising the events (answering questions such as “why
data. These assumptions are also valid is it happening?” and “what will happen?”)
when discussing the maturity analysis. before eventually creating self-optimising
Basically, the earlier in its data and AI responses to various events in the
journey that a company has ensured operations (e.g. production machines that
it follows all relevant legal and ethical adjust their power or speed automatically)
standards, the better positioned it is to and facilitating completely new disruptive
explore business model renewal; and the business models.
more confidently it can move forward
towards the champion stage. Maturity analysis plays an important role
for businesses in all sectors. Visiting other
The machinery industry provides a practical companies or industry events or talking to
example of the choice either to use data and external specialists, can help companies
AI to simply renew the existing business identify their data and AI-related strengths
or to develop completely new data and AI and weaknesses and potentially also help
services.82 Here, the typical path – which to clarify how to achieve a higher level
is still merely an ambition for many firms – of maturity. Visiting other companies or
involves manufacturing companies moving industry events or talking to external
from being a product-centric business to specialists, can help companies identify
 STAGE 1: BEGINNER STAGE 2: EXPLORER

Strategy & Data and AI are not present in Lack of focus and planning in
positioning company strategy. No roadmap data and AI-related activities,
for data and AI activities. even though progress is made
Laggard position

Commercial Value of data and AI and size of Basic understanding of data and
models & busi- commercial opportunity AI as assets. First pilot projects
unclear planned. Commercial models
ness acumen explored

Processes and No clear roles or processes Processes and roles for data
ways of working with data and AI. Lack of and AI getting clearer. Some
experimenting in service experimenting in service devel-
development. Organisational opment. Data and AI initiative
silos. Difficulties in getting portfolio getting clearer. Long
services ready time-to-market

Data & tech Data scattered and fragment- Data infrastructure and APIs
ed, not easily available and sporadically being built by
not easy to correlate. Data smaller teams within the
collection not applied system- organisation. Extracting and
atically connecting data from various
sources. Data queries and
reports. Data quality known

Organisation & No data science capability, Some individuals who can

competences dependency on vendors, play with data and analytics,
difficulties in attracting talent, scattered around the organisa-
lack of data literacy, fear of tion, wider adoption of the data
dealing with data and AI science skill is missing

Figure 17: Data and AI maturity model (source: Mika Ruokonen)

STAGE 3: INTERMEDIATE STAGE 4: ADVANCED STAGE 5: CHAMPION

Data and AI-related activities Clear understanding of oppor- Data and AI as critical assets of
are aligned with company tunities. Roadmap and business the firm, enabling transformative
strategy; roadmap and busi- case for all activities. Data future businesses and business
ness case becoming clearer. On and AI creating a competitive re-positioning. A leading player
par with its peers advantage

Data and AI commercialisation There is proof that data and AI Data and AI are harnessed in mul-
pilots in place. Data and AI to have been turned into business tiple ways e.g. to improve process
customers and/or partners value. Clearly demonstrated efficiency, to rethink the current
for free. Data and AI used improvement of internal business and to create new data
to improve current business operations and revenue from and AI-enabled business models.
but not to re-think the whole customers from data-related Some models are transformative.
business product/service offerings New data and AI-
enabled service concepts being co-
created with customers

Processes and roles with data Systematic data and AI initiative Clear processes and roles for data
and AI are clear. Some lean and portfolio management and lean and AI in place. Lean and custom-
customer-centric development and customer-centric service er-centric service development
projects. No systematic data development in place. Multi- in place. Systematic data and AI
and AI initiative portfolio disciplinary teams used but initiative portfolio management.
management further competence develop- Multidisciplinary teams is the
ment is needed. Time-to-market norm. Fast time-to-market
reducing

Data visualisations and Data is catalogued and Data is fully governed. Scalable
forecasting models. Situa- widely shared, data quality is cloud solutions with easy access
tional awareness. Training good. Common data platforms for users. Various data platforms
of algorithms. Data platform and tools enabling analytics. and tools in place. Continuously
initiatives launched. Transfor- Ability to process and use vari- improved advanced analytics and
mation to cloud technologies. ous types of data. Creation and AI models, including predictions
Data quality is satisfactory testing of custom algorithms. and optimisation. AI applications
Optimisation and automatisa- in place. Autonomous operations
tion of operations

Dedicated champion for data Data scientists available to col- Strong and institutionalised data
science. One group of people laborate with business. Access science capabilities. Data scientists
who know how to do analytics to BI tooling and data-enabled at the core of business. High data
and data science. Attempts decision-making in place. Data literacy and strong data-enabled
to raise data literacy more being analysed and discussed decision-making. Data and AI skills
widely. Attraction of data and in the organisation. Capabili- embraced in hiring and training.
AI-related talent ty-building programmes Strong support for harnessing
data and AI. Reinvention of ways of
working
their data- and AI-related strengths and certain level of maturity in terms of data
weaknesses and potentially also helps to infrastructure and management, analytics
clarify how to achieve a higher level of and data science competence, as well as
maturity. If they have already high level data commercialisation understanding.
of maturity in relation to data and AI Companies need to ask:
but customers are not yet ready for the • Is our target customer capable of buying
solutions, it doesn’t do any harm to take a and using our data and AI-enabled
step back. product/service offering, given the
format in which we are planning to sell
Of course, one can also look at maturity them and their current maturity level? If
from the customer’s point of view. What not, should we package and productise
level of readiness is required from them the offering further, so that it is easier
to adopt data and AI-enabled offerings to use and does not require deep data
and revenue models? Take Facebook as science skills?
a product, for instance. Even though its • Furthermore, is the revenue model that
platform is heavily optimised by data and we are planning to implement something
AI, its user interface is such that anyone that accurately communicates
can use it easily. Next, think of a big chunk our desired value proposition and
of raw data. In order for someone to buy something that the customer can
it and harness it, there needs to be a easily understand? If not, should we

INTERMEDIATE NEAR-TERM LONG-TERM LONGER-TERM

Operational New products Outcome-based Autonomous

efficiency & services economy pull economy

• Asset utilisation • New business • Pay-per-outcome • Continuous demand

• Operational cost models • New connected sensing
reduction • Pay-per-use ecosystems • End-to-end
• Improved produc- • Software-based • Platform-enabled automation
tivity, safety and services marketplace • Resource optimi-
working conditions • Product/service sation and waste
hybrids reduction
• Data monetisation

PRODUCT SERVICE OUTCOME PULL

Figure 18: Examples of a transformation to new business models in the

machinery industry
consider changing it to something that
is more familiar, in order to match it
with the customer’s readiness to adopt
something new?
4.3

Take Coca-Cola. Coca-Cola does not sell Next, take McKinsey & Company, which sells
direct to consumers. To date, there are tailored consulting projects. In contrast to
no sensors in its bottles or cans that Coca-Cola, McKinsey can get very close to
track usage, whether the bottle is half full its corporate customers, collect data on
or half empty. Through its distribution their business, on the projects delivered
network, however, Coca-Cola can estimate to them and on how those customer
consumption, demand and consumer relationships developed over time. McKinsey
preferences in each of its target countries. also has data from the “supply side”, e.g.
It can also use data from its website to which services and methodologies from its
undertake consumer profiling. Coca-Cola portfolio are most used by customers and
has also invested in tracking how its various which services and methodologies it might
products are discussed on social media, need in the future to serve its customers
including image recognition technology even better. The opportunities for McKinsey
so that it knows when its drinks (and to innovate and test new product/service
those of its competitors) are included in offerings and revenue models based on
photographs posted online. In addition, data and AI are significant, and – as with
Coca-Cola has launched self-service drinks the Coca-Cola /fountain example – co-
fountains that let customers mix their own creating with customers is a possibility,
drinks. The data from these fountains allows albeit in a B2B context. Overall, however,
Coca-Cola to identify the most popular the starting point for harnessing data and
flavour combinations and turn them into a AI is very different for McKinsey than it is
ready-made drink for a wider audience.84 for Coca-Cola.
 CONSUMERS (B2C)

Service delivery
Product usage data
data
Consumer profiling
Consumer profiling
data
data

Target
customers

Service delivery
Product usage data
data
Customer lifecycle
Customer lifecycle
data
data

COMPANIES (B2B)

PRODUCTS Offering SERVICES

Figure 19: Four categories of companies according to their offerings and

target customers (source: Mika Ruokonen)

By and large, companies (or their divisions and the conditions of our market, what valuable
or businesses) can be divided into four data and AI assets do we have already or could we
categories, as illustrated in Figure 19. This create in the future through our operations? How
categorisation is important to understand could those assets be transformed into sources of
because it lays out the broader business new revenues?”
context for the data and AI-enabled
business models that a company can start While the two-by-two matrix of target
to explore.45 customers vs. offerings looks stable and set
in stone, in practice it is often quite flexible.
Nearly all business leaders find a place for For example:
their business in the two-by-two matrix • Companies might have both B2C
instantly and intuitively: and B2B customers (for instance,
“We’re a product/service company, and our typical telecoms operators typically serve both
customer is a consumer/company”. consumers and companies).
• With a bit of adaptation certain B2B
The key questions for business leaders to products and services could be sold
ask are: “Given the characteristics of our business, to consumers too (during the recent
 Coronavirus pandemic, certain but often about “both-and”. Due to the
wholesalers who used to deal mainly revolutionary nature of AI technology, shifts
with restaurants started to sell in the target customers versus offerings
ingredients directly to consumers, for table are possible, and data and AI might
example). be the critical elements that enable a new
• Some services could be “productised”, relationship between a customer and a
and new services could be built around product. With a bit of creativity, data or AI
existing products (think of Boeing’s assets could be converted into any variety
aeroplane maintenance service, for of products and services that could benefit
example). various different customers.
• Some physical products could
be turned into services (think of
CDs and DVDs, versus Netflix and
Spotify’s data-enabled personalised
recommendations).

For companies wishing to renew their

business models with data and AI,
therefore, it is not about “either-or”

• Data aggregator
Vulnerable • Open source data
• Freely available algorithms

• Data and algorithms available only to selected players

Buffered • Platform builder

• Creator of unique data

Protected • Creator of unique algorithms
• Platform owner

Figure 20: Defending data and AI-enabled business models: three levels
4.4

When it comes to data and AI, the barriers very least, competitors will have to make
to entry may be fairly low, resulting in similar investments in data, algorithms and
commoditisation and fierce competition. platforms to catch up.
In this scenario, creating and sustaining
a profitable data and AI-enabled business In other words, the more a company invests
model could be a considerable challenge.85 in building something of its own, rather
than using something ready-made, the
So how is it possible to create and sustain bigger the business potential, assuming
a profitable data and AI-enabled business there is a genuine customer need for it (as
model? There are four important elements discussed in Chapter 2.4).
to consider – the model should:
1. Be built on unique assets The uniqueness of data and AI assets
2. Stack up against what’s on offer from depends on the context in which they are
the competition being harnessed, too. Even if some data
3. Link closely with the core competence of and algorithms are unique and therefore
the firm valuable to the company selling them, they
4. Have the potential to become a game- might not be equally unique or valuable to
changer in the given marketplace.72/76/85 the customer who is operating in a different
context. Creating unique data and algorithms
Figure 20 (original De Saulles,85 edited by does not automatically lead to high business
Mika Ruokonen) depicts three levels when potential. For companies selling data and
it comes to defending data and AI-enabled AI assets, it makes sense to test the waters
business models. The more a company relies with potential partners as early as possible
on aggregation of data or open source data, in order to understand how to create value
and on freely available algorithms, the more that stands out from the competition. The
likely it is that its competitors rely on them company doesn’t necessarily have to do
too. This means there is little differentiation everything alone; data and AI assets that are
from the competition.11/86 However, the game-changing and unique could be created
more the company bases its business via a partnership, in which various companies
models on the creation of unique data (and (e.g. other data and algorithm providers)
state-of-the-art algorithms) delivered via a contribute to the final result.37
company-owned platform (or other similar
proprietary channel), the harder it is for Different companies have different strengths
others to copy its business model. At the and weaknesses. For companies wishing to
implement new data and product altogether in the future). These
AI-enabled business models it makes sense companies might use data and AI to help
to identify their core competencies (i.e. the their customers to develop better products,
most critical strengths) then think of how more quickly.
to “codify” these into data and algorithms.
If the core competence of the firm is These examples illustrate that a company’s
already attractive to customers (and other core competence, when turned into data
stakeholders), it will also be attractive when and algorithms, can be a great source of
leveraged by relevant data and algorithms. valuable new product/service offerings.

Take consulting firms. Their core But what if data and AI are not built on core
competencies are talented people competence but on something else? Can
and unique methodologies – as well as they still be valuable to customers? Yes,
successful project delivery to customers; in potentially. A digital by-product of a firm,
other words, in operating a thriving service some data or algorithm created on the
business. If these companies capture data periphery of the existing operations, could
on their talent, methodologies and project potentially be valuable to some customers.27
success factors and use algorithms to But is it sufficiently unique enough and does
optimise them, they likely have something it deliver enough competitive advantage to
valuable to offer as an extension of their make it worth developing?
consulting business (and potentially as a
replacement in the future). They might, for Finally, it’s essential to work on business
instance, help customers to harness their models that are game-changers and enable
own talent or manage their own projects transformative growth. Mark W. Johnson,
more effectively too. Clayton M. Christensen and Henning
Kagermann72 have worked on business model
Next, think of R&D-heavy companies that renewal and state that “companies should not
develop high-tech products. Their core pursue business model reinvention unless they are
competence is in innovating, launching and
continuously improving their products; ”. The same goes for data and
those that harness the latest technologies. AI-enabled business models.
If these companies can capture data on the
development and usage of their products Business leaders needs to understand how
(e.g. how to build new products as quickly their new business model might harness
as possible, or how to improve the usage data and algorithms in a way that is superior
of their products using data) and use to the competition (remember the “AI-first”
algorithms to optimise those, they too thinking earlier in Chapter 4.1), and how
have something that could be offered as a these new models could enable completely
standalone offering and extension of the new types of businesses in the future. By
product-focused business (the data and using data and AI, the aim should be to
AI offering might eventually replace the reimagine (not just fine-tune) how something
in the customer’s daily life works today, would allow us to collect unique data or
helping them to become happier, wiser, create algorithms?
cooler, richer, freer, etc.39 • How could launching data and algorithms
as an extension of our current business
Then again, building game-changers needs and/or as a standalone offering be a
to be put in the context of the company’s game-changer? Or something highly
ambition around data and AI (as discussed valuable that people outside our
in Chapter 4.1). Does the firm want to use industry would be interested in?
data and AI to disrupt the industry or • What special elements are unable to be
find untapped space between industry easily copied by our competitors?
boundaries, for instance? Or is it sufficient • How could we use data and AI to
for the firm to “stay in the game” when it reimagine something that is not working
comes to data and AI, while also building optimally today; and help our customers
sustainable competitive advantage? This reinvent themselves?
could be by harnessing some other unique
assets that the firm possesses, such as
strong customer relationships or exclusive
access to certain raw materials. That is fine
too. So, even though creating game-changing So, in order to develop successful data and
business models using data and AI is a great AI-enabled business models, companies
ambition to aim for, for some companies a need to:
less ambitious plan might be sufficient. • Set their ambition
• Assess their maturity
To summarise, when developing new data • Build on business characteristics, and
and AI-enabled offerings, a company needs • Create something unique and
to ask: game-changing
• Do our processes and operations
currently allow us to generate unique Ambition is a state of mind, whereas
data? Or could they be improved to allow the last three bullet points depend on a
us to do so? thorough understanding of a company’s
• Do we already develop unique status quo. In our view, these four points
algorithms? Or could we start are key to developing a sound commercial
developing them? plan for data and AI.
• What are we really good at in our
company? What is our core competence? We return to these points in Chapter 5,
How can we turn that core competence when presenting different business model
into unique data? Or can we create options for companies (the playbook), as
an algorithm that captures our core well as in Chapter 7, when discussing the
competence? findings and insights from our co-creation
• Are there other processes or operations sessions with companies, illustrating
on the sidelines of our business that business model implementation in practice.
5 The playbook
of data and AI-
 Selling existing Selling data,
Selling data
product/service analyses and
platforms
offerings reports

Selling data and Selling outcomes

Selling AI assets
AI-based services or capacity

Figure 21: Data and AI-enabled product/service offerings (source: Mika Ruokonen)

As a reminder, Chapter 3 described

business models and business model
innovation as a multifaceted and sometimes
complex concept, encompassing several
different definitions. Our aim in this chapter
is not to describe every potential data and
AI-related business model under the sun:
instead we will deliver a simple presentation
of model options that are both practical and
implementable.

We start our analysis by outlining data

and AI-enabled product/service offerings,
answering the question “what is being
sold?” using six different models (see
Figure 21).
The idea:

An example: Tesla sells cars that are In this model, data and AI are considered
enhanced with data and AI capabilities, e.g. as features of existing service/product
autonomous driving, but what the customer offerings.
buys is still a car.
There is no need to educate customers
Reflections: From the business model point on how to buy data or AI assets from the
of view, this incremental approach is the company, as they are not typically sold
easiest. Remember our earlier discussion or priced separately (Tesla, shrewdly,
about the ambition level for data and AI? If harnesses the eagerness of early-adopter
a company doesn’t wish to “rock the boat” customers to test its various new data and
within its particular industry or value chain, AI-related products).
but instead strives to deliver excellent
customer service and products for its This model means that the core purpose
existing customers, then this model makes of the company does not change: but the
sense (and serving existing customers company using it could still introduce new
better is a worthy enough ambition alone). data and AI product/service offerings to
its customers in some way. It could use the company’s existing revenue model.
data and AI to create tailored offerings Whether it involves customers paying
for its customers from its existing for products directly, or via monthly
products or services; or make data-based subscriptions or licence fees, it all
recommendations on what products to buy remains the same.
next (think of online fashion retailers, for • That said, customers might not even
example). Returning to the Tesla example, be aware that the price of the original
the car effectively navigates by itself, which product also includes data and AI
is revolutionary in its own right. There are components – and this could become a
many opportunities to innovate with data problem if customers start to expect
and AI, even where products and their the latest AI technology “for free”,
revenue models remain more or less the coupled with the core product.
same as before.
While this selling existing product/service
According to a Microsoft study,41 many offerings model is a simple option from the
company leaders anticipate that AI will business model perspective, for many firms
have a bigger impact on their core business it is not enough. According to the Microsoft
when it comes to new business models and study,41 57% of European companies expect
products in the longer-term. In light of this, AI to have a big impact on business areas
it may make sense to practise data and that are entirely unknown today, meaning
AI commercialisation within the existing a majority are looking beyond their
business framework, without taking bigger current business model when it comes to
steps or risks. Once a company has more harnessing AI. So, it makes sense to explore
confidence and experience working with other product/service and revenue models
data and AI, business leaders may choose to that extend beyond a company’s existing
move onto more radical new models. portfolio. Let’s move on and look at other
options.
Revenue models:
• When it comes to generating revenue
from data and AI using existing
product/service offerings, customers
can “see” the data and AI elements
when using the product (in the case of
Tesla, the car comes with navigation
features). But typically, they don’t
need to pay anything extra for them –
unless, for example, they are paying for
an advanced version which has more
features.
• The revenue model for the data and AI
product/service offering tends to follow
The idea:

Examples:
• Vodafone sells anonymised and raw • Sievo, a Finnish software company, sells
location data about its customers procurement analytics solutions to its
to TomTom, which uses it to create corporate customers.
navigation services.
• Walmart sells point-of-sales data to Reflections: Selling raw data is a somewhat
various brands that supply products in basic and unsophisticated model. A
the retail value chain. company takes the data, packages it into a
• iRobot, the robotic vacuum cleaner manageable form, then sells it to another
manufacturer, sells data on homes to party (e.g. a company or organisation)
tech companies that develop smart more or less as it is, and gives the buying
home interfaces. company or customer ownership of the data.
• Finnish pharmaceutical distributor AI might play a limited role, for example in
Tamro sells analyses of drug purchases processing the data before it is sold.
by consumers to local pharmacies and
drug manufacturers. The added value of the data to the person
• Fitbit sells sleep quality and duration or company buying it might be very high
analysis to consumers. (for example, if it is scarce, or difficult to
get from anywhere else), or alternatively predict a customer’s health insurance
very low (the customer purchases data needs.
cheaply – perhaps because it sees some • iRobot, the company behind Roomba
additional value e.g. in combining and robotic vacuum cleaners, collects
analysing the bought data with its own mapping information of homes (e.g.
data). As a result, the price of data and the distances between sofas, tables, lamps
fees in this model can differ considerably. and other home furnishings) with a view
to selling it, with customer consent, to
It is common for data sales to occur within tech companies such as Apple to help
the value chain of the business supplying improve their smart home information.
it: in other words, the buyer is a known The tech companies then get to enhance
customer or partner, and so there is a pre- their future smartphone products and
established relationship between the two.87 improve accessories such as artificially
In this scenario, the company selling the intelligent voice assistants as smart
data can potentially reduce any legal risks home interfaces.88
related to the data sale by only selling data
to selected partners.34 An example in a B2B The point is to think creatively: “Who could
setting is the fast-moving consumer goods
business: retailers such as Walmart sell
their point-of-sales data to their suppliers,
such as Procter & Gamble, who then keep
an optimal stock of products in their A key question is whether selling data is
warehouses to deliver just-in-time resupply a smart move or not. Some companies
to supermarket shelves.31 actively avoid selling raw data in order
to protect their assets or reduce legal
However, there are also many opportunities risks.34 In this scenario, is it wise to give the
for firms to explore selling data outside company buying the data ownership of it,
their value chain or industry. For example: or would it be more savvy to turn the data
• Volumes of export and import goods in into other products, allowing the customer
harbours are a reliable early indicator of limited access to some subsets on a view-
a country’s economic activity, so some only basis? Could this approach reduce
harbour operators sell that data to the risk of inappropriate or unethical
companies, such as banks and financial use of a firm’s data? Also, if selling data
institutions, that want to forecast always requires tailored solutions and data
economic fluctuations. exchange agreements for each individual
• Sports clothes design and marketing third-party customer, is this lack of
companies like Nike collect data scalability blocking growth?31 These are
generated by smart sports garments. questions that companies selling data must
In the future they may start to offer consider very carefully.
that data to insurance companies (with
customer consent), which can use it to
If the company selling doesn’t want to of the business selling data, and the
sell all their data to another party and/or relationship the two parties form together.
wishes to add more value to the product
before the sale takes place, then the A good example of selling analyses is Fitbit,
analyses and reports model may be a the wearables and activity tracker. Fitbit
good option instead. The same data, when offers a premium digital monthly/annual
analysed and turned into reports, perhaps subscription that provides sleep quality and
with the help of AI, could be sold to various duration analysis to consumers, as well as
parties as a bespoke proposition with some personalised health and fitness guidance,
tweaks. based on their own data.89 Fitbit’s core
business is the sale of its activity tracker
The sale of analyses and reports has the products, but the provision of analyses is
potential to be a very scalable model, also an integral part of its services that
provided that specific customer groups customers want to buy.
or individual customers do not require
labour-intensive report tailoring. In this Finnish software company Sievo Ltd,
model the key thing is to choose between meanwhile, sells procurement analytics
one-size-fits-all (the business selling data solutions that help its corporate customers
says: understand more clearly how much they
), one-size-fits-many ( spend on various categories of purchases.
This solution is particularly useful for
) and large multinational firms that have various
one-size-fits-none ( Enterprise Resource Planning (ERP)
systems; meaning data on purchases is
often scattered in various silos around
). The more tailoring the company. By using its customers’ own
there is, the closer the business selling data and applying AI, Sievo creates weekly
data can get to the needs of the company or monthly reports for those customers,
purchasing it, and the more the service and applies an annual licence fee for its
starts to resemble consulting work. The reporting services. In addition, it generates
more standard the report, the easier it is to project revenue from the implementations
scale up the business model quickly: so it of its solutions within the customers’
might be tempting for many companies to complex IT environment, especially ERPs.
keep tailoring to a minimum.
Revenue models: There are several models
When selling analyses and reports, the that could be used to sell data, analyses and
supplier typically keeps ownership of their reports:
data and grants the customer a licence • If there is a data dump, e.g. a static set
to read the report or view the analysis. of information that is sold as a one-
The depth of the information given to the off, it could be sold like any product,
customer depends largely on the wishes with a fixed one-time fee. The fee
 could also be charged for each device As we can see from this breakdown, many
that is connected into the data supply revenue models that could be used for
chain of the supplier. Similarly, if the selling raw data might apply to selling
customer wants only one report or analyses and reports as well. The selection
several reports, they could be sold like of a revenue model also depends on what
products, with a price per unit, or price format the data and report offerings are
per number of readers. provided in, and what kind of commercial
• Alternatively, if a company is buying a goals the company selling the data and
constant supply of data, a monthly or its customers have. It’s a good idea for
yearly licensing fee might work well (see businesses selling data to join forces with
the example of Amazon Web Services their customers in exploring all available
Data Exchange in the reference section options.
of this book). As illustrated by the Sievo
example, if reports are sold regularly,
then a monthly or yearly subscription or
licence fee could be an attractive option.
• The freemium model (provide some data
or analyses or reports for free, convert
freemium users to payable users of
large or more advanced product/
service offerings) or project revenue
model (charge the customer based on
how much work is needed to collect the
data, or to create the analysis) could be
considered, too.
• Another option would be to tie the
price of the data to the benefits that
the business buying it gets after the
transaction is complete (the company
selling says:

).
• It is also worth considering a revenue
model we refer to as business model
overturn. In this model, data is the
only source of revenue of the firm and
everything else is given for free in order
to obtain valuable data that can be
sold on.90 A company may extend this
approach by turning the valuable data
obtained into sellable reports, too.
The idea:

Examples: Google are already providing AI platforms

• Facebook provides advertisers with a through which various companies harness
platform through which to buy targeted their data.10 Some industrial companies
advertising aimed at specific user are implementing customer portals that
groups, based on the data it controls. visualise and display data from their
• Apple’s App Store provides a platform product usage (motors, cranes, trucks and
where consumers can tap into data and various production machines, for instance:
AI assets. the portals show how the equipment is
• Irish company Eagle Alpha acts as a data performing and highlight potential issues
broker and middleman between two that may arise). The App Store for Apple’s
parties – data owners and investors. iPhone is arguably a platform that can
be used to tap into data and AI-enabled
Reflections: In this model, the data owner product/service offerings. There are various
builds a digital platform to provide data speech and face recognition applications
and AI assets to its customers or partners. to choose from, making the platform a
The platform could be a customer portal, “carrier” for data and AI capabilities.
dashboard or marketplace. There are
several companies following this path. Large The challenge with implementing data
companies such as Amazon, IBM and platforms is the amount of work and
investment needed to build them, and to for all necessary services (for more about
convince customers to start using them. platform-focused business models, see
Why bother if the customer only needs raw Moazed91 and Zhu and Iansiti92).
data or a few reports or algorithms to solve
their problem? Does the revenue potential One potential way to do business with
of selling data and AI assets on a platform data platforms is to assume a data broker
outweigh the investment needed to build, role. Eagle Alpha Ltd., an Irish company
launch and scale a platform business; established in 2012, is an excellent example.
as well as the costs of maintenance and The company acts as a middleman between
development? If not, are there other two parties; data owners who lack access
ways to monetise the benefits created to qualified investors, and investors such
for customers through implementing the as private equity companies, who need
platform, e.g. by raising prices or winning reliable data to help them make better
market share elsewhere? These are all investment decisions. With more than 1,200
questions that companies wishing to build datasets included in its “Data Machine”,
platforms need to consider carefully before Eagle Alpha facilitates the trade between
venturing into expensive development the two, solving key industry challenges.
initiatives. The company generates revenue through
various fees that it applies for its services.93
That said, if implementing a data platform So, if it is not commercially feasible for a
helps a company selling data to become company to develop its own data platform,
an orchestrator in its network, then it it may choose to sell its data via this kind
could be a winning strategy. This approach of data brokerage company: by doing so,
could include a platform that sells partner it leverages existing data platforms and
services, or one that is used by third sidesteps costly investments of its own.
parties such as startups to build products
on. Platforms can also be used as a critical The selling data platforms model is a
control point in a value chain or ecosystem: versatile one. Platforms can come in many
for example the platform creator uses it forms, e.g. they can enable and facilitate
to “own” customer relationships in a move matchmaking (like Uber and Airbnb,
that no competitor can bypass. In these for instance) or serve as a “carrier” for
scenarios, implementing a data platform data and AI assets that are provided to
could lead to a nearly monopolistic market customers (as in the App Store example).
position and create a virtuous cycle of Sometimes they do both of these things
more users – more partners – more data simultaneously, as seen in the Eagle
– more revenue. A company could use Alpha example. The key for any business
its platform to sell not only its own data leader is to nurture “platform business
and AI assets but potentially also the model thinking” in their company, i.e. help
assets of its customers, partners and employees identify opportunities for a
even competitors as well. It could provide platform business in any given industry or
customers or partners with a one-stop-shop marketplace.
Revenue models: When it comes to revenue are various service fees in place for data
models, data platforms offer an appealing owners and investors).
set of options to choose from:
• Many companies do not yet charge for As we can see, various different revenue
access to their data platforms, maybe models can be applied to selling data
because customers expect to receive platforms, and a business can run several in
data for free, so they bundle the price of parallel to achieve maximum impact.
the platform into the price of their core
product. Those that establish a separate
price for their data platform often
sell monthly or annual subscriptions
or licences (a practical example is
LinkedIn, which stores people’s CV data
and sells various subscriptions to both
consumers and companies against
additional features).
• Various handling or brokerage fees,
provisions or commissions are also
typical (the platform provider receives
money every time something happens
on the platform; Uber and Airbnb are
examples of this). In these cases, the
price that the platform user pays is tied
to the number or value of transactions
made through the platform, or to other
specific KPIs.
• Firms could also consider selling
advertising on the platform (if it has a
lot of consumers as users) or a business
model overturn (give everything else for
free to obtain valuable data that can be
monetised via the platform).
• Revenue models for different users of
a platform may vary. Some platforms
might be free for consumers to use, but
companies “on the other side” of the
platform have to pay (think of Facebook
– only the advertisers pay). The fees
for various user groups of the platform
may also vary in a B2B setting (e.g. in
the Eagle Alpha brokerage model, there
The idea:

Examples: have turned their ecommerce or software

• Amazon Web Services provides its ML/ capabilities into cloud computing providers.
AI capabilities, e.g. voice analysis and
translation modules, to all its users as This model is not solely reserved for large
part of its cloud offering. digital companies, though. Various smaller
• Firstbeat models heart function and suppliers are also developing their own
heart rate variability, and provides this AI assets, helping companies to lower the
algorithmic technology to sports watch barrier to accessing advanced analytics
manufacturers. (the company selling says:

Reflections: This model is for companies

that have advanced and powerful analytics, ).
machine learning or AI applications in As an example, technology startup Firstbeat
place. These applications may have initially has developed analytics technology
been developed for internal use, then that creates a digital model of a user’s
turned into sellable products for others physiology through advanced modelling of
to benefit from, too. Or, they may have heart function and heart rate variability.
been developed for external sale from the It is licensing this algorithmic technology
outset. to sports watch brands such as Suunto,
Polar and Garmin, to complement their
An increasing good quality of AI assets own products.94 If a company’s analytics/
are freely available in today’s market. AI assets are powerful enough, they may
Businesses only need to check the product/ enable other firms to build new innovations
service offerings of Amazon, Microsoft on top.
and Google to realise that analytics,
machine learning and AI assets for The sellable AI assets might come with data
various applications are available nearly (the business provides data that makes
or completely off-the-shelf.10/39 Here we its AI assets even more valuable to the
can see an intriguing example of business customer), or without data (the business
model innovation in which large companies aims to help the customer reap benefits
from its own data). There may be data • In order to create a recurring, stable
storage and ownership issues to deal with and scalable revenue stream, the
here, too. Some AI asset suppliers are happy licensing model is a good option.
to provide their technologies to customers
without requiring any ownership of data or
responsibilities to store the analysed data,
whereas some suppliers want to transfer
the data to their own servers/cloud before
the analysis can happen. There’s no right
or wrong option here: from the customer’s
point of view, it might just be easier to
take the AI assets without giving away
the ownership or storage of their data to
anyone else. Whereas sometimes storing all
the data in the supplier’s environment can
be a big benefit and reduce the customer’s
server costs (this is typically the case
when harnessing the AI capabilities of
cloud suppliers such as Microsoft Azure or
Amazon Web Services).

Revenue models:
• The most common revenue model for
selling AI assets is a monthly or yearly
licensing fee, which allows the customer
to harness some or all the capabilities of
the AI-enabled product/service offerings
for a certain group of licensed users.
• The business selling AI assets might want
to secure a share of the revenues that
the customer generates with them, or tie
the fee to the amount of data processed
with the AI assets. In this model, there
are certain KPIs or metrics in place that
define the price of AI asset use.
• Other models include freemium (the
basic AI features are free, the advanced
ones are payable), consulting fees
(related to technical implementation of
the AI asset) or product revenues (all
usage is free after a one-off payment).
The idea:

Examples: selling AI assets), were directly related to

• IBM provides its industrial customers products. But this latest category shifts
with equipment effectiveness towards data and AI-enabled services.
improvement and energy usage In this model, the business selling data
optimisation through its Watson IoT and AI-enabled product/service offerings
service. strives to help the customer to do things
• Finnish company Lassila & Tikanoja better, faster, smarter or more efficiently.
provides real estate owners with data- To do this, it sells consulting services that
enabled consulting services. are augmented by data and AI.
• Australian explosives company Orica
Ltd. sells various consulting services Essentially, the company selling says: “We
that help its customers to achieve
efficient and productive blasting.

Reflections: The previous three models

we explored (selling data, analyses and For the customer,
reports, selling data platforms and this is an easy option: there is no need
to start developing its own AI assets or The issue arising from these kinds of
invest in data collection or analysis efforts, service concepts is that the customer may
because the consultant comes in and takes seek to contain valuable learnings within its
ownership of the entire process (the benefit organisation, rather than lose them to the
is typically reflected in a higher price for company selling the solution (or to other
the service, too). companies or competitors in the same
industry). But the company selling would
The field of circular economy offers a never have benchmark data if its customers
practical example. Lassila & Tikanoja Plc. retained all the learnings in this way.
provides its real estate owner customers However it also has no right to use the data,
with a service concept called “Smartti” that even on an aggregated level, in any way
uses data from its customers’ buildings to after the project. So effective contracting is
help them lower energy consumption and needed to make this model work in practice.
save costs. The data is complemented by a
consultant who undertakes an analysis of Another topic worth discussing is whether
the building’s energy usage then delivers the customer receives rights to use the
analyses, reports, recommendations, consulting business’ AI assets once service
guidance and training for making delivery is completed. Normally the rights
improvements. There are also control room (if any) are limited to usage only instead
and automation solutions available for of full ownership. The company selling has
customers who wish to optimise energy often invested intensively in its AI assets,
consumption widely across their real and wishes to stay in control of them after
estate portfolio, e.g. to match weather its consulting gigs are complete. Again,
conditions.95 various contracting models are possible to
solve this issue.
Another example is Orica Ltd., an
Australian-based company that provides This service-based model is not only
commercial explosives and blasting systems suitable for a B2B setting; it can also apply
to the mining, quarrying, oil & gas and in a B2C context. For instance, IBM provides
construction industries. Orica collects data its Watson service to medical centres and
from its customers’ blasting operations on their doctors, so that they can analyse
its BlastIQ platform, which it has developed data, diagnose illnesses and provide health
for the purposes of blasting data analysis, consultations to patients. So, in theory, all
storage and sharing. The accumulated types of service work can be amplified with,
data enables the company to sell various or even replaced by, data and AI.
consulting services, including blasting
outcome improvement, hole loading, blast Revenue models: Several revenue models
loading, and technical blast management. can be considered:
All these services help customers improve • The most obvious option is to use
their efficiency and increase productivity in project revenues: the consulting
their blasting operations.96 business charges the customer on a
 time and material basis, depending
on the work needed to conduct the
analysis and deliver recommendations;
and also to follow up and support their
implementation.
• Another option is a performance-based
fee (the company selling says:

).
• Other options to consider include:
a freemium model (basic consulting
services and AI assets are used for
free, advanced services and features
are payable), a retainer model (a fixed
monthly fee for the improvement in
services delivered), or a business model
overturn (giving consulting services for
free in order to gain access to valuable
data that can be sold elsewhere).

Consultants are typically innovative, so they

may find creative ways to combine these
revenue models to meet the expectations
of their customers, while also guaranteeing
their revenue flow at the same time.

The last model we explore is arguably more

radical and disruptive than the previous
options. It typically seeks to challenge
existing market, competition and customer
expectations. Its implementation means
that the vision, strategy and identity of the
company selling could change quite a lot
along the way. It moves from selling data
and AI-enabled products or services (as
discussed so far) to guaranteeing a level of
performance or pre-agreed outcome. Now
things get really interesting.
The idea:

Examples: • ShareNow offers consumers a service

• Machinery company Kalmar provides enabling them to buy the right to use a
harbours with a performance contract car and pay per minute.
in which it takes full ownership of
operating its machines and software for Reflections: Harvard Business School’s
the customer. Theodore Levitt famously said:
• Agriculture machinery provider John
Deere uses data to sell better farming In the selling outcomes
outcomes, including crop yields. model, the company that owns data or AI
• GE Power offers guaranteed assets does just that. Instead of offering
improvements in efficiency for its products directly, it uses data and AI to
power producer and power distributor deliver outcomes (in other words, solutions)
customers. that are helpful for the customer.
• MAN Truck & Bus provides truck-as-
a-service for its customers; based on Remember the earlier discussion about
vehicle and driver data, it guarantees finding problems worth solving as the basis
uptime of the truck for a monthly fee. of data and AI-enabled business models?
Under the selling outcomes model, the and power distributor customers.99
customer wants to outsource some major, Meanwhile, instead of selling doctor
often critical, part of its operations (this appointments, healthcare service providers
model typically happens in a B2B setting). may start selling “health outcomes”, e.g.
The customer’s rationale is to reduce charging their corporate customers if they
risks related to capital investments or can show a proven decline in levels of staff
receive continuous improvement from the sick leave.
company selling outcomes. In this model,
the customer pays for the performance In addition to selling “quarter-inch holes”,
(of e.g. equipment, software and services) there is also an opportunity to guarantee
that the company selling outcomes and that the “quarter-inch drills” are available
capacity delivers.83 In this way, the supplier for the customer to use too, thereby selling
is incentivised and committed to improving capacity. This is also a service-oriented
productivity. business model. Here, the business selling
outcomes decides to always keep some
Data and AI play a crucial role in selling and products or services available for its
delivering outcomes here. By using data customers. “What is being sold” is not a
and AI to fully understand the customer’s product per se, but capacity and availability.
outsourcing needs, the supplier can Hiring equipment/motor/tool/ battery/
predict which customer-specific outcomes robot/tyres per hour are a good example
and service levels to commit to; and how of this model. They may be radically new in
much it costs to deliver and maintain the sectors where customers are accustomed
outsourced solution.97/98 If the supplier to buying products, not services. The
wishes to, it can give its customers access motives for the customer to move to a
to relevant data and AI assets as part of the capacity-based model are similar to the
outcome delivery, to prove that its services buying outcomes model already discussed:
are running smoothly or to demonstrate any outsourcing, reducing capital usage and
service improvements. mitigation of risks related to product use.90

Selling outcomes can apply to many While selling capacity is eminently suitable
different businesses. Agriculture machinery for B2B companies, some consumer
provider John Deere has personalised businesses also do it; it can be a viable
its services to farmers by collecting data model for any kind of company. ShareNow,
from them, then selling outcomes such as a car-sharing service that is wholly owned
more efficient operation management and by the car manufacturer BMW, is a good
an increase in crop yields.97 GE Power, the example. Customers can buy the right to
power generation and water technologies drive a car from one place to another and
division of General Electric, offers pay per minute, based on the time spent
guaranteed improvements in efficiency driving.
or output according to key performance
indicators provided by its power producer
As with selling outcomes, data and AI play costs related to guaranteeing outcomes
crucial roles in selling capacity. The better and capacity.
a company can predict its customer needs • Alternatively, there might be a fixed
(e.g. location, volume of usage, type of monthly or yearly fee for services
usage), the more accurate its forecast of provided, as part of the total
how the products or services provided work invoicing, complemented by a variable
in practice (e.g. wearing out, depreciation performance-based element.
of value over time, need for insurance • One-off fees are perhaps more
and maintenance), and the greater the appropriate if capacity usage is
opportunities to optimise the capacity sporadic and irregular, e.g. pay-per-
model for the customer’s benefit. As with hour or pay-per-day packages for the
the selling outcomes model, the company customer.
selling capacity may choose to give the • Finally, a more disruptive model worth
customer access to some of its data and testing might be a business model
AI assets here as part of the transaction, overturn: the company selling capacity
allowing the customer to further optimise provides equipment, software and
its operations. service performance or capacity to the
customer for free, in order to receive
Revenue models: Selling outcomes and valuable data in return that could be
capacity can be monetised in several ways: commercialised elsewhere.
• The delivery of the service is
commonly tied to certain jointly agreed It is increasingly common for business
performance-related KPIs that the leaders to want to pay for guaranteed
company selling needs to meet. The outcomes or capacity, instead of owning
company could sell monthly or annual products or tools. So for the company
subscriptions/licences that are tied selling outcomes or capacity, it makes sense
to KPIs around capacity, too. In both to offer revenue models that highlight
cases, Service Level Agreement-type and monetise the end result delivered as a
conditions can be applied. service.
• There might be a risk and reward
model in place for the business selling
the outcome or capacity.83 This could
incentivise the creation of better-than-
expected outcomes or, alternatively,
delivery of superior capacity to the
customer; but it would also penalise any
failures or delays in service.
• Various guarantees – even insurance-
like ones – can apply to ensure that the
customer is happy. For the seller, the
key is to learn to manage the risks and
In this chapter we have presented six Selling data, analyses and reports, selling
different potential data and AI-enabled data platforms and selling AI assets:
product/service offering models that These are product-oriented models. While
explore answers to the question “what is some services might be included in the
being sold?” in business model renewal. concept or solution that is sold to the
The objective here is not to encourage customer, data and AI assets are, to a large
companies to take a technology-focused degree, seen as a product; so the revenue
approach, or simply focus on pushing models related to product sales apply quite
products to the marketplace. Instead, we well. The business selling data says:
want to help business leaders understand
how they can solve important problems , so the product offering is
worth solving (as discussed in Chapter 2.3), standardised or even sold in online self-
by offering data and AI assets to various service channels.
customers or partners in a creative,
meaningful way. Selling data and AI-based services, selling
outcomes or capacity: These are service-
The data and AI-enabled product/service oriented models. In these models, data
models can be further divided into three and AI are one critical element of the final
logical categories, as follows (see also concept or solution that is sold, but not the
Figure 22). only one. Selling these services would be
difficult without proper data and intelligent
Selling existing product/service systems. Yet what the customer sees is a
offerings: The product or service is what service delivered, monetised and charged in
it is, with only an incremental change in some way. There is often a human involved,
terms of the business model. Data and tailoring the solution to the specific needs
AI are in the background as an enabler: of a customer (the earlier Kalmar and Lassila
they may be visible to customers in the & Tikanoja examples illustrate this well).
customer interactions, or they may not. Selling these solutions in online self-service
These services are typically bundled into channels is not impossible (think of the
the price of the original product unless ShareNow example), but it’s not common
there are reasons to set a separate price either, especially if the customer is a large
e.g. for some more advanced version of the company and the solution delivered is
product/service offering. complex.
 Selling existing Selling data,
Selling data
product/service analyses and
platforms
offering reports

Selling data and Selling outcomes

Selling AI assets
AI-based services or capacity

Product-oriented models Service-oriented models

Figure 22: Data and AI-enabled product/service offerings categorised (source: Mika Ruokonen)

This categorisation is merely indicative

and not mandatory or “black and white”.
For many companies, a change from selling
products to selling services (or vice-versa)
might be a profound and long-lasting
transformation that requires extensive
building of completely new capabilities and
careful change management. Moving further,
to selling outcomes or capacity, is again a
significant step towards new frontiers.

Finally, a major shift would be to allow

AI and machines to decide what the end
product/service and revenue model should
be (we humans love control!) There are few
examples of companies that have jumped
directly from product sales to selling
outcomes or capacity, let alone to a fully
automated and optimised business model.
But nothing prevents businesses from doing
this if their capabilities are nimble enough
and can support these changes.
For each of the six AI and data-enabled AI asset. Usually there is a recurring
offerings we have discussed, there are and predictable monthly or yearly fee
also revenue model considerations. This is charged. This is based on the number of
the other side of the coin in our analysis: it users or the amount or scope of usage.
provides an answer to the question “how • Marketplace, platform, brokerage,
is money made?” Here’s a recap of the performance: The company selling data
revenue models we introduced: controls and orchestrates data flows on
• Free, freemium: The company selling a platform/marketplace/service delivery.
data or AI assets gives them out for It charges customers a data access
free, or bundles them with an existing fee, a handling fee, a brokerage fee and
product or service. Alternatively, it provision or commission per transaction.
offers a subset of features for free and The fee can be fixed or a percentage
tries to convert free users to payers value attached to specific KPIs.
with an advanced version of the product. • Business model overturn: The company
• Product/transaction revenue: The selling gives an existing service for free
company selling charges a fee per each but charges for data and AI assets. The
product sold. For instance, there is a fee key here is to allow use of the original
for each dataset or report, or for a piece product without a charge in order
of algorithm sold to a customer. to collect the maximum number of
• Project revenue: The company selling customers and their data, then sell that
charges a fee per data analysis or AI- data to interested parties.90
related project delivered to a customer.
The final fee depends on project scope, As we discussed earlier, the licensing
duration, team size or other factors. revenue model is a logical and versatile
• Advertising revenue: The company option. Wherever there is continuous
selling uses data and AI to identify provision of data and/or AI assets, it is
target groups for advertising and to relatively simple for the company selling
optimise advertising space/frequency. to set up a monthly subscription model
It charges a fee based on ads shown, that delivers a scalable, recurring and
clicks by the user, or leads generated stable flow of money, in return for access
for the ad buyer. and usage rights. In addition, the licensing
• Licensing, subscription: The company revenue model might work as a continuous
selling charges the customer for access retainer or a fixed stable part of the fee
to a dataset/dashboard/platform or in service-oriented models such as selling
data and AI-based services, and selling The business model overturn deserves
outcomes or capacity. more of our attention as a source of
revenue. A prime example of this model is
In the B2C world, subscriptions are the Google: its search service, the core of its
norm – take Netflix, Spotify or digital operations, has always been free of charge,
newspapers for instance – which means but it has been used very effectively to
they will be familiar to customers in a create a highly successful advertising
professional context, too. business through AdSense and other
similar products. The core business is
The challenge of the licensing model lies in to attract audiences who generate data
persuading the customer to start buying while using the search, and right after
in the first place (a pressure that the the search the same audiences become a
freemium model mostly sidesteps) and target for advertising. The implementation
also in keeping the customer continuously of this model was intentional and data
satisfied (something that the use of commercialisation was by design right from
product or project revenue models can help the start. Google could have simply sold
to avoid). But overall, a licensing model is its raw data but instead the company built
risk-averse. The company selling has access new data-enabled businesses on top of it to
to a scalable, recurring and stable model. create even more value for its advertising
And the customer has control: they can customers.31 This type of a revenue model
generally end the contract whenever they is not entirely new: radio channels or
choose. newspapers have also provided content
for consumers for free in order to sell
The advertising revenue model will be advertising space. Yet, Google and other
challenging in some cases. A lot of users similar digital media firms have brought
are required in order to generate revenues particularly strong data and AI elements
from digital advertising on data or AI to the model: specifically with the efficient
platforms: a large number of people need targeting of online ads.
to see and click on the ads. This is harder
to achieve in B2B compared with B2C, For many business leaders, a business
given the low prices of digital ads, and model overturn is a mindset issue:
the tendency of the prices to decrease
over time (exceptions here are Facebook
and Google ad prices, which seem to
fluctuate every now and then). That said,
the advertising revenue model might Out of all the revenue models presented in
complement other revenue sources. A this chapter, the business model overturn
company may charge a licence fee for its is the most fresh and radical option. Yet,
data and AI platform usage, and then sell it could also be considered as a revamped
ads as well. freemium model, in which giving something
for free (the traditional service or product)
opens new business opportunities parameters such as willingness to pay,
elsewhere (selling data and AI assets). So, demand, internal capabilities, restrictions
why not merge these two models? The and ability to deliver, and then allows
reason is that the business model overturn algorithms to set and optimise prices
is completely dependent on a company‘s that are shown to the customers. This
data and AI assets whereas freemium is allows the business selling data to tap into
not, and can be delivered even without any profitability gains that might be hidden by
data or AI. The business model overturn as fixed or rules-based pricing. Typically, to
a distinct new model reveals the disruptive run dynamic pricing, the company doing
nature of data and AI, so it makes sense the selling needs relevant data on its
to position it as a separate item in our customers, data on its own capabilities
playbook. to deliver product/service offerings and
a self-learning algorithm that bases its
When choosing revenue models, it’s price recommendations on historical
important to discuss price perceptions: data.100 Given that this book is about data
• What is the perceived value of data and and AI-enabled business models, and
AI to the customer? not about price-setting as such, we will
• How high could the prices be, while not dig deeper into the topic of dynamic
remaining commercially feasible and pricing. But if a firm wants to use data
acceptable to customers? and AI to increase its revenues without
• If something is given for free now (e.g. launching new product/service offerings or
in the freemium and business model revenue models, dynamic pricing might be
overturn examples), is it possible to something to explore.
charge for it later?
• If something is given for free, does that
massively increase the ability of the
company selling to quickly gain more
data and AI users, so helping it to claim
a first-mover advantage and a dominant
position?

Business leaders must make some tough

decisions when setting a price that
is attractive for customers, yet also
commercially robust.

Finally, when talking about how data and

AI might change “how money is made”, one
potential tool for companies to consider
is dynamic pricing. Here, the supplier
uses data and algorithms to study various
AI is a disruptive technology. It can help Figure 23 represents a playbook of the
companies change the rules of the game combinations of product/service offerings
in their industry, and allows them to and revenue models we have discussed.
radically rethink their business models – if These are divided into two categories;
doing so proves necessary or lucrative. “what is being sold” and “how money is
So, in addition to emphasising existing made”. Some options may be less feasible
strengths, implementing data and AI is than others, but they are still worth testing
also an excellent opportunity to challenge with potential customers. The idea is to
orthodox thinking and look for completely highlight the combinations that are most
new approaches. This is the beauty of data attractive for data and AI as assets.
and AI in business.
 “What is being sold”

“How money EXISTING

DATA, DATA AND OUTCOMES
is made” PRODUCT/
ANALYSES
DATA
AI ASSETS AI-BASED OR
SERVICE PLATFORMS
AND REPORTS SERVICES CAPACITIES
OFFERING

depends
on existing
FREE,
product/
FREEMIUM
service
offering

depends
PRODUCT/ on existing e.g. a data
one-off use
TRANSACTION product/ dump or one one-off
of capacity
REVENUE service report
offering

depends e.g. how

on existing much work
PROJECT one-off per hour/
product/ is needed
REVENUE project day
service to deliver
offering assets

depends
on existing
ADVERTISING
product/
REVENUE
service
offering

depends
continuous
on existing
LICENSING, data, analy- continuous continuous e.g. a fixed
product/
SUBSCRIPTION sis or reports use retainer part of fee
service
provision
offering

e.g. at-
depends
MARKETPLACE, tached to perfor-
on existing e.g. attached attached to
PLATFORM, the size of mance
product/ to number of the improve-
BROKERAGE, the revenue against de-
service transactions ments gained
PERFORMANCE generation fined KPIs
offering
opportunity

BUSINESS sales of
MODEL additional
OVERTURN services

Likely model A model worth testing

Figure 23: The playbook of data and AI-enabled product/

service offerings and revenue models (source: Mika Ruokonen)
When it comes to both product/service groups (for example, selling raw data to
offerings and revenue model options, partners versus selling data platforms
there is no particular order of importance, directly to customers). There might be
maturity or feasibility, nor a predefined path different jobs to be done in the marketplace,
that a company should automatically follow. requiring distinct models each time.
The point is simply to illustrate the vast
range of alternatives that any given business Likewise, various revenue model options
has for its data and AI-enabled business (seen in the rows of Figure 23) could be
model renewal, providing inspiration for used in parallel as well: for example, giving
future discussions with customers. some data and AI assets for free (using
the freemium model) and also charging for
What if an innovative company aspires to some of them (using the licensing model)
combine data and AI-enabled services? or applying data to deliver services (using
Thinking creatively, what if it strives to the outcomes and capacity sales model).
build a platform, on which to sell AI assets A company’s data and AI assets are likely
through a model that guarantees AI asset to evolve constantly, so whatever revenue
capacity to its customers? Which model is model is chosen, it should cater to the
this? Well, that depends on your point of dynamic nature of data and AI-enabled
view. One could argue that the platform products and services, allowing for an
is the core product being sold in this case incremental change over time.101
as it provides the environment from which
business is generated. Someone else might ---
argue that,
, or they
might point out, There are various obstacles for companies
. There’s no when exploring new business models
right or wrong when it comes to data and and moving from one to another. Often, a
AI services. They don’t necessarily need to business case for a new model might be
rule each other out: various combinations uncertain or difficult to build as there is
could work. The key is to identify the job to limited understanding and evidence as
be done at the customer end, then design a to the implications of the change. What’s
service that matches that brief in the most more, the new business models might not
effective and meaningful way. constitute “easy wins”, but instead demand
long-term commitment. This may include
Often, the product/service offering options accepting certain reputational or financial
(seen in the columns of Figure 23) do not risks, investing in data and AI maturity,
exclude each other. Instead a company can introducing a culture change, and so on.6
offer several of them in parallel (for example, Finally, a company might not be in a credible
by selling both raw data and reports to or trusted position to sell both its existing
customers, depending on what they need and services and its data and AI-enabled
when), to the same or different purchasing
concepts under the same brand or legal Often, deciding which data and AI-enabled
entity. business models to use is a choice that
does not fall exclusively to the company
Revenue cannibalisation is also a common creating the product/service offering.
issue with data and AI-enabled product/ Many consumer-facing firms, for instance,
service offerings and revenue models. If a struggle to gain access to customer data
company’s existing business model is solid, because their products are sold by resellers
e.g. selling products, does it make sense who control the customer interface and
to start selling outcomes or capacity? keep the customer data. Also, many
Or to try a business model overturn, or industrial firms do not operate in a vacuum
implement a freemium model? If there is but instead need several partners to deliver
no external pressure to do so, then the their solutions to their customers. What’s
answer might be “no”. But if new or existing more, some companies use other firms’
competitors are about to disrupt the platforms as a “carrier” for their datasets
industry, the answer is more likely to be that are then commercialised with the end-
“yes”. It makes sense to cannibalise your customers (take the Eagle Alpha example,
own business before someone else comes or the Amazon Web Services example102 in
and cannibalises it instead. Understandably, the references of this book). The ecosystem
these management decisions are often and partnership play therefore becomes
tough to make, given the uncertainty and crucial.37/38/103 Business leaders should
risks related to cannibalisation. consider:
• What data and AI assets do we have in
It is important to consider who is in the our firm? What assets do our partners
driver’s seat when it comes to data and (or competitors) have?
AI-enabled business models. Is it the • What would be the optimal, unique and
customers creating a market pull as they valuable solution for the end-customers,
eagerly await new products, services and given the ecosystem and the potential
revenue models? Or is the implementation partnerships? (Note: one typical issue
of data models a market-making effort is that the end-customer wishes to
by the company selling, who wishes to deal with one partner only and wants
experiment with new models even though to avoid logging into various systems
the customer has no clear perceived need or using various applications to see
for them (yet)? As discussed in Chapter relevant data).
2.4, when describing the customer’s • What is the system for bringing various
perspective, both options are possible. pieces of data and algorithms together
From the point of view of the business in the ecosystem (e.g. uploading, data
selling data and AI-enabled solutions, it pipelines, common tools)? How can we
makes sense to explore, test and co-create make systems and data flows work
new models together with customers, to together to deliver what the customer
gain more understanding of their needs and needs?
preferences.
• How can we build revenue models at develop more capabilities that allow their
different points in the ecosystem so customers to harness their data assets in
that the contributing players get their new or better ways going forward. This is an
fair share of the value generated? area where it’s worth seeking legal advice.
• How does the ecosystem evolve and
grow over time so that it continues to In some cases, data and AI assets are
deliver added value, using data and AI, traded between two parties without money
and is hard for competitors to replicate? changing hands. In this context, data and
AI move beyond sellable items to become
When designing data and AI-related tradable assets. Here are some questions to
offering and revenue models, it is vital to consider in this scenario:
understand the business landscape in a • If a company gives data and AI assets
given industry. Key players within individual to another firm, what similar or
companies and the wider industry sector complementary assets could it receive in
must build interoperability and interfaces return?
to other relevant systems, and design and • If a consumer receives data and AI tools
implement data flows between parties to free of charge from a company, would
make sure that the data and AI-related goals they be willing to give their own data for
of the business are adequately met. The free in return?
more dominant and connected the company • Would the two traded assets be of equal
is in its ecosystem and market, the more value?
data and AI-enabled business options it will • Would the combination of these two
have access to. assets form the 1+1 = 3 equation?
• Would it be beneficial for various parties
To make the situation more complex, not to allow data and AI to be traded without
all companies use their own data and AI payment?
capabilities when building new business
models. A significant number of companies Many of the data and AI-enabled product/
are using their cloud providers (Azure, service offering models listed in this book
Amazon Web Services, Google Cloud, and could apply to asset trading, although the
the like) to deliver data warehousing, revenue models we’ve discussed generally
processing and analytics tools, instead of do not.
building their own models. These companies
need to read the contractual terms of their
cloud supplier agreements very carefully
to understand whether they limit “what
is being sold” and “how money is made”, To summarise the key learnings of this
especially if the companies plan to trigger chapter: when it comes to data and AI as
industry shifts and enter completely new an enabler, it looks as though the bigger
markets with completely new products. impact is on “what is being sold” (the
It’s likely that cloud providers will strive to product/service offerings) as opposed to
the “how money is made” (revenue models).
Data and AI provide companies with a new
set of technologies, tools and capabilities
to renew and amplify their product/service
offerings, yet it seems that the existing
ways of making money remain largely
unchanged.

The reason for this may be simplicity. The

easier a revenue model is to explain to
customers, the easier it is to buy/adopt.
Companies may not wish to make their
revenue models too complicated because
– no matter what is being sold – they want
their customers to know what they are
paying for. Alternatively, data and AI might be
such a new phenomenon for many businesses
that they haven’t yet worked out what the
impact on revenue capture might be.

For any business planning to harness data

and AI, understanding emerging business
models is important. The playbook we discuss
is designed to help by creating new insights
and providing a structure for fresh and
creative approaches.

In the next chapter, we move from the

playbook towards tangible examples and
case studies that show how companies have
applied these models in practice.
6 Co-creating data
and AI-enabled

So far, we have explored how data and AI work and how

Typically the potential of data and AI-enabled Our approach is different: we have
business is demonstrated via case studies deliberately focused on studying Europe-
involving large Silicon Valley companies based companies predominantly from more
such as Google, Facebook, Uber, Netflix and traditional industries.
Amazon. However, the success of these
technology behemoths has little relevance
to more traditional companies based outside
Silicon Valley that do not have access to the
same level of financial resource, data science
expertise and technological competence.

GROWTH REINVENTED 105

 Co-creation
methodology
To test and validate our data and AI-enabled conglomerates with more than 100,000
business model playbook, in Spring 2020 employees.
we organised 20 co-creation sessions • When describing their experience in
with companies from various industries digitalisation starting from very basic
and different European countries. Our knowledge all the way to fully-digital
hope was that these companies would operations, we applied the following
reveal important insights that we could categorisations: Beginner – Exploring –
use to assess whether our approach was Transforming – Experienced – Digital by
meaningful and relevant. definition.
• With “target customers” and “offerings
• Various industry sectors were covered. focus” we refer to Chapter 4.3, in which
A maximum of two companies per we discussed industry characteristics
industry sector were included in the as a starting point for data and AI
sample to ensure enough industry business models. As we can see from
variance. In cases where two companies Figure 24, various customer-product/
came from the same industry sector service offering combinations were
there were significant differences when included in the sample.
it came to how they implemented data • Finally, referring to Chapter 1.6, in
and AI-enabled business models. which we discussed how companies can
• The size and geographical scope of be in very different starting positions
the companies varied greatly, from strategically, in our sample we identified
small local or regional companies three major groupings: traditional
employing dozens of people to global companies, companies undergoing
 INDUSTRY SEGMENT GEOGRAPHICAL SCOPE NUMBER OF EMPLOYEES

Company A Equipment manufacturing Global ~700

Company B Digital media Local ~100

Company C Digital platform Regional 30+

Company D Telecoms Regional 20 000+

Company E Healthcare Local 500+

Company F Automotive Global 100 000

Company G Automotive Global 300 000

Company H Construction Local 300+

Company I Machinery Global 10 000+

Company J Financial services Regional 10 000+

Company K Consulting Local 70

Company L Healthcare Local 5 000+

Company M Insurance Local 1 500

Company N Energy Local 1 000

Company O High-tech Regional 50+

Company P Cleantech Global 450

Company Q Energy Regional 20 000+

Company R Machinery Global 500

Company S Equipment manufacturing Global 1 800

Company T Raw materials Regional 5000+

Figure 24: A list of companies we held co-creation

sessions with (source: Mika Ruokonen)
 STATUS IN DIGITAL
TARGET CUSTOMERS OFFERING FOCUS
TRANSFORMATION

A Transforming Companies Products

B Digital by definition Consumers & companies Products

C Digital by definition Consumers & companies Products

D Experienced Consumers & companies Products & services

E Exploring Companies Services

F Exploring Consumers Products

G Exploring Consumers & companies Products & services

H Exploring Consumers & companies Services

I Transforming Companies Products & services

J Experienced Consumers & companies Services

K Beginner Companies Services

L Experienced Consumers & companies Services

M Beginner Consumers & companies Services

N Exploring Consumers & companies Services

O Exploring Companies Products

P Beginner Companies Products & services

Q Transforming Consumers & companies Services

R Transforming Companies Products & services

S Exploring Companies Products

T Exploring Consumers & companies Products & services

 digital transformation and data and whenever they became relevant in line with
AI-native companies. the company’s specific circumstances and
situation.
The companies involved in the co-creation
sessions varied significantly in terms In a few sessions, only one person from
of their profile and digital maturity. the company was present, but typically
The resulting sample provided us with there were between two and five company
a valuable window into the thinking of representatives discussing the topic. The
business leaders from a wide range of most comprehensive setups included more
backgrounds. than 10 people in the room. Participants
included many C-level executives as well
The shortest co-creation sessions were as business development executives, VPs
sparring sessions and discussions that of new business, heads of data science
lasted for roughly 1.5 hours, while the and data science team leads. The sessions
most extensive co-creation project was a turned out to be a good opportunity to
series of deep dive workshops that lasted bring business-minded people together
a total of four days. With a few companies, with data science specialists.
there were two consecutive sessions; the
first one to discuss data and AI-enabled
business models in general and the second
one to dig deeper into the company’s
specific opportunities and challenges. Most
of the sessions were held face-to-face
immediately before the Covid-19 lockdown
came into force. Only a few sessions were
held via a remote video call link.

Appendix 4 summarises the discussion

themes that were used in the co-creation
sessions. First, we briefly reviewed the
current situation in the company, then we
started to co-create data and AI-enabled
product/service offerings and revenue
models, and finally we had a closing
discussion in which we shared feedback
and planned the next steps. In terms of
topics to be covered we we followed a
broadly similar template: current situation
– co-creation – closing. The themes were
not discussed in a strict format, one
after another. Instead, topics were raised
For the purpose of the co-creation business model options according
sessions, we created four canvases, to which has the greatest potential.
building on Futurice’s Lean Service There are two axes: the size of the
Creation methodology.104 The four canvases desired impact (e.g. potential revenues,
comprised: societal impact) and the difficulty of the
• Business Goals & Drivers: This canvas problem to be solved (e.g. anticipated
is useful where a company is not very work effort and costs). The sweet spot
advanced in terms of planning data and is in the top left-hand corner where
AI-enabled business models. It helps a the potential impact is large, and the
company define its ambition, the impact problem can potentially be solved easily.
it aims to achieve with business models, • Data & AI-enabled Business Models:
and the potential problems worth The first three canvases are designed
solving at a high level. It also helps to to help the company structure its
identify which of the company’s data thoughts or prioritise initiatives, while
and AI assets can be commecialised. this last canvas really gets to the point,
• Customer Groups: This canvas is useful helping the firm to describe its data and
for exploring potential customer groups AI-enabled product/service offerings
for data and AI-enabled product/ and revenue models in more detail.
service offerings and revenue models. It aims to help the company identify
Companies can add several customer the specifics around anticipated value
groups, and describe their desired propositions, product/service offerings
behavioural and/or mindset change, as and revenue models for each of their
well as anticipating their needs. customer groups.
• Impact & Implementation Optimiser:
This canvas helps companies to Small diagrams of the canvases (source:
prioritise the data and AI-enabled Mika Ruokonen and Futurice) are presented
in Figure 25. If the reader wants to use the of what kinds of new product/service
canvases in their own company, Appendix offerings or revenue models could be
5 provides them in full-page layout, along created in which types of companies.
with brief instructions on how to use the Some of the sessions provided insight into
canvases in co-creation work. companies’ different strategic choices,
their business characteristics and their
During the co-creation sessions, the role levels of readiness and maturity around
of the canvases was to help the companies commercialising their data and AI assets,
taking part to uncover business goals, as and how these kinds of starting points
well as to identify potential target groups, impacted business model implementation
value propositions, revenue models and in practice. Some of them helped us to see
implementation opportunities related to which data and AI-enabled business models
data and AI-enabled business models. In our work, and the steps needed to innovate
study sessions, the canvases were printed and implement them. Finally, some of the
out in A1 size and Post-it notes were used to co-creation sessions helped us to refine
fill them with ideas and to summarise key the canvases and create tools to be used in
conclusions. future sessions.

We were selective about how we used the Finally, we also held a co-creation session at
canvases: in some sessions we skipped Futurice. As a consulting firm, the business
the Business Goals and Drivers canvas is built around recruiting and retaining
entirely as its content was self-evident to talented people, delivering projects to
the people in the room. Instead we focused customers, and continuously innovating
on evaluating the best problems to be methodologies and improving practices.
solved with the Impact and Implementation These operations generate a lot of data
Optimiser canvas, and designed solutions on an on-going basis. In the Futurice
with the Data and AI-enabled Business co-creation session, we talked about the
Models canvas. In other sessions we opportunity to deliver data and insights
avoided using the canvases altogether to our existing customers, to help them
because the discussion flowed seamlessly to see which projects were run together,
and we tackled key points without them. who was involved and how methodologies
The role of the canvases was to structure supported the work in practice, and
the discussion and inspire people to helped our customers to learn new skills.
generate new ideas whenever relevant. It was refreshing to hear both how other
companies are tackling data and AI-enabled
The contribution of the co-creation sessions business models and also to explore how
was multifaceted. Some of them provided our own business could be transformed
interesting case studies that are shown using this kind of thinking.
in Appendix 6. Some of them helped us to
evaluate how our playbook could be applied,
by helping to further our understanding
Figure 25: Images of the canvases used in our co-creation
sessions, (source: Mika Ruokonen & Futurice)
Appendix 6 includes seven case studies of • These companies provided a good
companies that took part in our co-creation cross-section of the entire sample of
sessions. Most of the case studies (#1-5 our study. They include small and large
and #7) feature one company, however firms, companies at the beginning
case study #6 includes insights from two of their digital transformation and
automotive industry companies. This digital-by-definition firms, local and
explains why, while eight companies are global companies, and companies from
listed in Figure 26, there are only seven case different industry segments.
studies as one of the case studies features • Not all co-creation sessions in our
two companies. sample provided new information:
repeat themes started to emerge
Figure 26 summarises the companies that had previously been discussed
included in the case studies (note that elsewhere. The seven case studies
these are the first eight companies listed in reflected and highlighted the key points
Figure 24). of all the co-creation sessions while
avoiding overkill.
Why were these case studies selected as
opposed to other companies in our sample? All the case studies featured in our study
And why only seven case studies, not more? are anonymised to protect company
There are several reasons: confidentiality.
• These seven case studies provided the
best practical examples of companies For a genuinely deep dive, the reader is
on a journey to build data and AI- strongly encouraged to read the seven
enabled business models: so they case studies in Appendix 6 which provides a
offered the most illuminating insights powerful insight into the opportunities and
from our sample group. Crucially, challenges that companies encounter when
these companies did not just provide implementing their data and AI-enabled
a “polished” view of their data and AI business models.
initiatives, they also shared specific
challenges that they had encountered
on their journeys. This made them
particularly valuable case study
candidates.
 COMPANY

INDUSTRY GEOGRAPHICAL NUMBER OF STATUS IN DIGITAL TARGET OFFERING

SEGMENT SCOPE EMPLOYEES TRANSFORMATION CUSTOMERS FOCUS

Equipment
A manufacturing
Global ~700 Transforming Companies Products

Consumers &
B Digital media Local ~100 Digital by definition
companies
Products

Digital Consumers &

C platform
Regional 30+ Digital by definition
companies
Products

Consumers & Products

D Telecoms Regional 20 000+ Experienced
companies & services

E Healthcare Local 500+ Exploring Companies Services

F Automotive Global 100 000 Exploring Consumers Products

Consumers & Products

G Automotive Global 300 000 Exploring
companies & services

Consumers &
H Construction Local 300+ Exploring
companies
Services

Figure 26: A list of companies that were selected to be included in the case studies
7 Closing the loop:

insights from

sessions
For example, telecoms Company D saw data related efforts and opened new growth
and AI as a logical and relevant new frontier opportunities.
that built on its existing digitalisation-
A similar view was also evident in other commercialise every asset it has (including
large firms. Many large companies data and AI) and to prove the financial
expressed concern that their existing impact of new business models as fast as
business was under pressure; for instance, possible. These companies are typically in a
prices were eroding, competition was rush and in turmoil.
intensifying and customer expectations
and attitudes were changing. The feeling In addition to the two groups already
that “we need to do something” was often discussed, we also identified a third group
present in large firms (see Chapter 1.1): in co-creation sessions: firms that planned
they were forced to actively look for new to further accelerate their growth using
business opportunities, and data and AI data and AI. Healthcare service provider
were seen as valuable assets in this process. Company E is a prime example as it
At the same time, it is important to note proactively searched for growth avenues,
that among the large firms we spoke to, even though its core business was already
none of their existing business was about healthy and growing.
to “fall off a cliff”. Instead, they anticipated
a gradual and slow erosion, which reduced Without exception, all the companies
the sense of urgency to quickly innovate considered data and AI-enabled business
new business models. This was especially models a matter of strategic importance
obvious in automotive Companies F and (see Chapter 1.4). It was something that
G. There isn’t always enough pressure to they were seriously considering, and not
warrant a fundamental transformation. only as a sideline or an operationally small
It’s worth noting here that almost all of the issue (the insurance company was the
co-creation sessions were held before the only exception. It saw data harnessing as
Covid-19 outbreak; the pandemic may well a strategic issue, yet the regulation of the
have altered leaders’ perceptions of how industry, as well as the rather traditional
stable their businesses are. company culture, made it difficult to do
much in practice). Of course, there is a risk
By contrast, small companies, especially of selection bias in our study approach: it’s
digital-by-definition ones, were often possible that only those companies that
entirely built around their data and AI consider data and AI-enabled innovation
assets. In these cases, building data to be of strategic importance, volunteered
and AI-enabled business models was not to take part in the co-creation sessions.
really a strategic choice but instead a Another potential selection bias in our
fundamental necessity and the name of sample is that the employees involved in
their game. Yet, small firms were also the co-creation sessions often held a data
actively looking to expand their current and AI-related role in their respective
business, spurred on by increasing use organisations. So they had a tendency to
of data and AI. It is easy to appreciate consider all data and AI-related topics as
how a small startup that might soon run being strategically important.
out of cash will do its utmost to quickly
The issues of data and AI-enabled business better financially compared with
models resonated well with the companies companies that are not investing.
that took part in our co-creation sessions. • It was hard to judge whether new data
It is an area that is urgent, relevant, and and AI-enabled models have led to
interesting to many forward-looking strong revenue growth (the positive
European business leaders. Companies impact) versus revenue cannibalisation
often see an opportunity to use data and elsewhere in the business (the negative
AI to get closer to their customers and impact); or under what conditions
to build deeper and more long-standing revenue growth would outweigh
relationships with them, so creating potential revenue cannibalisation.
significant added value. Companies
are also moving beyond using data and More time – and perhaps a follow-up study
AI in their existing businesses (e.g. to – is required to identify any concrete
achieve operational improvement or commercial successes brought about by the
efficiency gains) and are starting to implementation of new data and AI-enabled
consider how these assets can be used product/service offerings and revenue
to build new products, innovations and models.
growth opportunities that will deliver real
competitive advantage.

However, this step forward is not always

easy. As one participant put it:

“We have been working on data and AI a lot

to focus on new business generation has been a

It is clear that many companies are still on

a journey, and are not yet ready to fully
implement data and AI growth models:
• We did not find strong evidence that
the firms we talked to have successfully
created a “second leg” for their
business using data and AI offerings
and revenue models; nor that companies
investing in these models perform
 How

As suggested in Chapter 2, the business new product/service offerings and growth

leaders we spoke to increasingly saw opportunities.
data and AI not only as assets to improve
internal operations, but also as tools to Many companies had large repositories of
help to build new services and revenue relevant data, and, as such, data collection
models. Data and AI-related matters was not really an issue. But many were also
seem to be a hot topic right now. In our disappointed by the large amount of work
co-creation sessions we were never needed to consolidate, process and analyse
required to define data or AI as concepts, their data before it was ready for any AI
or explain what they can enable in a applications (see also Appendix 1). “We have
business context. so much data, but we do not yet get enough value
was a typical response. Datasets
These concepts and their applications in automotive Companies F and G were also
were already quite familiar to the people scattered around the large organisations, so
in the room. Therefore we could jump the first step with them was to understand
directly to talking about data and AI as which assets could be harnessed, and who
would need to come together to advance Companies tended to ask questions such as,
data-related opportunities.
when the
There was not much discussion around the obvious answer would have been, “Go and talk
ethics of using data for business purposes,
but often legal matters were addressed (we
return to these in Chapter 8.2). Business This standard approach to
leaders were typically well aware of laws service design work is typical for all digital
regulating business and data use in their development, not just data and AI.
given industry. There was often a desire
to operate as a respected, trusted and One of the two energy companies we
responsible player when it came to applying worked with was a clear exception to
data and AI. Digital media Company B, in the rule: it heavily emphasised problems
particular, emphasised careful contracting worth solving – reminding us that we
as a key element of any data and AI-related should be cautious about making broad
business, and wanted to restrict the use of generalisations.
its assets by customers in order to avoid
reputational risks or potential litigation. The In addition, making a business case for data
insurance company, meanwhile, stated that and AI-enabled product/service offerings
new business models were a difficult topic and revenue models was not always
in general, given the heavy regulation and common practice in companies: many found
control mechanisms in place in the financial it easy to embark on various pilot projects
sector around harnessing data and AI. without having any commercial plans in
Overall, the companies in our sample could place.
be described as risk-averse whenever it
came to ethical and legal questions around It’s possible that some of the companies
data and AI. in the sessions lacked the skills to create
comprehensive and convincing business
Designing new data and AI-enabled cases for their data and AI-related
solutions (as described in Chapter initiatives. It was difficult for them to judge
2.3.) was generally a new concept for who would purchase and pay for data and
many companies. Many were somewhat AI-enabled product/service offerings; what
technology- and inward-oriented, which the final product/service offering would be;
is not necessarily a bad thing, because and what revenue models and price points
companies need to realistically assess would look like for them.
which assets they can commercialise
and which can wait. But often we needed While this shortage of customer-centred
to stress the importance of looking at and commercially focused thinking was not
things from the customers’ perspective, necessarily a major flaw, it was a surprising
identifying problems worth solving and outcome of the co-creation sessions:
thinking of jobs to be done.
• A more multifunctional approach to range of perspectives. The discussions
data and AI business models, including could represent the views of the business
tech, design and business perspectives, selling data, their customers and their
would have been helpful in many firms. end-users, all at the same time. This
• There were some instances in the co- approach would ensure that the customer
creation sessions where data science perspective is front of mind, allowing an
people and business people from the instant evaluation of the value created for
same company had never met before, subsequent business planning purposes.
which demonstrates the need for a more
cohesive cross-team approach. In a handful of firms (including digital
• The setup of the co-creation sessions media Company B, the raw materials
mattered a lot in this respect: potential provider company and one of the machinery
customers were in the room for just companies) potential customers took
one session. For the rest of the time, a proactive approach, by asking for
it was just internal company people, data or AI assets, new services and new
representing the perspective of the revenue models (see Chapter 2.4). But
company selling data and AI-enabled most companies in our sample had to
solutions. make the first move by educating their
• The canvases we prepared for the co- customers about how to harness data
creation sessions were useful in this and AI for commercial purposes in their
regard, because filling them out forced respective organisations. In these cases,
the companies and the various people the market seemed immature in regards
in the room to think of their customers, to adopting data and AI-enabled product/
and the value created for them. service offerings and revenue models,
forcing the companies selling AI and data
The co-creation sessions were eye-openers to stimulate market creation before new
for many business leaders in our sample. business could really take off. Customers
Our discussions encouraged them to seemed to welcome new product and
approach data and AI in a multifunctional revenue offerings and early feedback from
way, taking into account service design and them was typically positive. So provided
commercial perspectives. It also persuaded that firms manage their market-creation
them of the need for more customer/buyer activities successfully, they have the
validation in order to succeed. Several potential to create a first-mover advantage
companies concluded that more thorough and a unique market position that could be
business planning would be a logical next difficult for competitors to challenge.
step.

Going forward, it would be useful to

organise sessions in which companies and
their customers come together to discuss
data/AI commercialisation from a broader
The concept of a business model was business model in general, or reviewing the
familiar to the participants in the co- entire Business Model Canvas (as in Chapter
creation sessions, and did not require 3.2). This strategic choice regarding the
much explanation. We sensed that business co-creation setup means our findings from
leaders were very eager to talk about data the sessions are mostly focused on specific
and AI as technologies, but they were often data and AI-enabled product/service
not so keen to discuss business models offerings and revenue models rather than
as tools to structure their thoughts. In on business models as a general concept.
our sessions, we often jumped straight to
talking about product/service offerings As there was a large variety of data and
and revenue model options available (per AI-enabled product/service offering
the playbook of Chapter 5, and the four and revenue model configurations (see
canvases laid out in Chapter 6) without Chapters 3.2 and 3.3), it was relatively
thinking about the various elements of a straightforward for our partners to
innovate them, and there was plenty of Internal communication and information
room for imagination in the co-creation sharing was often missing, and the
sessions. The exception was the case study organisational structure did not
featuring the two automotive Companies encourage people to work on data and
F and G: both struggled to come up with AI in a holistic way. Internal politicking
ideas. Typically there were no predefined and lobbying was also required to get
business models to fit with data and AI new innovative data and AI-related
assets. However, some options were seen projects approved.
as easier to implement than others. Where • While large companies generally had
relevant, to help companies prioritise plenty of resources, they tended to
we used the Impact and Implementation get stuck when it came to getting their
Optimiser canvas. business model innovation projects off
the ground and they struggled to do
While our sample of 20 firms does not allow things quickly.
us to make wide generalisations, we found • Some of the large companies openly
there were some differences between the admitted that they were looking into
small and large firms. Small firms typically partnering with small startups within
had “less to lose” as they were not yet their industry. They felt that they would
well-established. This meant that they get results more quickly by working
were often eager to rethink their business with nimble external partners. Some
approaches altogether, and to use data and large companies were also forced to
AI to reconfigure their existing business recruit entrepreneurial-minded people
models. Given the small size of the existing because they could not find enough
businesses, establishing a new “second leg” people capable of driving business
from data and AI also seemed like a realistic model innovations from within their
undertaking for these companies. That organisation.
said, small companies often struggled to
assemble sufficient resources to run both Given these obstacles, we often heard
their existing businesses and their new decision-makers in large firms say:
data and AI-enabled ventures at the same “We have this nice data and AI-related business
time (these conclusions line up well with our
discussion in Chapter 3.4).

In large firms, business model innovation Deciding whether new business model
activities sometimes seemed chaotic: innovations should be run within the
• There was often a rather siloed existing business or managed separately
approach in large firms: parallel units was a theme that came up in many of our
tended to work simultanously but co-creation sessions, for companies large
separately on their own data and and small (see Chapter 3.4). The new data
AI-related pilot projects, designing and AI-enabled product/service offerings
new products and revenue models. typically competed for resources and talent
against the company’s existing businesses.70 to succeed with data and AI, the story has
Unless they delivered immediate financial to be engaging and clear. Those in charge
gains, they tended not to get enough need to move from being reactive and
managerial attention and were seen as “too opportunistic towards a more proactive
future-oriented” and hard to understand. and strategic position. Typically when the
So, many company leaders considered plan is being put together, leaders need to
setting up a separate unit or legal entity for answer the question: “How do all these data and
their data and AI-enabled business, in order AI-related activities contribute to, and align with,
to demonstrate that they were, indeed, .
building a credible “second leg”.
Finally, it would be a good idea for all firms
Scaling up emerging data and AI-enabled to nurture business model innovation
business models was a common issue, capability. There are usual several reasons
too (see Chapter 3.5). In order to succeed why a data and AI-enabled “second leg”
in a large company context and claim fails to gain traction and these need to be
more funding and resources for scaling, tackled holistically and persistently. The
innovation teams had to show evidence more a company can make business model
and proof of early success (as already innovation an established common practice
mentioned, they often had to lobby, too). across its entire organisation, the better
This is not a surprise as providing evidence positioned it is to harness data and AI-
around a new initiative is normal business enabled new offerings and business models.
practice in large firms. Sometimes, as in the
case of the raw materials provider company
in our sample, it was a case of chicken-and-
egg. People in the co-creation session were
keen for more funding for their data and AI
initiatives, but they had very few success
stories to show to management that would
justify such investments – because they
lacked the initial financing to get the work
started.

For many companies, building a coherent

data and AI business model narrative (or
simply a business plan) would help explain
why the issue is strategically important,
and lay out the anticipated benefits. In
large firms, individuals rarely achieve much
if they cannot get senior management
on-board; let alone external stakeholders
such as customers and partners. In order
 How

Companies involved in our co-creation so the basic ambition to drive business

sessions varied in terms of ambition growth was in place. Decision-makers were
levels (see Chapter 4.1). Business leaders actively looking into building completely new
were mostly keen to start innovating and products out of their data and AI assets. At
describing new data and AI-enabled product/ the same time, many also tried to implement
service offerings and revenue models; data and AI in their existing product/service
offerings (N.B: the Covid-19 pandemic might Existing customers seemed like a logical
impacted these ambitions since then). target group to start with, as they offered
a low-risk way to explore new product/
Typically, companies saw existing service offerings and revenue models within
customers as high-priority potential a familiar setting. Automotive Companies
customers for their data and AI assets. F and G, in particular, considered their
When it came to selling data and AI assets current customers – the car owners – the
to B2B customers, the right contact within most important buyer group. That said, the
the customer organisation was typically canvases did prompt leaders to consider
someone new and unfamiliar from the several customer groups, including new
person they usually dealt with. This was ones.
considered a major challenge for the sales
teams, as they needed to start identifying None of the companies we worked with
and contacting completely new people – expressed a clear ambition to use data
even while working within their existing and AI to enter a new industry sector. The
corporate customer base. exception was healthcare service provider
Company E, which was transforming
There were also examples where completely from a local service business towards
new customers were proactively identified an international technology licensing
and targeted. This was particularly the case business through a spin-off (a bold and
in digital media Company B, digital platform transformative move). The overall lack of
provider Company C and healthcare ambition among companies was surprising,
services provider Company E. These firms given that data and AI are often expected
were either continuously responding to to revolutionise industries, blur boundaries
inbound requests from completely new and enable completely new approaches to
companies (the case for digital media business:
Company B) or considering a major business • Was this lack of ambition to enter
model overhaul (this applied to digital new industry segments due to lack
platform startup Company C and healthcare of imagination, courage, resources
services provider Company E). or open-mindedness? There were
quite strong conventions within the
In terms of the traditional Ansoff Matrix industries we sampled, potentially
(see Chapter 4.1), all the boxes in the two- making it hard for people to see beyond
by-two matrix applied to the companies the obvious, and reach outside their
in the sample, meaning that both existing existing customer base, competition and
and new product/service offerings were business parameters.
considered, as well as both existing and • Alternatively, business leaders may
new markets. Overall, there was a slight have considered that there was
preference for product development sufficient potential to create new value
(new products/service offerings aimed at by launching new product/service
existing customers) over the other options. offerings and revenue models without
 moving outside their own industries. were focused on continuing to provide data
They may have concluded that claiming to their car owner customers via the car
a new position in an existing value chain dashboard display, but had no clear plans
(as Company E was planning to do by to launch separate data product/service
moving from a service business to a offerings or revenue models to groups
technology business) allowed the same beyond their core customers.
or better opportunities as entering
completely new sectors. In other words, As a general rule, data and AI
they did not see any reason or potential commercialisation within our sample of
for expanding their operations outside companies always took place within a B2B
their current industry/value chain. setting, and as a transaction between two
• Maybe the reluctance to embrace firms. Why is this?
new industry sectors was due to • A simple explanation is that consumers
our sample selection? After all, we are not used to buying data or
talked to European firms that were algorithms, but instead prefer them
not particularly high-tech focused. to be “bundled” with more familiar
They were often incumbents rather services (e.g. served as a by-product
than disruptive niche players. This of cars, home appliances or websites).
explanation makes sense when we Alternatively, consumers may expect
consider that small high-tech companies data and AI services to be free of
from Silicon Valley and China have charge. One of the energy companies in
clearly demonstrated a great appetite the sessions certainly considered this
for disruption compared with their to be so: it wanted to bundle its data
European counterparts in recent and AI assets into other services before
decades. offering them to the consumer market.
• From a legal point of view,
In addition, only a few of the companies companies selling data may find it
we sampled clearly expressed an interest easier to do business with other
in selling data and AI product/service companies. Contracting may be more
offerings direct to consumers. An exception straightforward. Both parties are
was the digital platforms provider startup probably more aware of the legal terms
that operated fully online and already in the transaction, and there are likely
offered data that its consumer customers to be fewer risks related to litigation or
paid for. It wanted to explore new similar bad publicity if something goes wrong.
services. Some companies, such as the • Some companies in our sample only
healthcare service provider Company E and dealt with corporate customers. So for
one of the energy companies, saw potential them, providing data and AI assets to
in providing their consumer customers with consumers was far-fetched.
data and AI, and had already considered
various options, but were not yet sure how When it comes to companies’ data and AI
to do it. Automotive Companies F and G maturity, there were big differences even
in a small sample such as ours. The case difficult to proceed with commercialisation
study of construction services Company plans. Low-maturity companies, or those
H in Appendix 6 is revealing. Because the based in low-maturity sectors often needed
company, its partners and the industry as to start by building various enablers. Data
a whole is not yet proficient in using data infrastructure, capabilities and tooling had
or quantitative information at an advanced to be put in place before they could reap
level, it was hard to start building data and value from data and AI-enabled business
AI product/service offerings or revenue models. As the case study comments that
models. Current industry constraints were follow show, some decision-makers still
an obstacle, and we did not manage to struggled to ensure that their data and AI
explore new business models that would maturity was sufficiently advanced to allow
have allowed Company H to enter new new business models to be created:
industry sectors, let alone start serving
consumers with data and AI. It was a similar
story with the cleantech company in our that the data and AI-enabled solution really works
sample. It lacked many critical elements, and delivers the recommendations that we want to
including IT systems, data infrastructure
and data-related initiatives, that made it

STAGE 1: STAGE 2: STAGE 3: STAGE 4: STAGE 5:

BEGINNER EXPLORER INTERMEDIATE ADVANCED CHAMPION

Strategy &
positioning

Commercial
models
& business
acumen All companies were
All companies somewhere within these None of the
Processes & had passed this three stages, depending companies were
ways of stage already partly on who we spoke to truly at this
working from the company and what stage yet
dimension of this model was
evaluated
Data & tech

Organisation

Figure 27: The data and AI maturity status of companies involved in our sample
very advanced, so the key question is: can we create work, let alone imagine how it could be
turned into new business models and
growth opportunities. Simply bringing
data dashboards to the office and teaching
infrastructure so that it enables new business people to demand up-to-date data before
they make decisions, is an easy step that
could potentially spark innovation in terms
of creating new business models.

Figure 27 shows our interpretation of the

In digital-by-definition companies, by companies’ data and AI maturity status (as
contrast, the situation is different. As discussed in Chapter 4.2).
the business has been built around data
and algorithms, it is very easy to start In many of the co-creation sessions, we also
commercialising them, for example by talked about the maturity of the customer
finding new customers, creating new purchasing data and AI-enabled product/
product/service offerings and building new service offerings. In the case of digital
revenue models. Basically, nothing stops media Company B, the customers wanted to
these companies from making data and AI a buy raw data and had data science teams
significant business opportunity. in place that could enrich and analyse the
data after it was purchased. In other co-
In our sample, there were numerous creation sessions, it was evident that the
examples of firms that represented customers were nowhere near this level of
the middle ground. That is, they were maturity. This meant there was plenty of
not digital-by-definition but they were work to do for the businesses selling data.
not laggards either. What this shows is They had to educate their customers on the
that data and AI maturity unfolds over value of data, then develop product/service
a spectrum of stages that companies offerings that were simple enough for
progress through as they gradually improve customers to grasp, and finally implement
their capabilities. revenue models that were very easy to
adopt. Having said all that, customers’
For many companies, a simple way to low level of data/AI maturity could also be
increase the level of data and AI maturity turned into an opportunity. As one company
in the workplace is by making data and leader put it:
algorithm-induced insights more easily
available to employees. If data literacy in “As our various stakeholders have a low level of
the company is low, people will not be used
to viewing and using data dashboards or
making day-to-day data-enabled decisions.
It is hard for them to understand how data
and AI could help them in their current
It is important to note that in our sample revolutionary new data and AI business
there was very little correlation between models. They are doing their best to make
a company’s data and AI ambitions, its most out of the assets they have. But
strategic approach and its level of data and companies should resist getting too stuck
AI maturity. The case study of construction in their industry-focused thinking or ways
services Company H in Appendix 6 is a good of working. As one business leader said:
example of this. Participants from this
company were very ambitious about their
aims to revolutionise the industry using
data and AI, and their related initiatives control mechanisms for decisions. Because of that,
were highly strategic. Yet the starting
point and the maturity in the industry was We often fail to get things done and miss out on data
comparatively low. On the other hand,
telecoms provider Company D and digital
media Company B were good examples of Going forward, it might be beneficial to
the opposite scenario. These companies invite companies from various sectors to
had achieved a high level of data and AI meet and discuss joint opportunities at
maturity already, yet a more ambitious the edges of their respective industries.
and strategic approach could have helped Tapping into this kind of cross-sector
them to extract additional value from their collaboration might be eye-opening.
data and AI assets. Finally, healthcare
services provider Company E had all three When it came to their use of data and
elements in place: high ambitions coupled AI-enabled business models, the 20
with a strategic approach and good data companies we spoke to were generally not
and AI maturity. Various combinations overly concerned with needing to create
therefore seem possible as companies something game-changing or unique (see
explore the potential of data and AI growth Chapter 4.4). The large companies, in
opportunities. particular typically ignored the competition,
perhaps intuitively considering themselves
All the companies we talked to operate dominant and therefore unique already.
within a specific industry and the business
characteristics of that industry really However among our 20 companies, there
matter (see Chapter 4.3). The practices, were some exceptions where decision-
conventions and customer relations within makers felt that simply creating new data
each industry have a tangible impact on and AI-enabled product/service offerings
what data is available, how it is processed, was not enough. Instead, these companies
and what is generally considered possible wanted to create something of their own
using data and AI. It is difficult to be that would significantly differ from other
too judgemental of leaders that do not potential solutions in the marketplace.
demonstrate the ambition, imagination Two of these companies (digital platform
or courage needed to push forward provider Company C and the consulting
firm) were small and keen to find their own
niche market.

Among the bigger companies, those that

expressed a desire to create something
unique included the large healthcare service
provider and the large energy company,
which were interested in the opportunities
around finding unique problems to be
solved.
 Which

Co-creation sessions revealed that the contexts, not just as a business model
playbook we created ( Chapter 5) seems option (for more about platforms, see
to work very well in a company setting. Moazed91). “What do you mean by selling data
Participants found many of our data and was a question we were asked
AI product/service offerings and revenue a few times. We had to explain that in
models easy to understand. our context, it means digital platforms
that have been created for the purpose
The selling data platforms model of providing customers with data and AI
(Chapter 5.3) sometimes needed more assets.
elaboration. This might be because a lot
of “platformisation” is already taking It’s fair to say that a significant part of the
place within companies. Platforms are product/service offering models discussed
often considered to be relevant to various in the co-creation sessions were, in essence,
company functions, operations and different ways to sell data; whether as raw
data, reports, via data platforms (Chapters portal. And healthcare service provider
5.2 and 5.3) or bundled with consulting Company E was considering spinning off an
services (Chapter 5.5). Many companies in international technology licensing business
our sample had implemented these models using an algorithmic service. While
already, and others were considering them. equipment manufacturing Company A was
Various data platforms, for instance, were planning to sell its AI assets to existing
in place or being developed already. Often, customers, healthcare service provider
the product and services selection was Company E was interested in approaching
made according to the format in which completely new markets. In both cases,
the customers would potentially wish to algorithms were not sold in isolation, but
receive data (with or without AI involved). instead bundled with other services such
When providing products or services to as data, analytics tools, customer portals
customers, many companies seemed to and related technologies.
prefer easily scalable models that required
a minimum amount of manual work. Yet, But why were only two companies
as one participant put it, companies were interested in selling AI assets? Our
often still not sure how to use their data in interpretation is that the AI assets that
a way that provided the most value to their companies have created for themselves
customers: are typically bespoke, and tailored to a
very specific application and context.
“Our raw data seems uninteresting and looks ugly They are deeply embedded in the internal
IT infrastructure, which makes them
difficult to export elsewhere without
massive adaptation. Bespoke AI assets
are therefore difficult to commercialise.
Some companies were also unwilling to
commercialise their algorithms. They
considered them to be part of their
Algorithms and AI assets were commonly core competence, and so constituted
considered as tools to refine the data the company’s IP. So, in our sample
and increase added value before sales. of companies, there was rarely any
Few companies that we worked with willingness or ability to sell AI assets to
were interested in selling AI assets, as other parties. It might be that this model
described in Chapter 5.4. However, two is only suitable for specialised companies
companies were planning on following this that build their business – in one way or
path. Equipment manufacturing Company the other – around algorithm sales. These
A was considering commercialising an might include companies such as Peltarion
algorithm that helped its customers to and Aito.ai, and AI and cloud service
eliminate human errors and variance in providers such as Microsoft Azure, which
manufacturing work, and bundling this provide algorithms as part of their cloud or
service with its existing digital customer platform services.
In many companies, we identified a tension when a company gains capabilities in data
between protecting proprietary data and AI and AI, it wants to leverage them across its
assets for their own competitive advantage, core business and also use them to enable
versus sharing them with potential new business – because there might be
customers and partners to generate more synergies and economies of scale in doing
revenue. Some business leaders felt that both.
selling data or algorithms could leave them
competitively exposed, by revealing the The service-oriented models selling data
strengths or weaknesses of their business and AI-based services and selling outcomes
model to other firms. At the same time, or capacity were also relevant for many
however, they expressed an interest in firms in our sample as they considered
putting their valuable data and AI assets opportunities for selling something “as-
up for sale. Balancing these two opposing a-service”. One of the manufacturing
forces is a key managerial challenge. companies and one of the machinery
Of course, data and algorithms can be companies were thinking of various options
reusable and commercialising them doesn’t as they sought to become “total solutions
necessarily hinder the selling company’s providers” instead of “equipment sales
ability to use them. But the competitive companies”. Telecoms operator Company D
advantage inherent in them may be lost.35/86 was also considering a consulting business
that would be built in addition to its existing
The selling existing product/service data and algorithms. Also, the two energy
offerings model (Chapter 5.1) was evident companies were considering selling energy
in small digital platform provider startup consumption monitoring equipment and
Company C, which had already applied data analysis as bundled service concepts
strong data and AI capabilities in its core to their customers.
online services. Digital media Company B
harnessed data extensively in its classified So, as predicted in Chapter 4.3 when
ads business, while manufacturing and talking about the business characteristics
machinery firms also coupled their physical of different industries, companies were
products with data components. Companies interested in using data and AI to make core
such as telecoms service provider Company transformations in their product/service
D, healthcare service provider Company offerings (in these cases, from products
E and automotive Companies F and G to services) to be able to keep pace with
used their data and AI assets to enhance their customers’ changing requirements
competitiveness in their core products. It (it is reasonable to assume that the recent
is likely that many, if not all, companies are Covid-induced challenges in many firms will
already striving towards complementing have increased the urgency in rethinking
their core services with new data and AI their business models in a similar way).
elements. However, that doesn’t rule out the Yet, while these models were interesting
option of using these assets as standalone to many companies in our sample, they still
new product/service offerings too. Often, tended to be regarded as ambitious plans
rather than something that had been fully Our study does not include quantitative
harnessed. There were also doubts over data collection; so our suggestions are
whether transformative service-oriented merely an educated guess. However, based
models could really work in practice. As one on our discussions with decision-makers
participant put it: in the co-creation sessions, there are our
findings (the € signs indicate the size of the
opportunity):

to become those weird, visionary things discussed Evaluating different data and AI-enabled
by management. But if we look at harsh reality, we product/service offerings in terms of
their business potential:
• Selling existing product/service
offerings (€€€): A major area
of interest for a large number of
firms. This model offers plenty of
opportunities for delivering added
value and a better user experience
As a comment by another participant to customers. It helps companies
reveals, it may also be difficult to differentiate from the competition, and
get customers to understand the gain market share. Nearly all companies
transformation from products to services: should consider using data and AI to
augment their existing product/service
“Our ambition is to sell outcomes. The solutions we offerings.
• Selling data, analyses and reports (€):
A healthy additional source of revenue
for many companies. It complements
As our solutions eliminate manual work, they other businesses but it does not always
integrate well with them. Instead it
represents a standalone opportunity for
many firms, with limited revenue and
value creation potential.
• Selling data platforms (€€): Potentially
a huge opportunity in cases where
a company becomes a network
Finally, it is useful to evaluate all these data orchestrator or a critical control point in
and AI-enabled product/service offerings its value chain or ecosystem. At its best,
from the point of view of their business a platform business model can lead to
potential within various firms, as well as a nearly monopolistic market position
their impact on the global economy. Which and huge profits. But at its worst,
of them have the most potential? Which data platforms are given to customers
will create the most new economic value? for free, as part of the total product/
 service offering of the firm, which limits and interesting. It represents a major
its revenue, value creation and the business transformation opportunity.
potential for differentiation. Yet, as discussed previously, some
• Selling AI assets (€): This model is companies view this model as far-
not yet a mainstream opportunity. It fetched and merely a pipe dream.
seems most usable for niche players
that specialise in AI capability provision.
Algorithms are not always easy to re-
use after they have been developed and
the customer needs to have significant With regards to revenue models (Chapter
technical abilities to use them. They 5.8), the discussions were diverse. All
may also need to be bundled with other seven revenue model options were
services before they can be sold. Many interrogated in the co-creation sessions.
companies wish to protect their AI Typically this took the form of an ideation
capabilities because they consider them mode in which they were proposed then
a trade secret, or a significant factor in coupled with certain selected data and
their competitive advantage. AI-enabled product/service offerings.
• Selling data and AI-based services Very seldom were any decisions taken on
(€€): This model presents a significant revenue models in the sessions. Instead,
new opportunity for growth, as data and the companies that we spoke to typically
AI-enabled services can complement noted that several models could be
firms’ existing product/service applied to data and AI assets, and several
offerings. There are many kinds of new could also be run in parallel to test their
service concepts that companies can adoption with different kinds of potential
invent. These service concepts may customers.
help firms to expand their role in the
value chain or marketplace, e.g. from However, one revenue model stood out:
a products provider towards a full the licensing/subscription model. Business
services provider role, in which they leaders saw this as a means to generate
help their customers to use products stable and predictable revenue, in a way
better. that was already familiar to potential
• Selling outcomes or capacity (€€€): customers. So the licensing/subscription
This model has huge potential. If/when model was favoured over the six other
industrial multinational conglomerates models regardless of the product/service
start transforming their business offering in question.
models from goods provision towards
“as a service” models, there is plenty In contrast, the advertising revenue model
of new economic value to be created. did not come up so often in the sessions.
In addition to industrial firms, many As mentioned earlier, data and AI-enabled
other companies in various industry business models were often implemented
sectors might consider this model viable in a B2B setting, where generating large
 “What is being sold”

EXISTING
DATA, DATA AND OUTCOMES
PRODUCT/ DATA
ANALYSES AI ASSETS AI-BASED OR
“How money SERVICE
AND REPORTS
PLATFORMS
SERVICES CAPACITIES
OFFERING
is made”

FREE,
FREEMIUM

PRODUCT/
TRANSACTION
REVENUE

PROJECT
REVENUE

ADVERTISING
REVENUE

LICENSING,
SUBSCRIPTION

MARKETPLACE,
PLATFORM,
BROKERAGE,
PERFORMANCE

BUSINESS
MODEL
OVERTURN

Almost always discussed Regularly discussed Sometimes discussed Rarely discussed

Figure 28: A heatmap of data and AI-enabled product/service offerings and revenue
models in our company sample
consumer audiences needed for advertising In conclusion, the co-creation sessions
would be a challenge. revealed that companies are very
innovative, even playful, when it comes
The most radical of our revenue models, to imagining new data and AI-enabled
the business model overturn, involved product/service offerings (the columns in
giving the traditional product or service the playbook). They more or less considered
for free in order to collect valuable data everything under the sun and came up with
for commercialisation purposes. This various options. By contrast, the discussion
model came up in our discussion with around revenue models (the rows in the
digital platform provider Company C, which playbook) was commonly very conservative
was trying to identify its most feasible and people naturally gravitated towards
product-market fit and was therefore open the most familiar revenue model types,
to disruptive models. Healthcare service such as licensing fees, product revenues or
provider Company E also considered the project revenues. Maybe business leaders
business model overturn as an alternative, were hesitant to “rock the boat” by creating
should it start serving consumer customers confusion among their customers. This may
more extensively (currently the firm’s clients have meant that they were only willing to
are almost all corporate). In the rest of the favour a smooth transformation to new
companies that we talked to, the existing revenue models in the context of a wider
core business seemed solid enough that transformation. This was an anticipated
there was no desire to start disrupting or outcome of the co-creation sessions, as
cannibalising it by offering a business model suggested in Chapter 5.9.
overturn to the market (remember the
pre-Covid timing of most of our co-creation After conducting the co-creation sessions
sessions). In these cases, the companies with our chosen companies, we find it
could not consider giving their traditional difficult to judge which models in the
core service/product offering free of charge playbook (either product/service offerings
as a viable option. Overall, our observation or revenue models) would work especially
was that business model overturn was well either in: particular industry sectors, to
generally considered by companies as meet a particular level of ambition within a
something that was done in relation to firm or, to match to a company or sector’s
consumer data. They perceived it in terms level of data and AI maturity.
of a product/service offering given for free • It seems that the playbook for data and
to consumers in order to collect data that is AI-enabled product/service offerings
later commercialised in a B2B setting. and revenue models can apply to all
companies.
Figure 28 provides a heatmap of the • Companies from certain industry
product/service offering and revenue sectors might favour some models over
models that companies typically discussed others, due to their internal culture and
in the co-creation sessions, along with the processes. But this should not rule out
options that did not crop up as often. considering all options in the playbook.
 The business model choice in any given were often quite vague, appearing as early
firm – independent of its industry ideas, preliminary discussions or pilot
segment, size, ambition level or data projects with other potentially relevant
and AI maturity – is closely related to companies. We did not witness any strong
company-specific factors such as: evidence of data and AI-related contracting,
› Market conditions ( in which several players had come together
). to share their assets to create a radically
› Existing customer relations (“What new type of value in the marketplace. And
). the first and foremost thought for most
› Relations to third parties (“Who could firms in our sample was how to expand
). their own business. This is not entirely
› The strategic agenda and ambition surprising. But at the same time, many
in place (“Where should we be companies found it hard to envisage all the
). rich growth avenues that may have opened
› Maturity ( up as a result of working with other firms in
). an ecosystem or consortium.
› Imagination of the key people
involved (“Could we invent a business
model that clearly stands out from the
).
• In this respect, there will be both
leaders and laggards in all industries:
some companies will be more mature,
forward-looking, ambitious and
innovative, and others less so.
• By using the playbook created in this
book, each company needs to find its
own way when it comes to data and AI-
enabled business model implementation.

In the co-creation sessions, many

companies considered an ecosystem play,
as well as data and AI-related partnerships
or consortiums with other firms, as a viable
model for implementing new business
models.
was a common outlook, because
firms saw other companies as a way to
provide important assets for them, or as
a means to help them commercialise their
own assets. Yet, plans for partnerships
 their data and AI assets. In the course of our
We call these prioritised models “killer research, we
concepts”, i.e. data and AI-enabled business
models that seem to work very well in identified four killer concepts. They are
several companies already. While the name described in more detail here:
killer concept is slightly provocative, it • Sell customers their own data
captures how widely these models can How it works: Collect data on your
be applied, as well as their ingenious customers → Enrich and analyse it
and transformative potential for helping to create insights, observations
companies to create new value. In applying and recommendations → Turn the
these killer concepts, there is often a recommendations into a sellable
change in both “what is being sold” (the product → Apply a monthly licence fee
product/service offering) and “how money for access to the recommendations.
is made” (the revenue models) when This is a killer concept that builds on
comparing them against traditional models the earlier selling data, analyses and
that companies used before harnessing reports model in the playbook, as well
 as the licensing/subscription fee model, new platform business in a B2B context
to generate recurring revenue. There that brought various companies
were several companies in our sample together to trade goods and services.
using this killer concept, ranging from The foundation of this killer concept
equipment manufacturers all the way is in the selling data platforms model,
to healthcare and media companies which is commercialised through a
(see also the examples of Fitbit and commission (paid after each successful
Sievo in Chapter 5.2). The customers transaction in the platform). Obviously,
of this killer concept appreciate the the role of the matching algorithm is
added value that the company selling is critical in making sure that customers/
able to generate by consolidating and users can easily find the assets to
analysing their data and then providing be traded — or shared in the case of
recommendations, comparisons, Airbnb. This will enable transactions to
reports and dashboards that help them happen as efficiently and conveniently
identify opportunities for improvement, as possible.
e.g. when benchmarked against other • Build data and AI-optimised service
similar businesses or individuals. Some concepts that replace ownership with
customers were aware that data from access
their own activities was being sold back How it works: Collect data on the use
to them, but they rarely complained of your product → Identify ways to
about it as the added value of the data provide the same value without selling
service was so high. products anymore → Provide access to
• Build a data and AI-optimised platform the products and optimise provision
to match supply and demand with algorithms → Invoice customers
How it works: Identify unused assets → based on outcomes or capacity. This
Find ways to match supply and demand killer concept builds on the selling
for them → Create an algorithm that data and AI-based service and selling
does the matching work automatically outcomes or capacity models in the
→ Build a marketplace that delivers playbook. The essence of it is to replace
assets on demand → Apply a commission product sales with solutions sales.
fee. This is the classic Uber and Airbnb The customers are typically buying
killer concept in which there is a data a sense of comfort and convenience
and AI-optimised platform that brings that things work as planned. They do
various parties together and enables not have to worry about owning or
transactions between them.39/91/92 Large maintaining their products anymore.
global digital-by definition firms don’t There are multiple contexts in which
have a monopoly on this killer concept: this killer concept can be applied. It is
other companies can apply it in their common in the industrial sector, where
respective sectors too. For instance, companies are working towards a pull
one industrial company in our sample economy (as explained in Chapter 4.2).
was considering setting up a radical Consumer car-sharing services are also
 good examples of data and AI-enabled consulting services, partner services and
businesses in which ownership is hardware come together. So really we are
replaced with access. seeing “total concepts” in which data and
• Turn your core competence into an AI play important roles as enablers, but
algorithm also need to be complemented by other
How it works: Understand what you features. This allows the company to build
know/do the best in the world → Think solutions and value propositions that
how to build rules for the way that customers find meaningful and desirable.
knowledge/activity works → Create
an algorithm that does the job → Turn Some companies might want to try creating
the algorithm into a sellable product these total concepts without developing
→ Apply a monthly licence fee. This sophisticated algorithms, thereby
killer concept was widely considered implementing “only” other components.
by companies in our sample including That is totally fine. But the more algorithms
healthcare service provider Company there are in these concepts, the more
E, equipment manufacturing Company usable, optimised, automated and valuable
A, an energy services provider firm and they become.
a small consulting firm. The essence
of this killer concept is to understand It is possible to argue that merely building
deeply the core competence of the personalisation into a firm’s existing
firm (as discussed in Chapter 4.4), then portfolio of products constitutes a killer
use data and algorithms to map and concept in its own right. Many companies –
describe it. This helps other companies from websites to mobile apps, advertising
to do the same by licensing the assets. companies and service providers – are
This model resembles selling AI assets currently doing this. Data and AI-enabled
and selling data platforms in our recommendations mean that they strive
playbook, depending on how the final to create yet more personalised user
product is delivered to customers. While experiences, which drives up service
the monthly licence fee is not the only purchases and customer engagement
revenue model applicable in this killer overall. Even though personalisation is
concept, often companies consider it valuable and transformative, this concept
to be a business idea that resembles seems rather basic, especially from the
technology licensing, and therefore they business model point of view. Therefore it is
prefer recurring and stable licensing not included in our list of killer concepts.
fees in this context.
All four killer concepts outlined allow
It’s important to note that data or AI companies to start building a scalable
are not the only elements in developing “second leg”. We highly recommend testing
these killer concepts. Instead, companies them to see whether they could be used
typically create complete packages in which more widely in your business.
data, algorithms, technology platforms,
These journeys may entail a move from product/service offerings and revenue
product-related activities to providing model ideas, e.g. by using our canvases.
services, from selling products and services • Certain companies were already
to platform business models, or switching exploring data and AI-enabled product/
from one-off deals to scalable, stable services and revenue models. These
and recurring revenues.7 In some cases, firms often had plenty of ideas about
the journeys could even involve a number how to turn data and AI into new
of these transformations at the same product/service offerings and revenue
time, bringing about a change both in the models, and how to solve important
product/service offerings and the revenue problems in their marketplace. But
models. These journeys might not always be it was not yet clear which solutions
entirely due to data and AI assets, but they would be most financially or technically
often play a significant role in making the viable or desirable in terms of solving
journey possible. real problems. So, the next step was to
start prioritising the ideas that were
In general, the journey towards data and AI- already in place (e.g. by using the Impact
enabled new business models, as captured & Implementation Optimiser canvas);
by our co-creation sessions included several and testing and validating them in
steps: collaboration with potential customers.
• There were some beginner companies • Several companies were clearly in
that did not have enough data and acceleration mode. They had several
AI-related business ideas yet. It was pilot projects, proof-of-concepts or
not clear to them how data and AI could MVPs on data and AI-enabled product/
enable business growth, and the topic of service offerings and revenue models.
this book was new to them. So there was Sometimes they already had their
a need to inspire the business leaders first customers and tangible revenue
to think about the opportunities, and generation in place, too. It was
help them to generate more data and AI important for these companies to start
 scaling up the business, e.g. by creating catering for various abstraction levels (e.g.
a coherent narrative for their activities, culture transformation versus narratives
searching for more customers and versus details of projects) at the same time,
dedicating more resources to initiatives. while pushing overall towards new data and
• Finally, some companies were ready to AI-enabled business models (again, this is
drive a full culture transformation in typical of a digital transformation journey,
their organisations. Key people in these too).
companies were already fully onboard
with the creation of data and AI-enabled Often, the further a company proceeds on
new businesses. But the next challenge its journey towards data and AI-enabled
was to engage a larger group of people business models (as described earlier
in the organisation to work on data and in this chapter), the more inevitable and
AI commercialisation, and eventually urgent corporate culture matters start to
deliver a more thorough transformation be for business leaders. This, in turn, brings
that reached across the entire firm. a company ever closer to implementing a
complete culture change.
The journey seen here is similar to the
digital transformation journey that many Figure 29 represents three management
companies are already pursuing, in which paradigms that are important to
ideation leads to first projects that are later understand with regards to culture
scaled up, until eventually the organisation change, and that may be brought about by
as a whole tranforms. Often the data and implementing data and AI-enabled business
AI-enabled journey is not entirely separate models (source: Futurice materials, see also
from the company’s digital transformation Fountaine, McCarthy & Tamim53)
journey (if it exists). One journey can be • Starting from the left, the classic
parallel and complementary to the other, or industrial management paradigm
completely integrated into it. cherishes top-down management,
large functional silos, detailed plans
This journey is not necessarily one-way, and meeting strict targets. Here,
either. Some companies might want to the leadership is mainly focused on
start with culture transformation (e.g. by improving the core business but it is
building common narratives and engaging not too interested in fostering and
people more widely). Then once traction encouraging experimentation and
has been created across the firm, they risk-taking.7 In order to succeed with
move onto innovating product/service business model innovations (related to
offerings and setting up pilot projects. This data and AI or otherwise), companies
approach perhaps resembles a “think big, typically need to move away from this
start small” strategy that is guided by a model as much as possible. It seems
strong vision that binds activities together. counterproductive and outdated,
Obviously in many companies there is and will often kill all innovation,
also an opportunity to go back and forth,
 experimentation and entrepreneurial The different management paradigms are
spirit. not necessarily black and white: shades of
• The column in the middle presents grey are possible if companies choose to
a modern innovation-focused combine elements from several of them,
management paradigm that cherishes and create their own unique approach (if
experimentation, MVPs, failing fast that is the case, management needs to be
and autonomous and entrepreneurial extra careful in bringing clarity and not
teams. This approach is crucial in confusion to the integration effort).
unleashing business model innovation
and generating ideas when it comes to Also, a company culture, especially in a
data and AI-enabled business models. In large corporation, can involve several
this paradigm, the leaders understand management paradigms in parallel. Due
how innovation works and they invest to the nature of their work, different
substantial time and energy into departments or operations might be
supporting it.7 Without this element managed in very different ways. Since
in the company culture, it becomes conflicts between different sections might
difficult to proactively spot new data be difficult to avoid, larger organisations
and AI-related opportunities, or react should strive towards one management
quickly to changes in the marketplace; paradigm; or at least be mindful and
for example if competitors launch new accepting of the way that different sections
data and AI-enabled businesses. operate.
• The column on the right-hand side
presents a data and AI-focused When thinking about new data and AI-
management paradigm. This strives enabled business models, our co-creation
towards finding relevant data and sessions revealed that it is not always
generating data-driven facts through enough for a small group of people within
observation, prediction and automation the firm to get excited about them;
activities, as well as modelling especially if their activities are restricted to
company operations, to find an optimal a laboratory or a dedicated business unit.66
configuration. This management Instead:
paradigm, coupled with a modern • It might be necessary to develop the
innovation-focused one, is required by company culture and overarching
all companies wishing to build new data organisation in such a way that it
and AI-enabled business models. It can’t understands and broadly applies data
always be applied to all operations in and AI in its various daily operations.
all firms. But as a rule of thumb, the • We noted that the sales organisations
more a company learns to use data of our sample companies were often a
and AI widely in its operations, the bottleneck for new data and AI-enabled
higher its data and AI literacy, and the business model implementation, as they
more prepared it is to implement those need to embrace the completely new
technologies in its business models. product/service offerings, revenue
 MODERN INNOVATION-
CLASSIC INDUSTRIAL DATA AND AI-FOCUSED
FOCUSED

Guiding Incremental change,

Create new Automate
principles operate

Handling
Create plans Experiment Find data
uncertainty

Product
Hero’s vision MVP + validation Observation
planning

Guidance Detailed plan Vision Probabilities

Small teams and Data and AI-enabled

Team work Large functional silos
collaboration coordination

HIPPO – Highest paid, AI modelled and data-

Decision-making Autonomous bottom-up
directive top-down driven

Metrics Revenue, profit, RoE Innovation accounting Realisation of probabilities

Incentives Corporate Entrepreneurial Continuously learning

Budgeting Project Venture Predictive

Clear industry
Market
definitions, logical Creating new markets Uncovered from data
definition
segments

Invest into Assets People Data source

Learn, unlearn,
Mantra Meet your targets! Fail fast!
re-learn!

Embrace Capital and hierarchy Knowledge and opinions Data-driven facts

Figure 29: Three management paradigms

models, buyer groups, sales models and product management, marketing and
sales tactics required to succeed. HR – also needed to be brought onboard
• The sales organisations were not the too, in order to achieve a wider impact
only area of our sample companies to with data and AI-enabled business
create bottlenecks. Other departments models.
and functions – for instance R&D,
• Various training programmes are for results to materialise may prove
probably necessary to raise overall data overwhelming for many, hindering
and AI literacy in firms, and these would progress in culture change.
be particularly beneficial if their content • Companies that have relied on detailed
is tailored to the needs of people across plans, strict targets, corporate rules
different departments, and with varying and incremental change (as in the
skills, career levels and roles.48 classic industrial management paradigm
• As discussed in Chapter 2.3, organising in Figure 29) will often find it hard to
people with different functional thrive in the data and AI age, where
backgrounds into small, focused business is being driven by observation,
teams that work on relevant business iteration, prediction and continuous
problems, and design data and AI- learning based on new data; a model
related solutions for them, would help that is very new to them.
break silos and cultivate the sharing of • The change management effort to move
various learnings across the company. from one paradigm to another might be
a large and long-lasting one, involving
The move towards data and AI-enabled the breakdown of change barriers such
business models is often a journey and a as habits, resistance and organisational
culture change that is about creating a alignment.
different kind of company altogether. In
order to obtain results with data and AI, it Having said all this, a crisis such as Covid-19
is common that business leaders need to might just be the catalyst that decision-
re-architect the way that their firm works. makers and their organisations need. As
This includes changing how it gathers and management guru Peter Drucker said
uses data, reacts to information, makes back in 1985:
operating decisions and executes operating This
tasks.66 truism still applies now. Like all potential
crises, Covid-19 and the subsequent
Business might encounter major obstacles economic downturn can provide a window
to a change in culture and the creation of of opportunity for companies to change
a new company that harnesses data and their culture for good, especially if the
AI business models. These obstacles often post-crisis period requires new types of
come from within the organisation and leadership, principles and ways of working.
include the fact that:8/27/32 Over the coming years, we will see what
• Not all companies fully grasp the extent future-proof cultures companies have
of the transformation needed when developed, whether prompted by data and
applying data and AI to business. AI initiatives, emerging crises or other
• Even though commercial opportunities factors; and also the time required to make
related to data and AI may look these changes happen.
lucrative, the time and effort needed
8 Conclusions and
recommendations
In this chapter, we put the pieces of the puzzle together
and present our recommendations for data and AI-enabled
business models. We also explore plausible scenarios for
future development.
8.1 Plausible ways forward
This book has been about exploring and Figure 31 summarises the company
co-creating data and AI-enabled business examples that have been presented
models. At the beginning of the book, we in various chapters of this book (note:
set out to address the following topics: these are not to be confused with the 20
how data and AI can transform the way anonymous companies we used in our
companies create, deliver and capture sample study). As we can see, when it
value, through changing what they sell comes to the routes that companies are
(product/service offerings), and how taking to make their data and AI-enabled
they make money (revenue models). business models a success, there is no
Through the playbook (Chapter 5) and “one-size-fits-all” approach but a series of
the co-creation methodology and case equally viable alternatives. This depends on
studies (Chapter 6 and Appendix 6), we factors such as how to create value, which
have built an overview of the various product/service offerings to build, and
opportunities and challenges for companies which revenue models to apply to capture
in this area (reviewed in Chapter 7). Figure maximum revenue. Exploring opportunities
30 summarises the opportunities and iteratively together with customers and
challenges for businesses tackling data and partners is a good way to see which of the
AI-enabled growth. avenues would be most feasible, technically

OPPORTUNITIES CHALLENGES

• New stable, scalable and recurring revenue • Significant development of staff competences
streams in data management, analytics, and selling the
• A disruptive opportunity to escape from the new product/service offerings
limitations of the legacy business • Educating customers to buy new things, buy
• An extension to the digital strategy already them differently or pay differently
in place • Need to balance the legacy business with the
• Ability to deliver significant new value to new data and AI initiatives
customers and/or to win new customers • Continuous change management and
• Ability to position the company as a data and tolerance of uncertainty and risk
AI-harnessing entity, for a positive impact on • Company-internal transformation to scale up
company brand and valuation the business once there is evidence that the
new concepts work

Figure 30: Opportunities and challenges related to data and AI business models (source: Mika Ruokonen)
COMPANY
INDUSTRY SEGMENT IMPACT OF DATA AND AI
NAME

Data and AI as a core product/service offering of the firm: ag-

CB Insights Information services gregation and analysis of data by using algorithms, provision of
analyses and reports

Aeroplane Opening access for customers to equipment data so that they can
Boeing
manufacturing discover new insights or improve their operational performance

IBM and Logistics and A partnership of two major companies that have jointly created a
data platform for a certain industry sector (cargo and shipping) to
Maersk shipping harness

Increasing company valuation by using data and AI as assets in the

WhatsApp Software
balance sheet (Facebook acquired the firm)

Increasing company valuation by using data and AI as assets in the

Minecraft Games
balance sheet (Miccrosoft acquired the firm)

Personal Data and AI changing the behaviour of consumers (using taxis,

Uber owning a car), so providing value to customers (e.g. cheaper,
transportation smarter)

Media and Radical business model renewal (from DVD rental to content
Netflix streaming and to content creation), data and AI supporting the
entertainment change

High data and AI ambition both within the industry segment (weld-
Kemppi Welding ing) and potentially also in a new industry segment (data services
provision)

Entry into a new industry segment (from automotive to

Tesla Automotive
insurance provision), thanks to data and AI

Entry into a new industry segment (from telecom to banking

Orange Telecom
services provision), thanks to data and AI

For customers with low data and AI maturity, the product has been
Facebook Digital media made convenient to use, e.g. data and AI elements embedded and
hidden

Industry segment in question impacting what data is available in

Coca-Cola Beverage the firm, hence what data and AI business models could be
implemented

Industry segment in question impacting what data is available in

McKinsey Consulting services the firm, hence what data and AI business models could be imple-
mented

Keeping the core offering the same (cars) but augmenting it with
Tesla Automotive
data and AI assets (e.g. autonomous driving and navigation)

Sales of anonymised and raw data about customers (location) to

Vodafone Telecoms another firm (TomTom) who uses it to create new services
(navigation)

Sales of data that happens in the company’s (Walmart) value chain

Walmart Retail
(to various brands such as Procter & Gamble)

Sales of data (mapping information of homes) to a player in another

iRobot Home appliances industry segment (tech companies that develop smart phone
products)
 COMPANY
INDUSTRY SEGMENT IMPACT OF DATA AND AI
NAME

Sales of analyses (sleep quality and duration) to consumers, pro-

Fitbit Wearable devices vision of personalised services (health and fit guidance) based on
data

Pharmaceuticals Sales of analyses on supply and demand of products (drugs) up and

Tamro
distribution down in the value chain (to pharmacies and to drug manufacturers)

Using the data on customers to generate and sell reports on their

Sievo Software
own activities (procurement spending)

Data and AI-optimised platform on which to do business (buy tar-

Facebook Digital media
geted online advertising space)

Provision of a platform (App Store) on which to tap into data and AI

Apple Digital devices
assets (e.g. speech recognition and face recognition applications)

Consolidating various alternative dataset sources to one platform

Eagle Alpha Software
(data brokerage between data providers and investors)

Applying subscription revenue model to an online platform (access

LinkedIn Digital media
to advanced data features for consumers and companies)

Amazon Web Provision of off-the-shelf AI assets (e.g. voice analysis and transla-
Software
Services tion module) as part of a cloud offering, enabling new innovations

Provision of algorithmic technology (heartbeat analysis) to other

Firstbeat Wellness and health
players (sports watch firms), so enabling new innovations

Provision of consulting services that have been built on top of data

IBM Software
analysis and AI technologies (Watson)

Lassila & Provision of data-enabled consulting services (energy consumption

Circular economy
Tikanoja analyses and training) to customers (real estate owners)

Explosives and Provision of data-enabled consulting services (optimisation of blast-

Orica
blasting ing outcomes) to customers (e.g. mining and quarrying companies)

Equipment Using data and AI to predict performance (of equipment), provision

Kalmar of performance contracts to clients (harbours) and guaranteeing
manufacturing outcomes

Agriculture Collecting data from customers (farmers), then selling outcomes to

John Deere
machinery them (e.g. increase in crop yields)

Power generation
Collecting data from customers (e.g. power producers), then sell-
GE Power and water
ing outcomes to them (e.g. improvements in efficiency)
technologies

Man Truck & Using data to understand activities (vehicles and drivers),
Automotive guaranteeing uptime of products (trucks), selling capacity instead
Bus of products

Predicting customer needs (e.g. location) and product usage (e.g.

ShareNow Car rental
maintenance needs), selling capacity instead of products

Using “business model overturn” as a revenue model for a data and

Google Digital media
AI offering (search functionality free-of-charge)

Figure 31: Company examples that have been referenced in this book
viable and desirable in terms of solving real data and AI into their product/service
problems.106 offerings and revenue models.30

As we can see from Figure 31, implementing By implementing data and AI-enabled
data and AI-enabled business models is product/service offerings and revenue
not only reserved for Silicon Valley-based models, companies can also build resilience
technology behemoths such as Google into their business. They can reduce their
and Facebook. There are several examples dependence on legacy businesses and
of companies from traditional industries assets, and open new avenues for growth
in various parts of the world that have that differentiate them from competition.
implemented them successfully. As we see in the case studies in Appendix
6, many companies have already realised
Throughout this book, we have challenged this. The economic difficulties caused by
business leaders to consider the idea that Covid-19 have demonstrated that this kind of
data and AI should not be seen just as assets resilience is necessary now more than ever.
to improve efficiency or capital usage, or to
cut costs in their existing business. At their It is critical to keep in mind that your firm’s
most radical, data and AI instead present existing business models might become
standalone new business opportunities: obsolete once data and AI technologies
they are tools for expanding business and arrive. So it is important for decision-
disrupting industries. While many companies makers to continuously examine and
initially look to data and AI to deliver interrogate the efficacy of their current
incremental cost-cutting and productivity business models. Ask questions such as:
benefits, it is likely that their biggest future “How vulnerable is our business model to data
value will be derived from their potential and AI disruption, and how soon might the change
to deliver transformative growth. Revenue happen? Instead of waiting for it to happen, should
gains can be a particularly powerful catalyst we be driving that change?”
for embedding data and AI deeply and more
broadly into the structure of any given Our belief is that business model renewal,
business.30/49/51/107 driven by data and AI, is challenging,
transformative and complicated. At
There may be several obstacles and the same time, however, it is capable of
constraints to commercialising data and delivering the biggest impact and the best
AI. And sometimes, tackling efficiency new opportunities (see also PwC51).
and optimisation may take priority for
business leaders, especially during the Harnessing data and AI for growth is
current turbulent times that have forced valuable not only for the companies that
many companies to cut costs quickly. Yet embrace it, but also for the people they
we believe that, in the long-run, companies employ to implement growth initiatives,
are missing a significant commercial and for wider society which benefits
opportunity if they do not actively introduce from increased job creation and income
from taxes generated and paid. Unlike “I think we’re in the phase where AI will
instances where data and AI is used to change pretty much every major industry.”
enable improvements in efficiency, where • Bill Gates, founder of Microsoft,
the aim is often to eliminate and automate highlights the value created and sees
tasks; with data and AI-enabled growth, no risks in the coming years:
the opposite often happens because there “In the next 10 to 20 years, AI is going to be
is more work for the employees to do. The extremely helpful in managing our lives… The
ripple effect from data and AI-enabled
transformation therefore has the capacity super-smart is way out in the future.”
to be as positive and revolutionary as the
transformation itself. If these assessments are accurate, the
technical foundation for implementing data
In this book, we have assumed AI to be and AI-enabled business models is likely to
the next wave of digital development in become very solid in future years, allowing
companies. With AI increasingly spreading companies to reimage their businesses
across different departments and levels of in many ways. And almost certainly, the
a company, we are seeing a phenomenon biggest opportunities have not yet been
that builds on top of existing digital exploited.10
transformation. AI also presents a new
level of abstraction whereby, instead of Today, many data scientists teach AI
humans doing all the work (e.g. coding, systems in the same they would train a
calculating, predicting), computers are dog, giving them continuous feedback on
augmenting what humans can do, so performance and rewarding the “right”
providing increased business value. We behaviour (as explained in Appendix 2, this
anticipate that companies will increasingly practice is referred to as reinforcement
strive to take greater advantage of their learning). That is about to change. Soon
data, too: so work on that front will also AI systems will not require that much
continue. guidance from us: instead they will learn
to teach themselves and adopt new skills
Various observers predict that the without our intervention. As predicted by
development of AI technology will Lovelock,108 eventually the roles may flip:
accelerate, with many of them anticipating the AI systems may come to view human
big opportunities rather than threats when employees as being the ones who need
it comes to implementing data and AI: constant guidance. In aeroplanes, this is
• As Gandhi and Ehl11 state optimistically: already happening: pilots are much more
“AI is likely to come faster than you might prone to error than the autopilot system.
think, and the impact on your business might be While humans have so far been responsible
far greater than you can imagine today.” for designing all the AI systems (and many
• Andrew Ng, adjunct professor at experts state that this should always be
Stanford University and former Chief the case), in the future we might witness
Scientist at Baidu says: supercomputers that are able to design
their own new extensions. This would make still a discrepancy between the ambitious
it difficult for humans to work out what blueprint set for AI as a revolutionary
is actually happening, and how to stay in technology (by many visionary authors)
control of computer-authorised activities and the practical plans and activities of
and decisions. We shouldn’t be arrogant European companies planning to use this
and ignore the rapid development of AI technology for business model renewal in
that is continuously taking place. Our the here and now.
jobs and daily lives are likely to depend on
these systems in the future, whether we It is possible, too, that the development
want them to or not. It is simply not easy of AI will be far from a linear success
to speculate what, exactly, this future story. AI has a history of being oversold
will look like. What will the mechanisms and the reality of what was achieved
of accountability and responsibility be as differs dramatically from what had
computers start to learn more complex been promised.109 Furthermore, it is
patterns, and also become more opaque? not guaranteed that the world will be
a significantly better place once AI has
A key question is also how business models become a more advanced and potentially
should be designed for AI systems that ubiquitous presence in our lives.11 We should
are capable of identifying new business have no illusions about this technology but
opportunities and designing related approach it pragmatically and sometimes
product/service offerings and business cautiously, to make sure it really delivers
models for themselves. Initially we can on the promise, and that its effects are
assume this will happen with help from positive.
humans, but later it may take place without
human intervention. It might be that
business opportunity identification and
business model optimisation designed
by data and AI will constitute new fields
of their own. Companies, therefore, will
increasingly ask their computers to work
out how to set up the business.

In the near future, humans will still be

needed to invent and implement data and
AI- enabled business models. As we have
seen in our co-creation sessions, companies
are far from being ready to implement
actual plans in this area. Instead they are
typically on a journey towards creating
more and bigger business opportunities
from their data and AI assets. So, there is
8.2 Legal, ethical and
sustainability considerations
While using data and algorithms as a standalone business
opportunity to generate more revenues makes perfect
commercial sense, it is still a controversial topic that may

troubling, as they are afraid of losing authority over their

personal (or their company’s) data. Or they fear losing
transparency over how the data is used. The fact that
there are more than 100 startups in Europe alone that help
consumers to manage their private data110 illustrates the scale
of this concern.

These fears are justified, too. The aim of

this book has certainly not been to propose used for both good and bad,11 and legislation
a “Wild West” setting, in which businesses must keep up with their development. Even
that trade data and algorithms lose their well-intended use of digital technologies can
control, ethical principles, or respect for the have unexpected negative consequences
law. We agree with Spijker31 who insightfully that need to be recognised rather than
points out that privacy is not only in most glossed over. Reassuringly, our society
cases a legal requirement, it is also an ethical has already developed various ways of
one. In successful ventures, privacy is, as limiting the potential damage caused by the
Spijker suggests, “not about secrecy, but instead irresponsible use of tools like data and AI.109
it is about being very clear and open regarding what
happens to data and who gets to use what”. Intangible digital assets such as software,
data and algorithms are likely to play an
The implementation of new data and AI- increasingly critical role in companies (as
enabled business models will be very much opposed to physical assets like machines or
dependent on the legislation in any given real estate). This is true on a country level,
country. It is important to note that data and too. Therefore policy makers will be tempted
AI are powerful tools that can potentially be to build more control and governance
for data and AI, to stay on top of their and mitigated. Among these, privacy
development and to protect national and consent issues are top of the list,
interests. especially if companies wish to monetise
data assets that are even vaguely related
Legislation typically lags some way to individuals (whether anonymised and
behind technology, however, and this may aggregated or not). But topics such as
especially be the case with AI, given the contracting and liabilities also need to
vast amounts currently being invested be addressed if a company wants to use
in developing AI applications. While its data and AI in launching new business
GDPR regulation has been implemented models, too. Assuming that systems could
in the EU (not yet globally), it is clear be constantly evolving of their own accord,
that governments are only starting to which contractual remedies should follow
realise what opportunities there might be if there are unwanted outcomes?
around data and AI. Therefore, companies
wanting to establish new data and AI- If there is legislation in place that restricts
enabled business models often need to the commercial use of data and AI assets,
apply self-regulation. Doing so will allow there will inevitably be legal obstacles
businesses to anticipate potential changes for business model innovation activities,
in relevant legislation and proactively too. That’s not necessarily a bad thing
mitigate their potential future legal risks. if the rules are clear to all and bind all
Historically, the need to regulate has companies equally. But it is important
often only became apparent after a rapid for governments to understand deeply
increase in the scale of a problem.109 The what implications these laws might
fact that there is currently only a limited potentially have on privacy, innovation,
amount of regulation for AI is a sign that competitiveness and asset management.
its implementation issues have not yet For instance:
been severe enough to trigger immediate • Where does the balance lie between
government action. protecting the privacy and interests
of individuals and companies that
Throughout this book we have assumed produce data, versus the economic
that a company has legal rights to benefits that might be gained through
possess, process and commercialise its implementing new data and AI-enabled
data and algorithms. Obviously, the more business models that deliver growth
a company wishes to use its data and AI and prosperity to the nation as a
assets outside its core business (e.g. to whole?
access new industry segments instead of • When data and AI assets are sold to
optimising its current operations, or to third parties, should the company
launch standalone data and AI product/ selling be held liable for the potential
service offerings instead of improving misuse of these assets, given that such
existing ones), the more it exposes itself to regulation would obviously discourage
new legal risks that need to be understood these kinds of transactions?39
These and other questions will need to be Do companies use data and AI to act
considered very carefully by governments responsibly and create positive outcomes
going forward. for society overall, including various
stakeholders and the environment?47 Do
What’s more, even if something is legal, they strive to help people reduce or avoid
it doesn’t automatically mean it is also polluting unnecessarily, or help them to
ethically sound. So new data and AI- do things smarter or better? Or do they
enabled business models should also be instead aim to maximise the sales of
evaluated from an ethical perspective. Is unnecessary junk at the expense of nature
the purpose of the new models sound, are and society, only driving value for their
they developed with the right intentions shareholders? Some data and AI-enabled
and does implementing them lead to business models might be better than
positive outcomes for all parties? Are they others in this respect. For instance, selling
free from biases and do they treat people outcomes or capacity models commonly
fairly? Is there enough transparency when enable the selling party to take ownership
implementing them, so that people can of system inefficiencies (e.g. energy
trust them and use the new models without consumption or wearing out of products).
concerns? Even though this book does not This gives them a built-in incentive to
put particular emphasis on the ethics of AI, strive towards efficiency and durability.
these are potential issues and pitfalls that The selling data platforms model may also
anyone designing new data and AI-enabled give business leaders an opportunity to
business models needs to take into account. put unused assets to work, and replace
more harmful ways of working. So it is
Finally, in today’s world, there is a up to individual companies to ensure
shared responsibility to look at new that new data and AI-enabled product/
business models from a sustainability service offerings and revenue models are
angle too. A major shift is happening developed with good intentions and do not
as many businesses are moving from impose further on our fragile environmental
a maximising profit approach (profit- ecosystems. Keep in mind that
driven) to maximising stakeholder value sustainability can be an opportunity and
(purpose-driven). Given the challenges we potentially offers a first-mover advantage
are facing in relation to climate change, for those who do it early and well.
loss of biodiversity and overconsumption,
to name just a few, all companies need to As this book is about data and AI-enabled
consider carefully whether their anticipated business models – and not about legal,
new product and revenue models create ethical or sustainability issues per se –
a positive or negative impact. The key we end with an appeal that companies
question is: What is their purpose, from a and their leaders act responsibly. Proper
sustainability point of view? governance involves ensuring that data and
AI assets are used to further the business
goals of the organisation in question while
also complying with local laws and ethical
and sustainability requirements.54 Luckily,
various organisations in different countries
are already discussing these issues, and
progress by both governmental and non-
governmental organisations is ongoing.
8.3 Nothing new under the
sun?
As we approach the end of the book, it’s worth asking
whether data and AI-enabled business models genuinely

the way that companies make money and create product/

and AI assets. Data and AI can therefore

As you might remember from the give companies a major competitive edge
introduction, we started to explore these in their product/service offering and
questions (Chapter 1.2) and organised co- revenue models, helping them to switch
creation sessions with companies (Chapter more easily between selling products versus
6), without fully knowing their answers. services, and capture more value from their
Rather than being seen as a mysterious customer base. It often helps them gain
or complex “black box”, we suggested more customers and access completely new
that all companies should master how to industry segments.
extract business value from data and AI.
And we created a pragmatic playbook for That said, some models presented in our
companies to do exactly that in Chapter 5. playbook are arguably not that revolutionary
after all. For instance, the models selling
On reflection, it is fair to say that data and existing product/service offerings and
AI do indeed change what companies sell selling data, analyses and reports have been
(product/service offerings) and how they around for some time. Likewise, many of
the revenue models, e.g. freemium, project
make money (the revenue models), to a revenue and advertising revenue are long-
certain degree. First of all, there are new established practices. Whether these have
things to sell: the data and AI assets that been used effectively is another issue but as
companies are starting to own. Secondly, models they are not new.
models such as selling outcomes or capacity
and business model overturn could be truly The point is to illustrate the many
revolutionary for many companies and are opportunities that companies have when
typically heavily dependent on, or can be it comes to data and AI, even if some are
significantly boosted by, the company’s not entirely radical. The playbook provides
ability to benefit from their valuable data guidance on what the data and AI product/
service offerings could be, and how to build dependent on the choices that each
new kinds of revenue models, but it is not individual company decides to make.
exhaustive: other options could be possible
as well. It would be interesting to conduct It is also essential that data and AI-enabled
a follow-up study five to 10 years from models aren’t treated as a silver bullet or
now to understand how transformational one-off solution, but as part of a company’s
the new data and AI-enabled product/ broader digitalisation agenda and business
service offerings and revenue models will strategy.51 Depending on the choices
become once creative businesses have made around digitalisation and business
truly embraced the potential of these strategy, data and AI can be used as a tool
technologies. In this sense, the playbook to amplify business performance in many
we present could perhaps be considered as different ways. The full benefits from
a “Minimum Viable Product”69 – something data and AI-enabled business models will
that can be further elaborated upon once only be realised if they are approached
there is more information and evidence strategically.
available.

The true answer to the question of whether

data and AI-enabled business models are
something radically new is dependent on:
• How radical does the company want to
be?
• What works for current or new
customers?
• What stands out from the competition in
any given industry? And what is required
to stay competitive in the marketplace,
going forward?
• What is feasible from the choices made
in the various boxes of the Business
Model Canvas? (Chapter 3.2).

The playbook presented in Chapter 5 can

be used either for conservative businesses
(selling existing product/service offerings
with the existing revenue model to existing
customers, but augmented by data and AI)
or very disruptive businesses (entering and
disrupting new industries with completely
new data and AI-enabled product/service
offerings and revenue models). It is all
8.4 A recommended six-
step path for companies
This book has been all about exploring new data and
AI-enabled business models and co-creating them with

topic and want to start implementing the models and the

playbook presented in your own business, then the following
recommended six-step path is a good place to start (also
Murillo, Levin, Nediak & Sergienko111):

1. Understanding your existing customers enhance consumers’ everyday lives using

and potential new customers: What is your data and AI assets? What alternative
going on within your customer base? What options do customers already use, perhaps
new trends are emerging? Is disruption on delivered by your competitors?
the horizon, changing how your customers
behave in the future? Which new needs Use a service design approach to
might your customers have? How might understand the needs, wishes, fears,
they want to buy from your company in the problems to be solved and jobs to be done
future? Who is a relevant supplier for them of your various potential customer groups.
in the future? How are your customers’ Identify the opportunities and explore the
processes, daily lives and ways of working desirability of emerging solutions among
about to change? What are they likely your potential customers.
to consider valuable in the future? What
future channels and interactions will your 2. Understanding data and AI technology
customers prefer to use? Are there jobs to maturity: What is going on in the field of
be done or applications that might benefit data and AI right now? What is feasible and
from your company’s data and AI? Would what is not? What will these technologies
existing customers or completely new ones and capabilities look like a few years from
potentially be willing and able to pay for now, and what will they enable? What
your data and AI assets? How could you kind of data and AI might disrupt your
potentially augment the capabilities of current business in the future? How well
the employees in those organisations, or do your customers already understand
 Your culture Your customers
• Innovation-focused culture • Problems worth solving
• Data and AI-focused culture • Jobs to be done
• Change management • Willingness to pay
• Core vs. new business • Customer selection

6 1
Your execution Your maturity
• Prioritised roadmap Six steps towards • Strategy and positioning
• Portfolio of initiatives data and AI- • Commercial models
• Development teams
5 enabled business 2
and business acumen
models
• Iterative development • Processes and ways
• Acceleration and scaling of working
• Data and technology
Go/No-Go 4 3 • Organisations and
competences

Your business case Your ambition

• Value propositions
• Common goals
• Investments
• Proactiveness vs. reactiveness
• Revenue model selection
• Executive commitment and sponsorship
• Financial impact
• Product/service offering selection

Figure 32: A recommended six-step path for business leaders (source: Mika Ruokonen)

these technologies? How big is the effort how are they unique (Chapter 4.4.), what
to achieve the data and AI technology is their overall readiness? Do you have
expertise that your company needs, and the required skills, capabilities, mindsets
does it make sense to develop it in-house and organisational ability to support the
or acquire it from third-party suppliers? development of those assets? What is
Where would your company get enough the “gap” between your current state and
relevant data to be commercialised, state-of-the-art new technologies that are
and to also use as raw material for AI coming to the market? If there is a perfect
development? What is the current maturity AI landscape in 10 years, what is the role
level in your own firm with regards to of your company within that, and what will
data and AI? How good are your own data your business model look like?11
and AI assets: what are their strengths,
Use technology research capabilities to data and AI-enabled initiatives. Then, make
understand not only the situation today, but sure your own ambitions are aligned with
also likely developments in years to come. the incentives of the key people involved.

3. Setting the ambition: Based on the 4. Making a business case: Is it realistic

maturity analysis, what does your company to produce the data and AI capability
want to achieve with data and AI, and what needed to renew your business and open
is a logical and realistic common goal for the desired new growth avenues alone or
development? How do your anticipated with partners? What level of investment
data and AI-enabled business models link is needed to keep your company relevant
with your overall business strategy and when it comes to commercialising data
objectives? What are the characteristics and AI, and to exploit all the relevant
of your industry (B2B versus B2C, product opportunities that these technologies
versus service business) that potentially offer? Can you make sensible but adequate
allow or limit your plans to commercialise investments? Given customers’ willingness
data and AI? Does it make sense for you and ability to pay, and the anticipated value
to proactively push forward, undertake propositions and revenue models, is there a
“offensive” efforts or instead “play strong and viable business case for moving
defence” and wait to see how the market ahead?
develops? Is this topic just a side issue
in your company or truly the core of your Use a business design approach and make
strategic agenda? What are your data your business case calculations against
and AI ambitions? How will you secure the your entire Business Model Canvas, catering
necessary commitment and endorsement for both the “left side” and the “right
from key players in your company for your side”. Make sure you take into account
data and AI commercialisation efforts all the changes needed in each of the
to take shape, whatever ambition and “boxes” to implement the new data and
strategy you choose?49/52 How does the AI-enabled business models. What is the
ambition and strategy of your firm relate overall financial impact of the activities
to the partnerships you might have in this implemented around data and AI business
area; who do you need to help you reach models?
your goals? And what exactly is the game-
changing product/service offering and ---
the consequent customer experience that
you want to launch to the market to reach At this stage, after the first four steps,
that ambition, and to differentiate from management should make a well-informed
competitors that might already be there? decision to go ahead or not. Given the
ambition level, how successful are these
Get your top management involved around data and AI-enabled business model renewal
the issue and make the decisions needed to efforts likely to be? Are the new product/
unleash your commercial potential around service offerings and revenue models
desirable, feasible and viable when it come and rigorously track progress towards
to being implemented within your business? the desired impact and manage projects
If the answer is yes, then it makes sense to efficiently. You might want to set an
move on to steps five and six. If not, then aggressive timetable for the initial launch
now might not be the right time to move of the new models to be able to prove that
ahead. Instead it makes sense to look for they produce results and demonstrate
alternative potential business opportunities feasibility.72 In order to scale up your new
that might prove more lucrative. data and AI-enabled business models,
you need to understand both the long-
5. Relentless execution: As the saying term transformation in your industry
goes, ideas are cheap, execution is (e.g. disruption caused, partnerships
everything. Think backwards from your needed, long-term data infrastructure
chosen ambition, business case, product/ and governance needed) and the short-
service offerings, competence, technology term customer-focused experiments (e.g.
and revenue models for data and AI-enabled iteration, validation and pilot projects). Put
business models: What is the journey that your best people to work solving problems
you need to take to get there? What is around data and AI and give them the
your “marching order”; the portfolio of support, autonomy and space to get things
initiatives with different time horizons done quickly.
and a prioritised roadmap for the required
activities that fulfil your targets and 6. Enabling and driving culture change: Be
ambitions? aware that many new data and AI-enabled
business models require a significant
You need to figure out your team and the culture change. The more radical the
competencies needed to make your plan new models adopted (for instance those
a success. This may include qualities such related to entering completely new markets
as entrepreneurial spirit, quantitative or building completely new product/
strength and customer-focused thinking. service offerings and revenue models),
You need to continuously and iteratively the more critical it is to embrace change
test and validate new hypotheses, pivot management. What resistance have you
offerings and revenue models as needed, observed? How can you shield innovative
and continuously create, scale, operate, projects against the inertia of the core
trade and kill your business models business, while also understanding that
whenever relevant. You need to train they might have implications for the core
your algorithms in the lab, test them business execution at a later point in time?
with a limited number of customers, How can you help people to understand
then progressively launch them more the value of data in daily operations and
widely into the marketplace and scale up decision-making, and as an asset of the
operations. You need to continuously stay firm? How can you ensure that the decision-
customer-focused when implementing makers in your company understand the
the new models. You need to continuously value of data and AI, and are committed to
the plan that you have in place for the new As illustrated in Figure 32, this six-step
business models? What data and AI-related path is not a one-way street; instead,
training should you organise to prepare the planning and implementation of new
people for changes, and to make room for data and AI-enabled business models
a data/AI-ready mindset? Assuming that is commonly an iterative process. For
not all the ventures are successful, how can instance, decision-makers often think
you ensure enough tolerance for risk-taking about the customer needs, the company
with the new data and AI-enabled models? data and AI maturity and the ambitions of
How can you ensure people come up with their firm (steps 1-3) simultaneously, as
data and AI-related new ideas and identify these considerations are often intertwined.
relevant market opportunities? Similarly, when firms have progressed
with getting going with their portfolio of
You might need to push forward both initiatives or with implementing culture
a data and AI-focused and innovation- change (steps 5-6), they often need to
focused culture in your firm simultaneously, return to the question of their customer
and find a delicate balance between the understanding, data and AI maturity
two. At best, your company will have a assessment and ambitions, asking: “Are our
thriving core business, complemented assumptions in these areas still valid?” It makes
by a transformative portfolio of data and sense for companies to continuously go
AI-enabled new business models. The back and forth and examine all these six
two will operate side by side, with both steps in combination, then judge which
being actively managed for growth and of them needs particular management
profitability, creating a 1+1 = 3 impact attention.
in which they amplify one another.7 This
may occur when the new ventures get an
“unfair advantage” in the marketplace by
using the valuable and unique data and AI As you might have realised by going through
assets of the core business yet being able the six steps, we are back at the Business
to successfully execute their own business Model Canvas devised by Osterwalder and
“free from the shackles” of the core.6 Pigneur19 (see Figure 33). If a company
is to design and implement successful
Based on the co-creation sessions with business models that are enabled by data
our sample companies, it is fair to say and AI, rather than optimising individual
that steps one (understanding customers activities, it needs to review all nine of the
better) and four (making a business case) canvas’s interdependent, systemic and
seem especially challenging right now holistic “boxes” from this perspective.75
(see Chapter 7.2). But fortunately many This includes assessing which value
are also moving in the right direction, e.g. propositions (especially data and AI-enabled
by launching validation discussions with product/service offerings) it wishes to
various customers, and making data and deliver to its various customers, through
AI-related business plans. which channels, and with the help of which
resources and partners (in our case, the So, why is there a new playbook and new
emphasis lies particularly on technology, canvases in this book? And why does
strong development teams, and data and our approach differ from the traditional
algorithm providers). It is also necessary to Business Model Canvas thinking, which
see what activities are needed to execute still seems valid and relevant for various
the plan (e.g. software and algorithm data and AI-enabled ventures: why all the
development and capability development), effort? It’s a deliberate choice for these
and consequently what revenues and costs reasons:
these activities entail. • For many businesses, data and AI
are still mysterious and difficult to
We have come full circle. At the end of the understand, and people are not used
day, implementing new data and AI-enabled to dealing with them: especially when
business models differs very little from looking at them from a commercial
implementing almost any other business perspective.
model. • AI is a revolutionary general-purpose
technology which means we probably
don’t yet fully understand its business

Key Key Value Customer Customer

Partners Activities Propositions Relationships Segments

Key Channels
Resources

Cost Structure Revenue Streams

Figure 33: The Business Model Canvas

 potential, let alone any significant So, while the traditional Business Model
potential increase in its value over time. Canvas is holistic, systematic and
In this sense, investments in AI might not comprehensive, our contention is that,
be typical of capital investments that in the particular context of data and AI-
companies have been used to making.46 enabled business models, a more tailored
• In this new and somewhat unexplored approach might be helpful. But if readers
context, the playbook of data and AI- prefer to stick more closely to the Business
enabled product/service offerings and Model Canvas thinking while tackling data
revenue models (the two key elements and AI technologies, the books by Gandhi
in focus) is intended as a practical and and Ehl11 and Langa1 are worth studying.
useful tool for decision-makers to use to
simplify their business model thinking.
In addition, the four canvases, building
on Lean Service Creation, are helpful in
terms of uncovering and prioritising the
needs of selected customer groups, then
designing product/service offerings and
revenue models accordingly.
• If company leaders manage to design
these first two important elements for
their data and AI assets in a customer-
centric way, they can also quickly start
to see some other important elements,
such as the emerging cost structure,
activities and partnering needs as well.
This is often a result of the product/
service offerings and revenue model
choices.
• Finally, various aspects of data and
AI development have been presented
in this book, including assessment of
data and AI maturity, understanding
business characteristics and taking
a journey towards culture change.
These are designed to complement our
playbook and canvases in covering the
many factors that need to be considered
when designing, building and launching
successful new data and AI-enabled
businesses.
8.5 A call to action for
business leaders
It is our firm belief that commercially and/or society as a whole? Which new attractive
focused leaders should play an active
role in embracing data and AI assets The data and AI-enabled mindset is very
for business value.52 Rather than data different from a traditional managerial
and AI implementation being seen as approach that often sees fixed industry
the domain of data scientists only, it boundaries and rules of the game. The
is important that it goes hand-in-hand latter overlooks the exponential power
with a broad school of cross-business that new technologies – as well as business
commercial thinking. Truly transformative model innovations – have.
businesses are never exclusively about
discovering and commercialising a great Applying data and AI-enabled new business
technology. Instead, their success comes models often means introducing systemic
from enveloping the new technology change to a company, as leaders strive to
in an appropriate, powerful business implement new skills and ways of working.
model.72 Adding some data and AI to The change is often much broader than
processes, practices or operations does just some demonstration projects or small
not necessarily deliver enough benefits if a fixes here and there. Typically when new
business model redesign is not considered technology is deployed, it amplifies the
at the same time. Often, the data and worst features of the old way of doing
AI capabilities are universal and can be business. Individuals and organisations
applied anywhere, so harnessing them for only gradually realise, through cascading
specific business purposes often makes a networks of innovation, how to put new
lot of sense because it creates a point of technology to work properly.39 A key
difference. managerial challenge lies in recognising all
the transformative opportunities that data
Decision-makers should continuously ask: and AI will provide in the future. This book
“Could there be game-changing openings within gives some answers, but in the future there
our market using data and AI technologies and, if will be an influx of insights and evidence
so, how could we take advantage? If/when we apply that will enable firms to go much further.
these technologies to our businesses, which new
business opportunities could we start creating? Is Data and AI should no longer be treated
there an opportunity to leverage these technologies as belonging to a separate set of skills,
not only within our own company but more widely specific job descriptions or business
across our customer base, industry, ecosystem functions. Instead, as time-honoured
Acknowledgements
1. Langa, Bert (2018): Building Successful Business Models Based on Artificial Intelligence.
A book available on Amazon: https://www.amazon.com/Building-Successful-Business-
Artificial-Intelligence-ebook/dp/B07CHLG9ZQ

2. Brynjolfsson, Erik & Rock, Daniel & Syverson, Chad (2017): Artificial Intelligence and the
Modern Productivity Paradox – A clash of expectations and statistics. Working paper.
https://www.nber.org/papers/w24001.pdf

3. Varadharajan, Deepashri (2019): State of Artificial Intelligence in China. Materials of a CB

Insights webinar, available here: https://www.cbinsights.com/research/briefing/china-in-ai-
trends/

4. Finn, Ed (2017): What Algorithms Want: Imagination in the Age of Computing. The MIT Press.

5. Bughin, Jacques & Hazan, Eric (2017): Five Management Strategies for Getting the Most
from AI. MIT Sloan Management Review article available here: https://sloanreview.mit.edu/
article/five-management-strategies-for-getting-the-most-from-ai/

6. McKinsey (2019a): Building New Businesses: How Incumbents Use Their Advantages to
Accelerate Growth. A research report available here: https://www.mckinsey.com/business-
functions/mckinsey-digital/our-insights/building-new-businesses-how-incumbents-use-
their-advantages-to-accelerate-growth

7. Osterwalder, Alexander & Pigneur, Yves & Etiemble, Fred & Smith, Alan (2020): The Invincible
Company: How to Constantly Reinvent Your Organization with Inspiration from the World’s
Best Business Models. John Wiley & Sons Inc.

8. Ruokonen, Mika (2019): Three Steps to Extracting Business Value from AI. A Futurice blog
post: https://www.futurice.com/blog/three-steps-to-extracting-business-value-from-ai

9. Johnson, Mark W. (2018): Reinvent Your Business Model: How to Seize the White Space for
Transformative Growth. Harvard Business Press.

10. Brynjolfsson, Erik & McAfee, Andrew (2017): The Business of Artificial Intelligence.
Harvard Business Review. https://hbr.org/cover-story/2017/07/the-business-of-artificial-
intelligence

11. Gandhi, Sharad & Ehl, Christian (2017): AI&U – Translating Artificial Intelligence into
Business. A book available on Amazon: https://www.amazon.com/AI-Translating-Artificial-
Intelligence-Business/dp/1521717206

12. Agrawal, Ajay & Goldfarb, Avi & Gans, Joshua (2018): Prediction Machines. Harvard Business
Review Press

13. Brynjolfsson, Erik & McAfee, Andrew (2014): The Second Machine Age: Work, Progress, and
Prosperity in a Time of Brilliant Technologies. W. W. Norton & Company.
14. Rouhiainen, Lasse (2018): Artificial Intelligence: 101 Things You Must Know Today About our
Future. CreateSpace Independent Publishing Platform.

15. Susskind, Jamie (2018): Future Politics: Living Together in a World Transformed by Tech.
Oxford University Press.

16. Tegmark, Max (2017): Life 3.0: Being Human in the Age of Artificial Intelligence. Vintage
Books.

17. Tunguz, Tomasz & Bien, Frank (2016): Winning with Data – Transform Your Culture, Empower
Your People, and Shape the Future. A book published by John Wiley & Sons.

18. Johnson, Mark W. (2010): Seizing the White Space: Business Model Innovation for Growth
and Renewal. Harvard Business Press.

19. Osterwalder, Alexander & Pigneur, Yves (2010): Business Model Generation: A Handbook for
Visionaries, Game Changers, and Challengers. John Wiley & Sons Inc.

20. Rissanen, Tommi (2019): Perspectives on Business Model Experimentation in

Internationalizing High-tech Companies. A Doctoral Dissertation at Lappeenranta University
of Technology.

21. Chollet, Francois & Allaire, Joseph J. (2018): Deep Learning with R. Manning Publications.

22. Gray, William (2019): Machine Learning with Python: An introduction to Data Science
with useful concepts and examples, step by step, learning to use Python. Independently
published.

23. Raschka, Sebastian & Mirjalili, Vahid (2017): Python Machine Learning: Machine Learning and
Deep Learning with Python, scikit-learn, and TensorFlow. 2nd Edition. Packt Publishing.

24. CB Insights (2020): Company webpage, available here: www.cbinsights.com

25. Boeing AnalytX (2020): Company webpage, available here: http://www.boeing.com/

company/key-orgs/analytx/index.page

26. Thistlethwaite, George (2018): IBM & Maersk Launch TradeLens Blockchain Shipping
Platform. An online article, available here: https://www.gbnews.ch/ibm-maersk-launch-
tradelens-blockchain-shipping-platform/

27. Lewis, Alan & McKone, Dan (2016): To Get More Value from Your Data, Sell It. Harvard
Business Review. https://hbr.org/2016/10/to-get-more-value-from-your-data-sell-it

28. McKinsey (2017): Fueling Growth Through Data Monetization. A survey report downloadable
here: https://www.mckinsey.com/business-functions/mckinsey-analytics/our-insights/
fueling-growth-through-data-monetization

29. Miller, Ron (2014): Actually, Every Company is a Big Data Company. A TechCrunch online
article available here: https://techcrunch.com/2014/05/22/actually-every-company-is-a-
big-data-company/

30. Gandhi, Suketu & Thota, Bharath & Kuchembuck, Renata & Swartz, Joshua (2018):
Demystifying Data Monetization. A MIT Sloan Management Review article, available here:
https://sloanreview.mit.edu/article/demystifying-data-monetization/
31. Spijker, Arent van ’t (2014): The New Oil: Using Innovative Business Models to Turn Data into
Profit. Technics Publications.

32. Bean, Randy (2017): How Companies Say They’re Using Big Data. Harvard Business Review.
https://hbr.org/2017/04/how-companies-say-theyre-using-big-data

33. Ruokonen, Mika (2017): Six Inspirational Ways to Make Money with Data. A Futurice blog
post: https://www.futurice.com/blog/six-inspirational-ways-to-make-money-with-data

34. Laitila, Miikka & Gustafsson, Robin & Parvinen, Petri & Pöyry, Essi & Rossi, Matti (2019):
Advancing Data Monetization and the Creation of Data-based Business models. Unpublished
research paper by Aalto University, forthcoming in Communications of the Association for
Information Systems

35. MIT Technology Review Insights (2020): The Global AI Agenda - Promise, Reality, and
a Future of Data Sharing. A research report available here: https://mittrinsights.
s3.amazonaws.com/AIagenda2020/GlobalAIagenda.pdf

36. Wikipedia (2020b): Data Monetization. A web page: https://en.wikipedia.org/wiki/Data_

monetization

37. D’Addario, Josh (2020): Seven Reasons Why Businesses Should be Sharing Data. The
Open Data Institute post available here: https://theodi.org/article/seven-reasons-why-
businesses-should-be-sharing-data/

38. Williamson, Peter & De Meyer, Arnoud (2019): How to Monetize a Business Ecosystem.
Harvard Business Review. https://hbr.org/2019/09/how-to-monetize-a-business-ecosystem

39. O’Reilly, Tim (2017): Wtf?: What’s the Future and Why It’s Up to Us. Harper Business.

40. Wikipedia (2020a): Artificial Intelligence. A web page: https://en.wikipedia.org/wiki/

Artificial_intelligence

41. Microsoft (2018): Artificial Intelligence in Europe – Finland Outlook for 2019 and Beyond. A
research report: https://pulse.microsoft.com/fi-fi/business-leadership-fi-fi/na/fa1-suomi-
euroopan-karkea-tekoalyn-hyodyntamisessa-silti-yli-puolet-yrityksista-vasta-suunnittelee-
tai-pilotoi-tekoalyn-kayttoa/

42. Jääskeläinen, Atte (2020): Predicting the Success of News Content: An Algorithmic
Experiment in the Real World. An online article available here: https://blogs.lse.ac.uk/
polis/2020/02/18/ predicting-the-success-of-news-content/

43. Crosman, Penny (2016): How PayPal is Taking a Chance on AI to Fight Fraud. An online article
available here: https://www.americanbanker.com/news/how-paypal-is-taking-a-chance-on-
ai-to-fight-fraud

44. Rosenbaum, Eric (2017): IBM’s Watson Supercomputer is Getting into Wall Street Stock-
picking. An online article, available here: https://www.cnbc.com/2017/06/16/ai-assault-on-
stock-market-ibms-watson-is-getting-into-etf-business.html

45. Futurice (2019a): Data & AI: Why it is Time to Stop Talking and Start Doing. A handbook
available here: https://hello.futurice.com/futurice-data-ai-handbook

46. Daugherty, Paul R. & Wilson, H. James R. (2018): Human + Machine: Working in the Age of AI.
Harvard Business Review Press.
47. Mayer-Schönberger, Viktor & Ramge, Thomas (2018): Reinventing Capitalism in the Age of
Big Data. John Murray Publishers Ltd.

48. Ross, Jeanne (2018): The Fundamental Flaw in AI Implementation. MIT Sloan Management
Review, 59(2), available here: https://sloanreview.mit.edu/article/the-fundamental-flaw-in-
ai-implementation/

49. Ransbotham, Sam & Gerbert, Philipp & Reeves, Martin & Kiron, David & Spira, Michael
(2018): Artificial Intelligence in Business Gets Real - Pioneering companies aim for AI at
scale. A MIT Sloan Management Review, https://sloanreview.mit.edu/projects/artificial-
intelligence-in-business-gets-real/

50. McKinsey (2019b): Global AI Survey: AI Proves its Worth, but Few Scale Impact. A research
report available here: https://www.mckinsey.com/featured-insights/artificial-intelligence/
global-ai-survey-ai-proves-its-worth-but-few-scale-impact

51. PwC (2019): 2020 AI Predictions: Five Ways to go From Reality Check to Real-world Payoff.
Survey results available here: https://www.pwc.com/us/en/services/consulting/library/
artificial-intelligence-predictions-2020.html

52. MIT (2020): How AI Changes the Rules: New Imperatives for the Intelligent Organization. MIT
SMR Connections research report, available here: https://sloanreview.mit.edu/offer-sas-
how-ai-changes-the-rules-2020/

53. Fountaine, Tim & McCarthy, Brian & Saleh, Tamim (2019): Building the AI-powered
Organization. Harvard Business Review, July-August issue: https://hbr.org/2019/07/
building-the-ai-powered-organization

54. Coveyduc, Jeffrey L. & Anderson, Jason L. (2020): Artificial Intelligence for Business – A
Roadmap for Getting Started with AI. John Wiley & Sons Inc.

55. Sinfield, Joseph V. & Calder, Edward & McConnell, Bernard & Colson, Steve (2012): How
to Identify New Business Models. A MIT Sloan Management Review article, available here.
https://sloanreview.mit.edu/article/how-to-identify-new-business-models

56. Workera (2020): AI Career Pathways: Put Yourself on the Right Track. A report available
here: https://workera.ai/candidates/report/

57. British Design Council (2005): What is the Framework for Innovation? Design Council’s
Evolved Double Diamond. A blog post about the design methodology here: https://www.
designcouncil.org.uk/news-opinion/what-framework-innovation-design-councils-evolved-
double-diamond

58. Costa, Rebeca (2018): The Double Diamond Model: what is it and should you use it? A blog
post available here: https://www.justinmind.com/blog/double-diamond-model-what-is-
should-you-use/

59. Marr, Bernard (2017): Where Can you Buy Big Data? Here Are the Biggest Consumer
Data Brokers. A Forbes article available here: https://www.forbes.com/sites/
bernardmarr/2017/09/07 /where-can-you-buy-big-data-here-are-the-biggest-consumer-
data-brokers/#21413f446c27

60. Burke, Elaine (2015): Data Scientists Rejoice! There’s an Online Marketplace Selling
Algorithms from Academics. An online article, available here: https://www.siliconrepublic.
 com/start-ups/data-scientists-rejoice-theres-an-online-marketplace-selling-algorithms-
from-academics

61. Chowdhry, Amit (2018: How Algorithmia Built the Largest Marketplace for Algorithms
in the World. A Forbes online article, available here: https://www.forbes.com/sites/
amitchowdhry/2018/01/22 /how-algorithmia-built-the-largest-marketplace-for-algorithms-
in-the-world/#463cab0f53fd

62. Magretta, Joan (2002): Why Business Models Matter. Harvard Business Review, May issue.
https://hbr.org/2002/05/why-business-models-matter

63. Lewis, Michael (1999): The New New Thing: A Silicon Valley Story. W. W. Norton & Company

64. Drucker, Peter F. (1994): The Theory of the Business. Harvard Business Review, September-
October. https://hbr.org/1994/09/the-theory-of-the-business

65. Taran, Yariv & Nielsen, Christian & Montemari, Marco & Thomsen, Peter & Paolone,
Francesco (2016) Business Model Configurations: a Five-V Framework to Map Out Potential
Innovation Routes. European Journal of Innovation Management, Vol. 19, Issue 4.

66. Iansiti, Marco & Lakhani, Karim R. (2020): Competing in the Age of AI: Strategy and
Leadership When Algorithms and Networks Run the World. Harvard Business Review Press.

67. Clauss, Thomas (2017): Measuring Business Model Innovation: Conceptualization, Scale
Development, and Proof of Performance. R&D Management 47(3).

68. Rissanen, Tommi & Sainio, Liisa-Maija (2016): Business Model Experimentation in Incumbent
and Startup Companies. Proceedings of XXVII ISPIM Innovation conference, 2016, Porto,
Portugal.

69. Ries, Eric (2011): The Lean Startup. Crown Publishing Group

70. Sund, Kristian J. & Bogers, Marcel & Villarroel, J. Andrei & Foss, Nicolai J. (2016): Managing
Tensions Between New and Existing Business Models. A MIT Sloan Management Review
article available here: https://sloanreview.mit.edu/article/managing-tensions-between-new-
and-existing-business-models/

71. Rissanen, Tommi & Karhu, Päivi (2017): New Organization Forms of Innovation: What is
Business Model Ambidexterity? Proceedings of ISPIM Innovation Summit, 2017, Melbourne,
Australia.

72. Johnson, Mark W. & Christensen, Clayton M. & Kagermann, Henning (2008): Reinventing
Your Business Model. Harvard Business Review. https://hbr.org/2008/12/reinventing-your-
business-model

73. Zawacki, Mark & Ruokonen, Mika (2017): Radical Innovation Must Come From the Outside,
Not Within. A Futurice blog post: https://www.futurice.com/blog/radical-innovation-must-
come-from-the-outside-not-within/

74. Ansoff, H. Igor (1957): Strategies for Diversification. Harvard Business Review, available
here: http://sgpwe.izt.uam.mx/files/users/uami/sppc/13O/Gestion_y_Control_
Estrategico_I/Lectura_10._Ansoff_Igor_Strategies_for_Diversification.pdf

75. Amit, Raphael & Zott, Christoph (2012): Creating Value Through Business Model Innovation.
MIT Sloan Management Review, Vol. 53, No.2.
76. Bessis, David (2018): Building AI-first Products. An article on Medium, available here: https://
medium.com/@davidbessis/building-ai-first-products-90d503ccd43a

77. Kemppi (2016): Kemppi Opens Up Welding Software APIs. Company news available here:
https://www.kemppi.com/en-US/news-events/news/kemppi-opens-up-welding-software-
apis/

78. Haselton, Todd & Kolodny, Lora (2018): Tesla Starts Selling its Own Car Insurance in
California. News by CNBC, available here: https://www.cnbc.com/2019/08/28/tesla-starts-
selling-its-own-car-insurance-in-california.html

79. Davies, Jamie (2019): Orange Bank is on a Roll. News by Telecoms.com, available here:
https://telecoms.com/495886/orange-bank-is-on-a-roll/

80. Jenkins, Patrick (2018): Orange Bank: is a Phone Company the Future of Fintech? News by
Financial Times, available here: https://www.ft.com/content/6bd8ac00-f7c4-11e7-88f7-
5465a6ce1a00

81. Nikolaeva, Maya & Sassard, Sophie & Barzic, Gwénaëlle (2017): Exclusive - Orange is the
New Bank? Telecoms Giant Ventures into Lending. News by Reuters, available here: https://
www.reuters.com/article/us-france-banks-telecoms-exclusive/exclusive-orange-is-the-new-
bank-telecoms-giant-ventures-into-lending-idUSKBN1D0298

82. GP.Bullhound (2019): Smart Manufacturing – The Rise of the Machines. A downloadable
research report: https://www.gpbullhound.com/insights/smart-manufacturing/

83. Kalmar (2019): Performance-based Contracts: The Next Step for the Container Handling
Industry. An online article by the Kalmar/Cargotec company, available here: https://www.
kalmarglobal.com/news--insights/articles/2019/20190920_performance-based-contracts--
the-next-step-for-the-container-handling-industry/

84. Marr, Bernard (2019): Coca-Cola: Driving Success with AI and Big Data. An online article
available here: https://www.bernardmarr.com/default.asp?contentID=1280

85. De Saulles, Martin (2018): Building a Data-driven Business Model. A blog post on Information
Matters website, https://informationmatters.net/data-driven-business-models/

86. Open Data Insititute (2018): The Role of Data in AI Business Models. A report available here:
https://docs.google.com/document/d/14g0p6KSyH1r1J_PrykJIXUX-rdeP1B4CLIffAyFPOnk/
edit

87. Tanner, Adam (2016): How Data Brokers Make Money off Your Medical Records. A Scientific
American article, available here: https://www.scientificamerican.com/article/how-data-
brokers-make-money-off-your-medical-records/

88. Reuters (2017): Roomba Vacuum Maker iRobot Betting Big on the ‘Smart’ Home. An
online article available here: https://www.reuters.com/article/us-irobot-strategy-
idUSKBN1A91A5?il=0

89. Fitbit (2020): How to Rest Easier with Fitbit’s Sleep Tools. A blog post by Fitbit, available
here: https://blog.fitbit.com/fitbit-sleep-tools/
90. Syrjänen, Tuomas (2018): How Artificial Intelligence Can Help Business Model Innovation. A
blog in Director’s Institute Finland web page: https://dif.fi/teema-artikkelit/2018/q4-2018-
hallitus-ja-tekoaly/how-artificial-intelligence-can-help-business-model-innovation/

91. Moazed, Alex (2018): Platform Business Model – Definition: What is it? An online article
available here: https://www.applicoinc.com/blog/what-is-a-platform-business-model/

92. Zhu, Feng and Iansiti, Marco (2019): Why Some Platforms Thrive and Others Don’t. Harvard
Business Review, January–February 2019 Issue. Available here: https://hbr.org/2019/01/
why-some-platforms-thrive-and-others-dont

93. Eagle Alpha (2020): Company webpage, available here: https://eaglealpha.com/

94. DC Rainmaker (2019): Suunto 9 Series Gets New Sleep & Fitness Metrics from FirstBeat. An
online article available here: https://www.dcrainmaker.com/2019/10/suunto-9-series-gets-
new-sleep-fitness-metrics-from-firstbeat.html

95. Lassila & Tikanoja (2019): Opetetaan Kiinteistöt Fiksuille Tavoille: L&T Smartti Parantaa
Kiinteistön Energiatehokkuutta – Olosuhteista Tinkimättä. Service concept introduction in
Finnish language on the company webpage, available here: https://www.lt.fi/fi/smartti

96. Orica (2020): Company webpage, available here: https://www.orica.com/

97. Connerty, Michael & Navales, Eric & Kenney, Chris & Bhatia, Tanay (2016): Manufacturing
Companies Need to Sell Outcomes, Not Products. A Harvard Business Review online article,
available here: https://hbr.org/2016/06/manufacturing-companies-need-to-sell-outcomes-
not-products

98. Grove, Hannah & Sellers, Kevin & Ettenson, Richard & Knowles, Jonathan (2018): Selling
Solutions Isn’t Enough. A MIT Sloan Management Review article, available here: https://
sloanreview.mit.edu/article/selling-solutions-isnt-enough/

99. Gillin, Paul (2017): Selling ‘Outcomes’: GE Shows That Old Rules Don’t Apply to Digital
Business. An online article at SiliconANGLE Media, available here: https://siliconangle.
com/2017/06/24/ selling-outcomes-ge-shows-old-rules-dont-apply-digital-business/

100. Avaus (2019): The Avaus Cookbook: AI in Marketing and Sales. A report available here:
https://www.avaus.com/cookbook/

101. KPMG (2015): Framing a Winning Data Monetization Strategy. A report that is available here:
https://assets.kpmg/content/dam/kpmg/pdf/2015/10/framing-a-winning-data.pdf

102. Amazon Web Services (2019): AWS Data Exchange – Find, Subscribe To, and Use Data
Products. A marketplace containing over one thousand licensable data products from over
80 data providers: https://aws.amazon.com/blogs/aws/aws-data-exchange-find-subscribe-
to-and-use-data-products/

103. Reeves, Martin & Lotan, Hen & Legrand, Julien & Jacobides Michael G. (2019): How Business
Ecosystems Rise (and Often Fall). A MIT Sloan Management Review article, available here:
https://sloanreview.mit.edu/article/how-business-ecosystems-rise-and-often-fall/

104. Futurice (2019b): Lean Service Creation. A curated set of canvases for creating services and
products, available here: https://leanservicecreation.com/
105. McKinsey (2014): Manager and Machine: The New Leadership Equation. McKinsey Quarterly,
September.

106. IDEO (2019): How to Prototype a New Business. A blog post available here: https://www.
ideou.com/blogs/inspiration/how-to-prototype-a-new-business

107. Khodabandeh, Shervin & Fehling, Ronny & LaFountain, Burt & Duranton, Sylvain & Gerbert,
Philipp & Paetsch, Annais & Reeves, Martin (2019): How to Win with Artificial Intelligence. A
BCG publication, available here: https://www.bcg.com/publications/2019/how-to-win-with-
artificial-intelligence-ai.aspx

108. Lovelock, James (2019): Novacene: The Coming Age of Hyperintelligence. Allen Lane.

109. Higgs, John (2019): The Future Starts Here: Adventures in the Twenty-first Century.
Weidenfeld & Nicolson.

110. Rose, Nicolas (2018): 100 European Startups that Manage Personal Data. An article in
Medium, available here: https://medium.com/xangevc/100-european-startups-that-manage-
personal-data-5f564db2aec8

111. Murillo, Jose A. & Levin, Yuri & Nediak, Mikhail & Sergienko, Ivan A. (2018): How One
Company Made its Analytics Investment Pay Off. Harvard Business Review. https://hbr.
org/2018/01/how-one-company-made-its-analytics-investment-pay-off

112. Seppälä, Timo & Hakanen, Esko & Lähteenmäki, Ilkka & Mattila, Juri & Niemi, Rasmus
(2019): The Resource Dependency of Data: A Prospective on Data sharing in Supply Chains.
A research paper available here: https://papers.ssrn.com/sol3/papers.cfm?abstract_
id=3493650

113. Shaw, Greg (2019): The Future Computed: AI and Manufacturing. A book published by
Microsoft. Available here: https://news.microsoft.com/futurecomputed/

114. German Federal Ministry for Economic Affairs and Energy (2019): Project GAIA-XA:
Federated Data Infrastructure as the Cradle of a Vibrant European Ecosystem. A statement
and a plan available here: https://www.bmwi.de/Redaktion/EN/Publikationen/Digitale-Welt/
project-gaia-x.pdf?__blob=publicationFile&v=4

115. IBM (2016): Extracting Business Value from the 4 V’s of Big Data. An infographic available
here: https://www.ibmbigdatahub.com/infographic/extracting-business-value-4-vs-big-data

116. Slotin, Jenna (2018): What Do We Know About the Value of Data? A study at Global
Partnership for Sustainable Development Data, available here: http://www.data4sdgs.org/
sites/default/files/ services_files/ Value%20of%20Data%20Report_Final_compressed_0.pdf

117. Weir, Daryl (2017): Fantastic Problems and Where to Find Them. A Futurice blog post:
https://www.futurice.com/blog/fantastic-problems-and-where-to-find-them/

118. Polson, Nick & Scott, James (2018): AIQ: How People and Machines are Smarter Together. St.
Martin’s Press.

119. Silo.AI (2019): AI and Traditional Analytics. A whitepaper by Silo.AI company, available here:
https://silo.ai/blog/#white-papers
120. Ajanki, Antti (2018): Differences Between Machine Learning and Software Engineering. A
Futurice blog post: https://www.futurice.com/blog/differences-between-machine-learning-
and-software-engineering/

121. CrowdFlower (2017): 2017 Data Scientist Report. A survey of which results are
available here: https://visit.figure-eight.com/rs/416-ZBE-142/images/CrowdFlower_
DataScienceReport.pdf

122. Shewan, Dan (2019): 10 Companies Using Machine Learning in Cool Ways. A blog post:
https://www.wordstream.com/blog/ws/2017/07/28/machine-learning-applications

123. Sullivan, Josh & Zutavern, Angela (2017): The Mathematical Corporation: Where Machine
Intelligence and Human Ingenuity Achieve the Impossible. PublicAffairs.

124. Oh, Jasmine (2019): Yes, AI Will Replace Designers. An article on Medium, available here:
https://medium.com/microsoft-design/yes-ai-will-replace-designers-9d90c6e34502

125. Brynjolfsson, Erik & Collins, Avinash (2019): How Should We Measure the Digital Economy?
Harvard Business Review, November-December, digitally available here: https://hbr.
org/2019/11/how-should-we-measure-the-digital-economy

126. Gerbert, Philipp & Reeves, Martin & Ransbotham, Sam & Kiron, David & Spira, Michael
(2018): Global Competition with AI in Business: How China Differs. A MIT Sloan Management
Review article, available here: https://sloanreview.mit.edu/article/global-competition-of-ai-
in-business-how-china-differs/

127. Lee, Kai-Fu (2018): AI Superpowers: China, Silicon Valley, and the new world order. Houghton
Mifflin Harcourt.

128. World Economic Forum (2019): 6 Things to Know about China’s Historic rRse. An online
article, available here: https://www.weforum.org/agenda/2019/10/china-economy-
anniversary/

129. Dai, Sarah (2019): AI’s Ability to Translate into Real Economic Growth Needs a Reality
Check Warn Chinese Scholars. An online article in South China Morning Post, available here:
https://www.scmp.com/tech/policy/article/3034341/ais-ability-translate-real-economic-
growth-needs-reality-check-warn

130. Casadesus-Masanell, Ramon & Ricart, Joan Enric (2009): From Strategy to Business Models
and to Tactics. A working paper at Harvard Business School, available here: https://www.
hbs.edu/faculty/Publication%20Files/10-036.pdf

131. Spieth, Patrick & Schneckenberg, Dirk & Ricart, Joan E. (2014): Business Model Innovation –
State of the Art and Auture Challenges for the Field. R&D Management 44(3).

132. Hartmann, Philipp Max & Zaki, Mohamed & Feldmann, Niels & Neely, Andy (2014): Big
Data for Business? A Taxonomy of Data-driven Business Models Used by Start-up Firms. A
working paper at University of Cambridge. https://cambridgeservicealliance.eng.cam.ac.uk/
resources/ Downloads/Monthly%20Papers/2014_March_DataDrivenBusinessModels.pdf

133. Keen, Andrew (2018): How to Fix the Future: Staying Human in the Digital Age. Atlantic
Books.
Image Credits
Icons in Figure 8
Noun Project: https://thenounproject.com/
Box by Justiconnic from the Noun Project
Factory Created by Maxim Kulikov from the Noun Project
Group by Gregor Cresnar from the Noun Project
Handshake Created by Silvo from the Noun Project
Dashboard Created by ProSymbols from the Noun Project
Cloud Created by Iga from the Noun Project

Photography in Figure 9
Unsplash: https://unsplash.com/
Pexels: https://www.pexels.com/
• Ching, Hong Y. & Fauvel, Clemens (2013): Criticisms, Variations and Experiences with
Business Model Canvas. European Journal of Agriculture and Forestry Research, Vol.1. No.2.

• Laitila, Miikka (2017): Data Monetization: Utilizing Data as an Asset to Generate New
Revenues for Firms. A pro gradu thesis at Aalto University: https://aaltodoc.aalto.fi/
handle/ 123456789/28934

• Ovans, Andrea (2015): What is a Business Model? A Harvard Business Review article:
https://hbr.org/2015/01/what-is-a-business-model

• Redman, Thomas C. (2015): 4 Business Models for the Data Age. Harvard Business Review,
https://hbr.org/2015/05/4-business-models-for-the-data-age

• Ruokonen, Mika & Syrjänen, Tuomas & Siilasmaa, Risto (2019): Mystiikan Verho Pitää
Repäistä Pois Tekoälyn Päältä. An opinion piece in Finnish language in the Talouselämä
magazine: https://www.talouselama.fi/uutiset/tebatti-mystiikan-verho-pitaa-repaista-
pois-tekoalyn-paalta/c8bd8e24-31bb-4b96-85b5-4c81406a6f5b

Futurice publications and blogs:

• Growth Reinvented: turn your data and AI into money
www.futurice.com/growth-reinvented

• Three steps to extracting business value from AI

https://www.futurice.com/blog/three-steps-to-extracting-business-value-from-ai

• Six inspirational ways to make money with data

https://futurice.com/blog/six-inspirational-ways-to-make-money-with-data

• Business model innovation helps industrial firms move beyond COVID-19

https://futurice.com/blog/business-model-innovation-helps-industrial-firms-move-
beyond-covid-19

• Radical innovation must come from the outside – not within

https://futurice.com/blog/radical-innovation-must-come-from-the-outside-not-within
What exactly is data? Our definition describe a larger phenomenon, and spark
considers data as the entire digital further creativity on how to use it.31 Going
footprint either of people (e.g. private forward, it is likely that new inventions e.g.
individuals or company employees) or of in areas of satellites, drones, mobile devices
machines (whether operated by people and other equipment will help companies
or autonomously). Data can be generated to gather yet more valuable data to benefit
as a result of online activity, e.g. someone their business and operations.
browsing a web page, a portal or an online
marketplace. Data can also be generated In order to set the framework for discussing
with the help of sensors, for example from the commercial aspects of data, we make
equipment that performs various tasks the following assumptions:
in production plants or warehouses, and • We assume that a company has data in
machines that measure weather conditions its possession, whether collected as a
or serve as part of the infrastructure of any result of its own operations or received
given city or country. In addition, data can from elsewhere for free, or through an
be generated via transactions: for example, exchange or trading process.
when companies or people sell goods, trade • The company has adequate legal rights
stocks on the stock market, or deliver pro- and user consent for storing, analysing
ducts or services to one another. And so on. and using the data (beware that this is
not self-evident in many cases).
There are therefore a multitude of digital • The data of the company is in a format
activities that generate data.29 Indeed, that can be used for commercial
various studies report that the amount purposes: it is, for example, parsed,
of data available has exploded in recent cleaned, aggregated and/or catalogued
years. This is due to a dramatic increase adequately, and kept up to date.
in sensor and GPS functionality, as well as • The company has made a deliberate
the level of always-on internet connections choice (or has an ambition or aim) to
in use.31 You only need to monitor your own commercialise its data, with a view
digital activity on any given day to realise to enriching its services and building
that we have all become “data generation revenue models around it.
machines” without really even realising it. • There are no legal or ethical matters
Often, data is like Lego toy blocks: it comes that would stand in the way of
in small pieces that can be joined to create commercialising the data: for instance,
a bigger shape. When combined it may privacy or security issues, or risk of
 harm to other people (again, this is not easier and cheaper than ever before: there
always self-evident and needs careful is no need to get rid of data once it is used.
consideration). Therefore, the value of the data collected
and stored often accumulates over time, as
When it comes to harnessing data for companies can make comparisons and see
business purposes, is some data better important trends develop (of course, this
than others? It depends. Some datasets accumulation is not inherently valuable: it
might be richer, more unique or more depends on how it is being used). As long as
complete than others. And some data might companies can continue to add computing
have legal or ethical problems that prevent and storage capacity to their on-demand
a company from using it commercially. cloud-based technology infrastructures,
Information on private individuals their data-enabled operations are almost
often falls into this category, especially infinitely scalable.66
sensitive data such as medical records
or financial or biometric information. In Often, any data problems that companies
addition, companies often sign contracts encounter begin after the data is collected,
which set confidentiality clauses which and they tend to revolve around practical
restrict the use of their data or make the access. Business leaders may come up
entire relationship between two firms against issues such as, “Who knows about
confidential.112 the various data that we have in the
company?”, “Where is the relevant data
From a purely technical perspective, data is stored?”, “How can we move data from
data. No matter how it’s collected or where various sources into a common repository
it’s stored, data for computers is “ones (e.g. a data lake) so that it is more usable?”
and zeros”, and so constitutes a standard and so on. Often decision makers do not
body of information that’s waiting to be anticipate the sheer level of effort required
processed. in moving from merely collecting data to
accessing and analysing it.66
For many companies, collecting data is no
longer an issue. Companies have learned And, of course, not all companies can
to design products that, when used by collect or access any data they want. Even
customers, generate the exact and unique with all the streamlining and availability
data that they need in order to develop developments we’ve seen in recent years,
their business. Sensor and analytics the topic of data ownership still provokes
technologies are widespread and used a lot of discussion among business
by most firms. The costs of storing data, leaders. It also creates conflicts between
whether in a company’s own servers or firms, and plenty of work for lawyers. In
in the cloud, have reduced dramatically addition, data access is far from a level
and are expected to decrease further in playing field among different companies,
the future. This means that collecting and or even different countries. Large global
storing data over long periods of time is digital giants such as Google, Facebook
and Amazon have built unparalleled data- How exactly is the value of a company’s
generation engines for themselves, allowing data measured? Many observers assume
endless opportunity to build sophisticated that the more data a company possesses,
analytics and intelligence on top of their and the better the quality and reliability
data.86 And companies in China, for example, of its data, the more value it has because
benefit from their large and homogeneous there are better opportunities to identify
home market. China’s limited privacy important patterns and build models from
protection for citizens gives businesses it. And certainly, quantity and quality of
free rein to collect massive amounts of data data both have an important role to play.
for the benefit of their AI development.12
At the same time, small companies and/ However, they do not provide the full
or companies from countries where strict picture. As well as big data, “small data” can
privacy regulation prevails must make do also be extremely valuable. Some industrial
with the smaller amounts of data at their companies with only a few large customers
disposal.86 might choose to analyse them in detail to
understand the customer behaviour, needs,
There are at least two potential solutions hopes and fears. In this scenario, unique
for this issue of data equality. The first one data on a micro level is more important and
is to get the players in a certain industry valuable than data on a macro level. The
to agree on “data commons” efforts that same logic applies to a start-up company
will provide various companies within that that runs an online shop as its only product.
industry with access to rich and diverse Here again, micro data on the shop’s
datasets that can be used to develop their performance is critical, whereas other
AI solutions (this already seems to be larger, more generic bodies of data do not
happening in the healthcare sector35/87) hold the same value.
The second option is to make public sector
data broadly available to help to accelerate In addition, real-time data may also be
AI development in a given country or extremely valuable, even if it’s only available
industry. The public sector could also in small quantities. The value of large
potentially promote methods, tools, rules datasets may decrease over time as data
and processes for combining public and becomes outdated, making it difficult to use
private data.112/113 For instance, the German for modelling or forecasting. In contrast,
government has already launched Project real-time data allows a company to respond
Gaia-X to promote open common European rapidly to changes in the marketplace or in
data infrastructure. The project covers its own operations (something that might
data sovereignty, interoperability and be impossible to do if data is weeks or
interconnectivity, the use of transparently months old).
secure and open technologies, and other
elements that could enable European So, value can be derived from a variety
companies to come together and use data of datasets. Ideally, a company will have
more efficiently.114 access to large, rich datasets in order to
understand dependencies, correlations and
causalities across wide-ranging events or
topics. And it will also be able to deep dive
into very targeted data on specific events
and topics. Real-time data is the icing on the
cake to these datasets, allowing businesses
to be reactive and make quick changes
wherever necessary.

As described by IBM,115 to get value out of

data, a company needs to think of four Vs
as follows:
• Volume (scale of data)
• Velocity (speed of data)
• Veracity (certainty of data)
• Variety (diversity of data)

In addition, the more prepared and well

packaged commercialised data is, the more
value it typically has.

As with other company assets, business

leaders can use various approaches to
assess the value and set prices for their
data. These include cost-based approaches
(“How much does it cost to produce the
data?”), market-based approaches (“How
much are the customers willing to pay?”)
and impact/benefit-based approaches
(“How much new revenue could be
generated, or how many inefficiencies
will be avoided, as a result of the data
asset?”)116 It is good to bear in mind that
unlike some corporate assets, data tends to
constantly evolve. It may even appreciate
and depreciate at the same time. So,
price-setting should be flexible enough
to recognise the dynamic nature of the
asset.101
Artificial intelligence (AI), machine
Artificial
learning and deep learning are three
Intelligence
terms that are often used interchangeably
to describe software that behaves
intelligently. However, there are clear Machine
distinctions between them. The term AI Learning
often comprises computer systems that
accomplish goals in certain task(s). So, AI
is not a single entity; it is a combination of
various technologies, including machine Deep
learning. Machine learning, in turn, is a Learning
toolbox of algorithms and techniques
that learns rules from data and is used
to implement AI in narrow tasks. And
deep learning is one of the most popular
tools in that machine learning toolbox: a
subfield of machine learning concerned
with algorithms inspired by a structure and Figure 34: The relationship between
function of the brain called artificial neural artificial intelligence, machine learning and
networks. It is particularly effective when a deep learning
dataset is very large. Figure 34 illustrates
the relationship between these three key how to predict a target value for each input
terms. value. There are an unlimited number of
algorithms for any given task: some better
AI is all about algorithms. An algorithm is a and some worse. The purpose of machine
set of steps that can be followed to solve learning is to choose the best algorithm
a particular type of a problem. In machine for the task. In a way, in AI development,
learning, algorithms are rules extracted the machine “programs itself” or “trains
from a dataset – these rules comprise itself”117/118 And how do we know what
the model for that data. The model is the best algorithm is? That’s where data
used to make predictions. For example, in comes in. Data is used to compare different
regression, the line is a rule that defines algorithms with each other, and how well
they fit the data or explain the variation and software development that has been going
patterns in the data.118 Machine learning on for decades. Traditionally, software
systems have a cost function that measures engineers have used their human ingenuity
the fit numerically. to come up with a solution and formulate
it as a precise programme that the
So how AI is different from the “traditional computer can execute. Data scientists and
analytics” that existed previously? One way AI developers, on the other hand, don’t try
to think about this is to list the activities to write a program themselves. Instead
needed to achieve an outcome, e.g. sensing, they collect input data and desired target
reasoning, reacting and learning. The first values, then instruct a computer to find
activity, sensing (e.g. computer vision and and select a program that computes an
natural language processing capabilities), output for each input value (see Figure 35
and the fourth activity, learning (e.g. the by Antti Ajanki120).117 Machines learn from
ability to adapt independently without examples, rather than being explicitly
specifically being reprogrammed when programmed for a particular outcome10
conditions change), are of particular Developing machine learning is often a
interest here, because they both more iterative and explorative process
differentiate AI from traditional analytics. than software engineering, and it requires
By contrast, the second and third activity working with a higher level of uncertainty
– reasoning and reacting – are essentially and variability.117/120 Typically, data scientists
the outcome of traditional analytics, e.g. spend 80% of their time understanding and
descriptive, predictive and prescriptive fixing the data. This happens prior to any
analytics. This makes analytics an approach algorithm development or deployment that
which is simple, rule-based and not always accounts for the remaining 20% of time.121 In
very sophisticated.119 data science, quick wins are hard to come
by.
Next, it makes sense to ask how AI
development is different from any typical

Traditional programming
Input
Computation Result
Program

AI development

Input
Computation Program
Desired result

Figure 35: The difference between traditional programming and AI development

 beyond, machines are making calculations
and predictions and telling us that
In their book Prediction Machines,12
authors Ajay Agrawal, Avi Goldfarb and . So, AI
Joshua Gans stress that AI is all about algorithms typically deal with probabilities
generating predictions. As they say, rather than certainties.118 Also, predictions
“ are always based on past data, which is why
they might completely miss potential future
outcomes, or alternatively fail to describe
” The authors them reliably or comprehensively.
explain that various techniques related to
AI, such as regression, neural networks, Figure 36 (original by Agrawal, Goldfarb
deep learning and reinforcement learning & Gans) illustrates the predictive nature
etc. are, in essence, prediction techniques, of AI, by breaking the prediction-making
which can be valuable in certain situations into different elements. In the diagram,
to generate missing information. Whether we use the predictive maintenance of
it’s by processing text, images, sounds, or industrial equipment as an example: the

Judgment

E.g. false positive

/ false negative
consequences

Input Outcome
Prediction Action
data data

E.g. is an equipment E.g. is a technician

E.g. data from the E.g. regular check-
defect likely and what are sent to study the
equipment in the points, equipment
the consequences of such equipment or is a
field user feedback
a defect? software update run
remotely?

Training
Feedback data
data

E.g. is there a pattern?

What can we do better?

Figure 36: Graph illustrating the predictive nature of AI

question is whether to send a technician learning.122 As you can see, there is a
to study the equipment, and whether to plethora of terms, applications and tools,
run a software update or not. Naturally, almost like a “toolbox”, that needs to be
the data input from the equipment used mastered by anyone wishing to become
enables a prediction. The prediction is an expert in this area (and, as this book is
possible because training occurred about about business models, it is useful to note
relationships between different types of that various elements in this toolbox might
data, and which data is best suited to the enable very different kinds of business
task. Combining the prediction with their models).
judgement on what matters, the decision
maker can then choose an action. The It is important to understand that AI-
action leads to an outcome (which has an based systems are often not static; the
associated reward or payoff). The outcome data, data models and algorithms become
is a consequence of the decision. It is outdated quite easily. Therefore, they need
needed to provide a complete picture. The to be continuously trained and monitored
outcome may also provide feedback to help to ensure their performance, while also
improve the next prediction.12 protecting them against bias, safety issues
and privacy violations.113
So, how do machines learn? Typically, in one
of the three ways. In supervised learning, Sophisticated algorithms that are
the human programmer sets defined constantly evolving (in neural networks
outcomes and provides the machine with in particular) are starting to become so
feedback about whether it’s meeting them. complex that even the technicians who
In unsupervised learning, the machine is originally developed them don’t understand
given data and left to look for patterns, how they really work; let alone us normal
e.g. correlations with itself. In this case, people who just enjoy their outputs (that’s
it will discover knowledge which the why the phrase “black box” is widely
human programmer is unaware of. Finally, used). Deep neural networks might have
in reinforcement learning, the machine hundreds of millions of connections, each of
is given “rewards” or “punishments” to which contributes to the ultimate decision
indicate whether what it did was right or made by the algorithm. So, from a human
wrong, and it then learns to self-improve1/15 perspective, the “interpretability” of such
In other words, the role of the human is a system might be low, and the rationale
typically to develop algorithms, provide behind its recommendations are hard to
data for the algorithms, give feedback establish.10 These complex algorithms may
to the computer that is about to figure be “true” and yet no longer understandable.
out something useful, and eventually This surreal situation is a matter of
judge the prediction (or other outcomes) concern, since all human-developed
that the computer has provided for the systems should be able to be diagnosed and
given context and application. Figure 37 fixed by humans in case of failure.
illustrates the concepts related to machine
 However, it’s also possible to argue that
a company can create business value or
other important benefits using “only” its
Data plays at least three different roles data or “only” its AI assets. Therefore, it’s
in making algorithms work. First, there is important for these two elements to be
input data, which is fed to the algorithm understood and dealt with in isolation from
to produce a prediction. Second, there is one another.
training data for the algorithm to use so
that it becomes good enough. Third, there In addition, sometimes it might be easier for
is feedback data that is used to improve the companies to “sell data” or “sell algorithms”
algorithm’s performance using experience.12 without worrying too much about how
Algorithms are fairly useless without data, these assets come together. So, the paired
and so a company must possess and control approach of data and AI-enabled business
the data that it needs to train and operate models, as presented in this book, is more a
its AI.86 This strong interplay explains why point of view rather than a mandatory path
we discuss data and AI as a “paired” enabler for all companies to follow (although, as
of new business models in this book. mentioned before, it is difficult to develop
AI without data from a technical point of
view).

Dimensionality Reduction
Machine Learning
• Feature Elicitation
• Structure Discovery
Regression • Meaningful Compression
• Advertising Popularity • Big Data Visualisation
Supervised Unsupervised
Prediction Learning Learning
• Weather Forecasting
Clustering
• Market Forecasting
• Recommender Systems
• Estimating Life Expectancy
• Targeted Marketing
• Population Growth Prediction
• Customer Segmentation
Classification Reinforcement
• Real-Time Decision Learning

• Robot Navigation
• Game AI • Identity Fraud Detection
• Skill Acquisition • Image Classification
• Learning Tasks • Customer Retention
• Diagnostics

Figure 37: Concepts related to machine learning

Why is AI such a hot topic right now? that made them run. The arrival of cloud
Various authors of AI-related literature computing in the past decade means that
agree that this is the result of three the computing power no longer needs to
factors: be located within the device itself. This
• Computing power has increased has major implications: even small devices
exponentially for many decades in a can now tap into large-scale computing
row, allowing computers to do more resources. And machines don’t need to
complicated calculations far faster than learn from their own separate experiences
in the past. Also cloud computing has alone; instead they can tap into the
become cheaper. learnings of all the machines in a swarm
• The amount of available data has also or fleet, with each machine adding to the
increased exponentially in recent years. collective intelligence of the whole.10/15
So, we now have both large and rich Working with large datasets requires a lot
datasets, and the computing power of processing power, which is why big AI
to analyse them. The role of increased systems often run on supercomputers or
data cannot be underrated. Up until specialised computer architectures.10 Yet,
recently, a lack of data had been a simultaneously, hardware such as sensors
significant obstacle in the development and mobile phones has also become more
of sophisticated analytics. capable, with the capacity to run faster
• While many algorithms currently used (note the term “edge computing”). So,
for AI may be decades old – some AI can be run both from the cloud and
were invented in the 1950s – they have from localised microprocessors capable
recently become more relevant in line of handling sophisticated algorithms:
with the level of data and computing whatever seems most relevant to a specific
power now available. In addition, the user need.
increase in data has not only made
existing algorithms more effective; it In their book The Mathematical
has also supported and accelerated Corporation,123 Josh Sullivan and
the development of new and better Angela Zutavern suggest a high-level
algorithms. categorisation of the technologies needed
to create AI. Their layered approach
The synergistic confluence and co- provides a description of the levels of
operation of these three factors has been abstraction needed when it comes to using
truly revolutionary, allowing researchers to this technology. In their view, it boils down
do completely new things that were simply to five factors:
impossible in an earlier age.10 • Data collection, storage and
preparation.
In the past, it used to be that the computing • Algorithms applied to data to learn new
power fuelling any particular system information.
was physically present in that system: • Algorithms applied to information
the devices contained the processors learned.
• Applications to visualise, interpret and
act.
• Infrastructure technologies to make There are four rules of thumb when it comes
everything possible. to choosing business problems or use cases
that can potentially be solved with AI, as
We are currently undergoing a transition follows:
from handcrafted machine learning models • There is a clear business need for the
to tools that make it possible for ordinary solution that is being developed.
developers to produce them. And once • The task is so complicated that rules
that happens, AI will infuse and change our cannot be written by hand.
entire society in the same way that mass • The prediction task can be defined
manufacturing transformed the 19th and unambiguously (e.g. a metric to
20th centuries.39 optimise).
• There are lots of representative data
available.

Data
Start acquisition and
Formulated
understanding
data science Production
problems pipeline

Datasets

Validation of insights Performant model

Business Modelling Deployment
understanding

Prediction
parameters
End-users
and their needs UX design
for trust
Human-AI
interaction Operation and
optimisation

Figure 38: A data science process model (source: Futurice)

Once these kinds of business problems level artificial intelligence, on the other
have been identified, from the data hand, has turned out to be more difficult
science perspective the work proceeds than expected. Computers seem to be
to data acquisition and understanding. struggling with creating original content,
The solution is then modelled and moves understanding nuances (e.g. detecting
finally to deployment, operation and and decoding emotional subtleties based
optimisation. The result is a service that on social cues), and filtering biases (e.g.
relies on machine learning model(s) whose based on social or ethical consciousness).124
predictions will be used by human end- Some experts estimate it will take several
users. Figure 38 illustrates this data science decades to develop AI technology that
process in more detail. achieves this breadth and depth of human
intelligence, while others contend that it
So far, AI systems have been generally is pure fantasy and will never happen. It is
able to give answers (based on prediction) hard to tell yet which view is right, and the
to questions that people have specifically whole topic seems controversial. But it is
asked them to address. But in the future, plausible that in the coming years AI will
computers may increasingly start to give have advanced to the point where several
answers to questions that humans did not learning models can be pieced together to
even imagine could exist. For instance, form more comprehensive systems. These,
they may be able to find patterns and in turn, will be able to augment many facets
correlations in data that seemingly has no of work environments, as well as everyone’s
relationship whatsoever, when studied via daily lives.54
the lens of standard human capabilities.
This would be a huge new frontier: As the topic of this book is to explore data
whereas now computers perform tasks and AI-enabled business models, our focus
that are designed by human ingenuity is on understanding what new business
and creativity, in the future it may be the opportunities these technologies can
computers that surprise us with their provide. However, AI technologies extend
capability to invent new things beyond our way beyond this: they are also about
comprehension. We should not be afraid of creating resources, tools, expertise and
this, but instead embrace the opportunity capabilities for people who are working
to tap into new growth avenues and make to address issues that benefit humanity.
the most out of them. For the moment, this These include solving tough problems
all lies ahead. in medicine and healthcare, augmenting
human capabilities with technology (hence
While there have been plenty of advances in the term “transhumanism”), space travel
AI development, most of the AI applications and exploration, providing clean and
we use today apply to narrow and specific affordable energy, preventing wars, crime
areas.10/11 These include image recognition, and violence, reducing hunger, poverty and
speech recognition, or matching or inequalities, and so on. AI technologies have
comparing several sets of data. Human- a wide range of impacts and implications
beyond business. These can play out at presents our recommendations for data-
the level of individuals, communities, related ethical issues, with an emphasis on
governments or states. There is a good data and AI-enabled services and revenue
reason to expect that AI systems will not models that are sold to customers.
only grow more powerful; their rate of
development will also accelerate over time.15
Eventually this will become ubiquitous,
hence its amplified impact in various areas
of human life.11
The impact of the arrival of AI on nations’
GDPs has been surprising. According
to Brynjolfsson, Rock & Syverson2, GDP
growth has slowed. GDP growth in the US
was, on average, 2.8% per year in 1995-
Potential ethical problems of using AI 2004, compared with 1.3% per year in
technologies in business relate to: 2005-2018. This is mirrored elsewhere: 29
• Purpose and impact (i.e. who do the of 30 OECD countries saw similar-sized
systems affect and how?). slowdowns after 2004. So, for all the
• Inclusion and fairness (i.e. who is recent developments in AI technologies,
included/excluded and how could bias there is not yet clear evidence that they
manifest?). will contribute to economic growth. As the
• Safety and privacy (i.e. is data collected authors argue, lags and delays are likely
and stored safely, and is user privacy to be the biggest reason for this paradox.
respected with required consent in Or it may be that the most impressive AI
place?). capabilities are not yet widely available,
• Trust and transparency (i.e. are or their full effects won’t be realised until
the decisions made by the systems waves of complementary innovations are
transparent, and do people trust developed and implemented.
them?).
Often, companies aim to apply “ethics by Alternatively, we could also argue that GDP
design” in their AI development, allowing is not an ideal metric by which to measure
them to pre-empt and eliminate potential the impact of AI technologies. According to
ethical problems when it comes to a study by Brynjolfsson and Collins,125 the
deploying it operationally. average American consumer would pay a
staggering $17,000 per year for the right
While the topic of ethics of AI is an
to use search engines (including algorithms
important field, this book only touches
and AI), even though they currently access
on it briefly. There is the vast amount of
them for free. So, we can see that in fully
literature already available for the reader
digital businesses such as these, there is
to study, so it’s unnecessary for us to
plenty of value creation taking place. Yet
repeat the same themes here. Chapter 8.2.
this does not always translate directly
into maximum value capture and revenue many different problems. The Chinese
generation (search engine providers government predicts that by 2030, China
deliberately chose to provide their assets will become the centre of global innovation
for free, allowing them to create and in AI, leading in theory, technology and
capture value within the advertising applications.126/127 But even here, a real
market). And value creation does not always impact on GDP growth due to AI has yet
have a tangible impact on GDP, either.66 So, to be proven. It seems much easier to
the jury is still out on how and where we can find predictions and promises around the
measure AI’s business impact. potential of AI (e.g. as presented at the
World Economic Forum128) rather than hard
So far, we have explored the application evidence.129
of AI technologies mainly in businesses
located in OECD countries. But it’s
important to address the role of China as an
AI superpower, too. China is ramping up AI
investment, research and entrepreneurship
on a historic scale. This effort is driven
by several factors including venture
capitalists and other investors pouring
in money, a supportive governmental
policy environment, relevant high-quality
education being available, and finally rich
and abundant data for companies to use;
for example, emanating from their large
home market and in the population of the
country overall. Chinese entrepreneurs
seem to be especially talented in
applying breakthrough AI technologies
to various practical applications, and to
Appendix 3: Useful business
model thoughts and
frameworks
A) Alternative business Typically, the value creation executed by
a company is seen as a system, in which
various components come together and
considerations have many interdependencies.75

When discussing business models in Tommi Rissanen’s doctoral dissertation 20 is

depth, it is good to start by stating that insightful , as it presents different business
“business value” and “economic value” model definitions used by various authors
are multifaceted and complex concepts. from different academic backgrounds,
But they can be boiled down to a few key including strategy, entrepreneurship and
elements: growth, profitability, innovation, innovation management. These definitions
return-on-investment, commercial models, are distinguished by their scope: some
pricing and company reputation. In the focus purely on the economics of a
real world, this picture quickly gets more business, while others aim to cover nearly
complex, as each element contains dozens all aspects of business, strategy and
of different levers and drivers. And organisation.9
management decisions and actions tend to
have an impact on each and every one.8

Stage 1 Stage 2
Strategy stage Tactics stage

Firm chooses the business model Tactical choices are made. The tactical
through which it would like to choices that are available depend on
compete. the business model chosen by the firm
in the first stage.

Figure 39: A framework for strategy, business models and tactical choices
To establish a distinction between frequently enough. It is rare that one
strategies and business models, Joan hears a salesperson, for example, starting
Magretta62 states: “A business model is a their presentation by saying: “Here are
description of how your business runs, but a the nine boxes of our company’s Business Model
competitive strategy explains how you will do Canvas”, or in any way explicitly stating
better than your rivals.” To integrate strategy, the various elements of a business model.
business models, and various tactical Instead, companies tend to talk about
choices that a company might have, their strategies, products, services and
Casadesus-Masadel and Ricart130 provide value propositions. Sentences such as
a two-stage framework as illustrated in
Figure 39. Their point is that to begin with, great products”, “We help you to...”, and “We are
companies choose value creation and value on a mission to...” are common in business
capture logic on a strategic level (i.e. a language, whereas business models are
business model). Then they make tactical seldom mentioned. This makes the concept
choices guided by their goals in executing seem somehow remote or hidden.
their selected model for maximum value.
In a way, business models are both artificial
Business models can be approached and practical at the same time. Artificial
from at least from three perspectives: 1) in a sense that they are illustrations of
explaining the business (understanding how companies or ventures that are rarely
a certain company works and makes profit), described as such in company speeches,
2) running the business (explaining how a websites or brochures. Yet also practical
certain company stays successful or afloat because it’s clear that all successful
in relation to its employees, management companies have a value proposition, a
and shareholders), or 3) developing the revenue logic and a cost structure and
business (understanding how a certain other business model elements in place:
company identifies new opportunities whether or not they want to reveal them.
and creates sustainable competitive
advantage).131 This approach attempts to If you want to find out more about a
cover both the existing business models company’s new business models, the
a company has, and the innovative new best source are the press releases that it
models they may be developing, as well as sends out. These typically contain details
any transition period that comes into play of a company’s new products or services,
when shifting from one model to another new pricing models or new partnerships.
over time. So, they at least provide a hint about how
they hope to generate new value for their
customers in the future.

If the concept of a business model sounds

a bit theoretical, we believe it’s because
they are not discussed appropriately or
 Customer
value
proposition

Key Key
Profit formula
resources processes

Figure 40: A business model by Johnson

B) Alternative business The purpose of this breakdown is clear.

model frameworks Whereas the Business Model Canvas
explored earlier in this book lacks metrics
Building on the traditional Business Model for business model execution, the IDEO
Canvas (as in Chapter 3.2), the design model is grounded in them (for instance,
agency IDEO106 breaks down business how much value is created and how many
ventures into three major areas of risks customers are adopting the idea). This
and uncertainty that need to be tested, allows companies to understand exactly
experimented and prototyped in order to how a new innovation is performing in
find an ideal business model configuration: relation to the these three key areas.
• Desirability: Reducing customer-
related uncertainty by testing In Figure 40, Johnson9 presents another
assumptions around the “problem” business model illustration, which is a more
(jobs, pains, gains), solution (value simplified version of the Business Model
creation), customer acquisition and Canvas approach, with only four boxes.
customer retention strategy.
• Viability: Uncovering the financial In Johnson’s view, these four elements are
opportunity and reducing the financial interdependent and constitute the basic
risk by testing assumptions around architecture underlying all successful
the financial aspects of the venture businesses. The idea is to devise a relatively
(revenue, pricing, and cost). stable system in which all four interact
• Feasibility: Reducing the in consistent and complementary ways.
implementation risk by testing A change to any one of the four elements
available or potential technology should affect all the others, and the entire
options, as well as resources, activities system. And, if there are conflicts among
and partners. the four elements, the execution of the
business model carries a higher risk of 2. Which new activities are needed
failure. to satisfy these perceived needs?
(business model content innovation).
Johnson9 also stresses the job to be done 3. How could the required activities
in the value proposition: it lies within a be linked to each other in new ways
firm’s ability to solve important problems (business model structure innovation)?
for its customers via a product or service. 4. Who performs the activities that are
In his view, a job to be done goes beyond part of the business model? Should it be
merely “listening to the customer”. It the company? A partner? The customer?
extends more deeply and reveals what a What new governance arrangements
company is trying to get done in the given could enable this structure? (business
circumstances, whatever the articulated model governance innovation).
“needs” are and whatever the existing 5. How does the new business model
products and solutions. Typically, jobs create value for each of its participants?
to be done are related to removing 6. What revenue model fits with the
barriers such as insufficient wealth (e.g. company’s business model, to
providing cheaper products or solutions appropriate part of the total value it
that customers can afford, or providing helps to create?
high-end products and solutions that
wealthy customers prefer over standard In the book Reinvent Your Business Model,9
ones); access (bringing the products or author Mark W. Johnson presents various
services closer to customers, or serving “levers” that are related to business model
them in a more convenient way); skill innovation, both in products/services
(providing simpler products or solutions and revenue models. Figure 41 (originals
that customers know how to use); or Johnson9, modified by the Mika Ruokonen,
time (providing the products or solutions see also Sinfield et al55) demonstrate some
faster).72 potential levers (or trade-offs if you wish)
that a company can use to vary its business
model and stand out from those used by
competitors.
C) Alternative business
In his book Seizing The White Space,18 Mark
model innovation W. Johnson stresses the importance of
frameworks finding empty space in the market for new
business model innovations. For example,
In order to build new innovative business this could be by fulfilling important but
model configurations (as discussed in unsatisfied jobs to be done for existing
Chapter 3.3), a company needs to ask the customers in existing markets (“within”),
following six questions: 55/75 for completely new customers (“beyond”),
1. What perceived needs can be satisfied or in the new terrain that emerges as a
via the new model design? result of a transformative market shift
 Offering levers

Product Service
TYPE

Focused set of Extensive set of

features features
FEATURES

Measurable or
concrete product or Intangible attributes
service attribute
FORM

No ability to be Can easily be customised for

customised individual needs or prefer-
CUSTOM ences

Single product/ Multiple products/services

service delivered delivered to customer simul-
to customer taneously
BREADTH

Consumed multiple
Consumed once
times
LIFETIME

Revenue model levers

The price of the The price of the prod-

product/service uct/service offering
offering is fixed can vary (e.g. auction)
PRICE

Customer pays
Customer pays for proportionally for the
each unit value extracted
UNIT

Customer makes
Customer makes one
more, smaller
complete payment
FREQUENCY payments

Customer pays before Customer pays after

receiving the product/ receiving the product/
service offering TIMING service offering

Figure 41: Product/service and revenue model levers related to business model innovation
(“between”). Innovations in the context of “basic” Business Model Canvas literature
a business model only count as such if they that was reviewed in Chapter 3 and in
uncover business opportunities that are Appendix 3 is useful as a starting point,
relatively untapped. Johnson’s contribution but it doesn’t cater to the needs of data
considers the role of marketplace and AI as technologies. The number of
competition, whereas the Business Model books or articles available on this topic is
Canvases do not. limited.132 In addition, the literature that
does exist doesn’t explore the full scope of
Finally, there are various ways of generating the issue. Authors tend to tackle “obvious”
growth and making money when innovating evolutionary business models that resemble
new business models. New business model traditional existing businesses, rather than
innovation could enable a company to: look at more radical approaches. It may also
• Generate more revenue from the be that some of these revolutionary models
current business (e.g. via an ability have not yet been invented. This clearly
to sell more of its existing products/ represents a gap, but at the same time it’s
services). also an opportunity.
• Generate completely new revenues (e.g.
through the launch of new products/ One existing and rather good approach
services or growth ventures). to data and AI-enabled business model
• Receive a steadier source of revenue renewal belongs to authors Sharad Gandhi
(e.g. by moving to recurring revenue and Christian Ehl.11 They turn traditional
models). business model thinking (e.g. based
• Receive revenue over a longer period of on Osterwalder and Pigneur) into the
time (e.g. by selling continuous services “language of AI”, describing the anticipated
instead of products). impact of AI technologies in each of the
boxes (see Figure 42).
All business model innovations could be
evaluated from this perspective: what The labels and terminology in this model
changes in revenue generation would differ from traditional business model
implementing them enable? approaches, but there is still some
resemblance to them. However, Gandhi and
Ehl’s approach is still very much focused
on AI benefits, rather than 1) what exactly
D) Scarcity of existing is being offered to customers using data
and AI, and 2) how the product/service is
frameworks for data and monetised, i.e. how exactly revenues are
AI-enabled business models generated, and where the money comes
from. As these two points represent the
Unfortunately, when it comes to data and focus of this book, Gandhi and Ehl’s model
AI-enabled business model renewal, there doesn’t quite satisfy the need for more
is not much literature to draw on. The answers.
Figure 42: Gandhi and Ehl’s11 approach to data and AI-enabled business models
 Replace
materials with
information

Services on Magical user

demand experience

Marketplace Reputation
liquidity Networked systems
marketplace
platforms

Managed by Augmented
algorithm Helping workers workers
to succeed

Alternatives to
full-time Talent and
employment resources on
demand

Figure 43: A business model map for the next economy86

Another useful approach to data and services on-demand to customers through

AI-enabled business models is presented a magical user experience. At the same
by Tim O’Reilly in his brilliantly named time, they also use data and intelligence
book WTF? What’s the Future and Why to augment work performed by employees
It’s Up to Us39. Based on his examination and partners. The mode of operations in
of new digital companies such as Uber, these companies typically resembles a
Airbnb and Lyft, the author identifies a marketplace or platform. And instead of
common “business model map for the trading physical assets (e.g. cars, houses)
next economy”, as shown in Figure 43 (this these companies tend to trade information
model also resembles the popular platform (e.g. what assets are available near a
business model approach e.g. Moazed91 and desired location at any given moment).
Zhu and Iansiti92). The essence of this business model map
is that everything is enabled by data and
The operations of these new digital algorithms. In fact, the entire company
companies are typically managed by might be built around the algorithm that
algorithms that steer talent, resources and matches supply and demand. With a little
 Data sources • Existing data • Crowd-sourced
Internal
• Self-generated data • Tracked, generated and other

External • Acquired data • Open data

• Customer-provided • Social media data
• Freely available • Web crawled data

Key activity • Data generation • Crawling

• Data acquisition • Tracking and crowd-sourcing
• Processing
• Aggregation • Descriptive
• Analytics • Predictive
• Visualisation • Prescriptive
• Distribution

• Data
Offering
• Information/knowledge
• Non-data product/service

Revenue model • Asset sale

• Lending/renting/leasing
• Licensing
• Usage fee
• Subscription fee
• Advertising

Target customer • B2B

• B2C

Specific cost
advantage
Focus areas in this book

Figure 44: A taxonomy of data-enabled business models48

bit of effort and imagination, it’s possible to Hartmann, Zaki, Fedmann & Neely132
detect familiar elements of Osterwalder and developed a taxonomy for data-enabled
Pigneur’s classic business model19 in Figure business models, shown in Figure 44. This
43, including elements such as customer model was originally built for start-up firms
relationships, channels, resources, to use and was tested by 100 of them. But it
activities and partnerships. Yet revenue can also serve as inspiration and a blueprint
models and tangible products/services are for larger firms. The framework presents
still non-existent. the dimensions, attributes and features
behind forming a comprehensive business
One area of increasing interest for model for a data-enabled venture. It’s
researchers and executives, while still useful in identifying and assessing available
in its early stages, is data monetisation potential data sources, with a carefully
or data commercialisation. Both terms considered sequence of actions that show
are currently used, and there seems to how to turn data into a sellable asset. As
be very little difference in their meaning. the authors explain, a business model that
Data monetisation is commonly seen as builds on data can have more than one
an opportunity for a company to open feature in each dimension, too. For instance,
new avenues of growth,28 using data for there might be various data sources that
new products and other innovations,32 and perform different activities.
to put existing data assets into use for
revenue generation.27 Lewis and McKone The model looks quite like a “decision tree”
consider data as a valuable by-product or a “pick and mix” framework that helps
in many companies, and something business leaders to identify different
that can help them expand into new options in a business model selection. In
businesses that are on the periphery that sense, it is like a “business model
of their existing core operations. The canvas for data-enabled ventures”, even
assumption is that if data is not monetised, though it looks a bit different from a canvas.
companies are leaving money on the table. It also proposes “pre-cooked” answers and
A McKinsey study28 suggests that data alternatives which canvases typically don’t.
monetisation is particularly evident in
the energy, financial services and high- Evaluating the model of Hartmann et al132
tech sectors. It is commonly used as a from the perspective of this book, the
source of differentiation, and it seems to obvious disadvantage is that it doesn’t
correlate with industry-leading business cater for AI technologies in any way. So,
performance. On the other hand, the the product/services dimension is rather
McKinsey study also identifies some narrow (as we explored in Chapter 5, there
common challenges in data monetisation, are more options available). And even
including lack of data strategy, failure to though the revenue model options are well
attract and retain data analytics talent, lack presented, this taxonomy ignores popular
of financial resources and a failure to get alternatives such as free/freemium or
real commitment from business leaders. project revenue options (again, Chapter
5 sheds more light on various options
here). So, while this model is useful as a
foundation, it could benefit from further
development and analysis.
Appendix 4: Discussion
themes in the co-creation
sessions
A) Your current situation d. Other new innovative approaches?
e. What is the customer experience
1. Please describe your role in the company that you wish to deliver?
and your involvement with data and AI 7. What opportunities, challenges or
so far. boundaries does your industry bring to
2. What are your real problems worth data and AI-enabled business models?
solving when it comes to renewing 8. What do you think is the maturity of your
business models using data and AI? And firm in relation to applying data and AI
the real problems worth solving for your to business model renewal?
customers/buyers? 9. Which initiatives have you started in this
3. What is your vision and ambition for area already?
data and AI-enabled business models? 10. How have you taken your new data
Incremental improvement? Radically new and AI-enabled business models to the
ideas? What do you expect to happen? market? Any launch tactics to learn
Where do you get your inspiration for from?
the business model renewal?
4. What is the link between data and
AI-enabled business models and your
strategic agenda? Are they a priority? B) Co-creating solutions
Or a side-track?
5. What is your appetite for exploring 1. We sent you a handout showing our
completely new business models and thinking beforehand – what were your
data and AI products/services? How key takeaways from it? Did the points
does this compare to your focus on mentioned there make sense?
current business? 2. We have come up with the “playbook”
6. What data and AI-related business for data and AI-enabled business models
models: 1) do you have, or 2) plan to (products/services vs. revenue models
implement, or 3) have seen elsewhere? shown to the people in the room). What
a. New products/services? do you think of it: is it useful to your
b. New ways of making money? company?
c. New industry segments?
 a. Product-oriented models (incl.
reports, platforms, assets): what
C) Closing
works, what doesn’t? Are they
relevant to you? 1. How did the canvases work in this
b. Service-oriented models (incl. session? How can we make them better?
outcomes, capacity): what works, 2. What is your feedback on this session?
what doesn’t? Are they relevant to Was it relevant, useful and helpful?
you? 3. Are there any useful books, articles or
3. Do you have what it takes to explore and blogs that you would recommend? Are
implement the new business models? there any existing frameworks for data
E.g. capabilities, entrepreneurial and AI-enabled business models that
mindset, curiosity, data and AI assets in inspire you? Or companies to follow? Or
place, etc.? people to talk to next?
4. What partners do you need to implement 4. Any final words? Did we forget to
data and AI-enabled business models? discuss anything important?
Or can you do it all by yourself?
5. When you apply the new models, what
change is needed in your organisation
to make it happen? E.g. competence
development, re-allocation of resources
etc.?
6. Should the new business models be run
within your existing organisation or as a
separate unit?
7. Is there something that is preventing
you from innovating new business
models based on data and AI?
8. What are the (anticipated) financial
outcomes of the new models? Size of
the price? Or revenues at stake? Do
you already have evidence that there
will be a financial impact as a result of
implementing new data and AI-enabled
models?
9. What do you think is the sensible path
forward in your company when it comes
to data and AI-enabled business models?
E.g. strategic level, operational level,
team level etc.?
Appendix 5: Canvases used
in the co-creation sessions
chain? Or perhaps they are players based
Business Goals & Drivers outside your industry segment?

This canvas is useful for companies that

are still in the beginning stages of planning THE IMPACT YOU AIM FOR
data and AI-enabled business models. It What is your initial understanding of
helps businesses to define their ambitions, the behavioural change that you want to
and the impact they aim to achieve with facilitate in your customer group? For
business models. It also identifies potential example, do you want to help them do
problems worth solving at a high level, along something smarter, faster, cheaper or more
with the data and AI assets available for efficiently; or maybe it’s about helping them
commercialisation. to avoid doing something? How could your
data and AI enable a change in society in
YOUR AMBITION general? Do you already have a method in
Which changes do you see in the mind for measuring the desired impact?
marketplace, and how do these shape your
ambition with regards to data and AI-enabled PROBLEMS THAT ARE WORTH SOLVING
business models? Do you want to boost your What is your initial understanding of
existing business models and products/ the problems worth solving among your
services with data and AI? Or do you aim to customer groups that could be tackled
start exploring areas beyond your current using data and AI? How might solving these
business, e.g. new markets or customer problems take you closer to your data and AI
segments or new products/services? Or do vision?
you wish to use data and AI assets to tackle
an entirely new industry segment? Note that TECHNICAL ENABLERS
you can list several potential ambitions that What is your current data and AI-related
your team might have here. technology infrastructure (e.g. cloud
technologies, data lakes, data pipelines,
YOUR CUSTOMER GROUPS IN FOCUS interfaces, analytics tools etc.)? Do these
What is your preliminary understanding of enable you to reach your ambition? Do they
the customers that there might be for your enable you to quickly experiment with new
data and AI assets? Are they your current data and AI-enabled business models? What
customers? Or perhaps some other parties factors are missing that need to be in place
e.g. suppliers or partners in your value to progress?
DATA AND AI ASSETS you think are the motivations behind your
What data is in place? What algorithms have customer group’s current behaviours and
you already developed? Which of these does mindsets? How can you use data and AI
your company possess and own? Which of to have an impact on those motivations?
these could you acquire from partners or Finally, is there also an opportunity to
third parties? What are the limitations of change the behaviours and mindsets of
these assets? How could you make your your customer group in a way that brings
data and AI assets stronger overall, to about positive societal impact?
enable you to reach your vision?
RELATED NEED WORTH SOLVING
What is your understanding of the needs/
problems worth solving among your
Customer Groups customer group? What is it that they cannot
yet do? Or that they could do better? Are
This canvas is useful for exploring potential these problems of strategic importance, or
customer groups for data and AI-enabled are they critical to daily life? What is your
products/services and revenue models. You understanding of the root causes for these
can add in several customer groups, along problems? How are these problems being
with describing their required behavioural solved currently, and how could they be
and/or mindset changes and anticipated solved better in the future using data and
needs. AI?

GROUP NAME & DESCRIPTION CUSTOMER GROUP PATTERNS

Here you can simply identify the anticipated Are there certain patterns that are common
customer group for your data and AI assets. to all customer groups listed (e.g. similar
These may be your current customers (e.g. characteristics, similar desired behavioural
a particular customer segment) or your or mindset changes or similar problems
suppliers, partners or other similar parties worth solving among all customer groups)?
who would benefit from your data and AI If yes, please put them here. These common
assets. factors may become good design drivers
for the data and AI product/service and
DESIRED BEHAVIOURAL CHANGE revenue models that we create later, e.g.
What is the behavioural or mindset change factors that always need to be in place.
that you want to enable in your chosen
customer group using data and AI? How can
you help them to do something smarter,
faster, cheaper or more efficiently, or
avoid doing something altogether? Or, how
can you help them to think about things
differently, or understand things that they
have not previously understood? What do
Impact & Implementation bigger/smaller gaps than others? Again,
try to analyse each business model option
Optimiser separately.

This canvas helps companies to prioritise Finally, you and your team should decide
and choose their potential data and AI- where to place your various data and AI-
enabled business model options. There are enabled business model options on the
two axes: the size of the desired impact (e.g. canvas. Prioritise the versions that can sit
potential revenues, societal impact) and the easily in the top left-hand corner of the
difficulty of the problem to be solved (e.g. canvas: i.e. those that have a large potential
anticipated effort and costs). The sweet impact and are easy to implement. Then
spot is in the top left-hand corner when the work your way down with the options that
anticipated impact is large, and the problem have less impact or are more difficult to
is likely to be solved easily. implement. Decide on the model options that
you want to start implementing right away
MEASURE THE IMPACT OF THE PROBLEM versus those that can wait or won’t be used.
YOU ARE SOLVING
Consider the problems that you are about
to solve using data and AI. Also think about
the desired impact that this will have e.g. on Data and AI-enabled
your customer group’s behaviour or thinking,
or on society overall. Can you estimate the
Business Models
size of the impact your anticipated data and
AI-enabled business models will have? For The canvases outlined so far are designed
instance, how many people are affected, how to help companies structure their thoughts
profoundly will their behaviour change, how and prioritise initiatives. This last canvas
much money could be made, or how much goes even further, helping firms to describe
better will society become as a result of the their data and AI-enabled products/
models? Try to analyse each business model services and revenue models in greater
option separately. detail. This is achieved by identifying
potential value propositions, products/
HOW DIFFICULT IS IT TO SOLVE THIS services and revenue models for each
PROBLEM? customer group.
Think about your current data and AI
maturity, your current data and AI assets, CUSTOMER GROUP
and your ambition to solve problems and As in the Customer Groups canvas, here you
make an impact. How large is the gap can simply list the anticipated customer
between what you have and what you want group for your data and AI assets. They can
to have? How much time, effort and money be your current customers (e.g. a particular
will it take to build each of your business customer segment), or suppliers, partners
model options? Which options contain
or other similar parties who may benefit VALUE PROPOSITION
from your data and AI assets. How will you create value from data and AI
aimed at this customer group? For example,
what problems are you about to solve?
How can you help customers do something
differently? How will you bring about a
change in mindset, or facilitate better
decisions? This is an opportunity to create
and discuss a variety of options for data
and AI-related value propositions.

PRODUCTS/SERVICES
Given your chosen customer group and the
value proposition that you want to build,
what product or service do you need to
create? What format should your data and
AI assets appear in, ideally? Will you be
selling raw data, or analyses or reports?
Or perhaps you’re looking at the possibility
of a platform or service? Go through the
playbook of data and AI-enabled offerings
presented in this book and weigh each
option up against your customer group’s
specific needs. This will lead you to a group
of potential products/services that you
can discuss and validate with your target
customers at a later point.

REVENUE MODEL
Finally, how will your data and AI-enabled
product or service capture value from your
chosen customer group? If you’ve already
chosen your product/service at this point,
then you can tie your revenue model to it.
Again, check the playbook for likely models
and options worth testing. Play around with
thoughts and ideas in your team before
choosing a suitable revenue model that
you could discuss and validate with your
customers.
Appendix 6: Seven company
case studies
Case study #1: An product sales towards the sale of solutions
and recurring subscriptions.
equipment manufacturing
company (Company A) Company A has already implemented
sensor technologies that collect data
Key themes: Manufacturing, equipment, from its equipment and pushes it to the
data insurance model, human error cloud. This has allowed it to start selling
elimination model, digital strategy its equipment as a swarm and ecosystem.
extension, sales transformation, So, Company A has a great opportunity
commercial capability improvement, to help its customers optimise their
customer education. use of equipment based on its data and
algorithms. Various reports are already
Company description: Company A has being generated, either for Company A’s
over 50 years of global operations to internal use, or for customers to use in their
its name, along with 700+ employees. daily work.
It manufactures equipment for use
by other companies in their industrial Customers generally feel confident about
production facilities. We met three people giving their data to Company A to be
from Company A, including the Chief analysed, as they see the benefits that
Digital Officer and two of his colleagues come with continuously improved products
representing the business units. or services as a result. Another factor in
the customers’ willingness to give data
Company A has worked on its digitalisation to Company A is its reputation as a long-
for a few years and has already made a lot standing reliable and responsible party in
of progress. However, it aims to scale up its the industry.
digital operations further in coming years.
Company A already has a customer portal
Digitalisation, harnessing of data and in place that serves as the backbone of
analytics and building automation are all its scalable digital offering. The portal
at the core of Company A’s strategy. Its contains several important features such
vision for the future is to increasingly move as customers’ equipment information,
towards production of fully automated user manuals, spare part information, and
equipment, selling outcomes instead of ordering and tracking of maintenance
equipment, and moving from one-off work or product updates. Other features
can also be added. The portal is offered was based on a parallel customer expectation
free of charge to customers, apart from for cutting-edge digital products.
one paid-for feature. Customer-specific
data and insights on equipment usage and Company A found it easy to describe its new
performance are available for a recurring product/service ideas technically, e.g. how
monthly fee. The portal was developed in data and AI would be used, and what the
close collaboration with two early-adopter production outcome would be. But setting
customers and selected external partners. up an equivalent revenue model was more
of a challenging topic, especially in terms
Opportunities and challenges: Company A of how to handle a gradual shift from its
identified several new data and AI-enabled current model (one-off sales) to its chosen
opportunities. Our discussion focused concept (monthly licensing), given most of its
on two of them. We tried to clarify the customers were unfamiliar with the licensing
emerging concepts and think of ways to model.
productise them further. We also considered
setting revenue models for them. The We concluded that the monthly licensing
concepts were related to: model wasn’t the only option here: Company
• Providing customers with data on their A could also consider other revenue models
own equipment when in use and even such as selling fixed fee data or tying
guaranteeing continuous availability of prediction results to a set of KPIs agreed
such data (for instance, to meet industry between Company A and its customers.
regulatory requirements). This concept
would operate as a “data insurance firm”. Much of our discussion focused on whether
• Using data and intelligence to to serve Company A’s data and AI-enabled
predict the quality of the products concepts as extensions within its existing
manufactured by customers with customer portal, or as standalone products.
Company A’s equipment. This would The consensus was that extending
eliminate human error and variance in the portal made sense since it offered
manufacturing, and help customers customers a seamless, comprehensive
ensure that they deliver high-quality and continuously evolving data and AI
end products to the market. experience.

Both concepts capture the core competence Another topic that we discussed was the
of Company A (i.e. a deep understanding of format in which the data and AI products
the manufacturing process in question). Both would be delivered to customers. As
are also based on unique customer-specific Company A’s Chief Digital Officer put it:
data that is generated as a result of the
manufacturing process. Company A saw its “Some customers would like to see their equipment
newest and most advanced equipment as data in their ERP systems. But so far we have tried
offering the best commercial opportunity for to keep the data in our own portal and keep the
new data and AI-enabled development. This
API building to the minimum because it requires so similar transformation journeys in more
detail in Chapter 7.7). In particular, it
may need to enable major changes in
We also discussed how to identify, prove its sales processes, to ensure that its
and deliver value to customers with data team there possesses the incentives and
and AI. Company A already had systems capabilities needed to sell new products
in place for this with its early-adopter under new revenue models (currently its
customers, but we identified room for salespeople are accustomed to selling
improvement. It’s a critical point in order “equipment per piece”).
to justify Company A’s new revenue model • If Company A could point to external
(monthly licensing). If Company A’s data examples of successful transformation
and AI products were to be scaled up, its in a similar equipment manufacturing
customers would need to be reassured setting, it would be easier for it to
that they’d get value for their money as a raise awareness and justify changes
result. This approach also has an impact on internally.
the future P&L of the company: if certain • Company A has some work to do in
revenues can be assumed, it is easier to terms of educating its customers on
calculate which investments make sense. the value of new business models, too.
Luckily, some customers had already
Key learnings: started to realise this, creating an
• The new data and AI-enabled business expectation for innovative new models
models identified by Company A seem that Company A can build on.
like a strong and effective way of • Company A is well positioned when it
extending its current digital strategy comes to technology. But in order to
(see also Chapter 1.6, in which we make its data and AI ambitions a reality,
discuss the relevance of synchronising its biggest areas of improvement lie in
between a company’s existing and the commercial side of the business.
future digitalisation). Challenges include sales skills,
• However, a shift to these new data commercialisation and productisation,
and AI-enabled offerings and revenue value creation and capture, and
models is also a complex process for systematic change management.
Company A, requiring lots of persistence
on behalf of key players involved. As
Company A’s Chief Digital Officer noted:
“We in this room love this new digital Case study #2: An online
technology and these new solutions but it
is tough to get others inspired by them, too,
(Company B)
customers who should be adopting them.”
• A comprehensive transformation awaits Key themes: Digital media, selling data, new
Company A going forward (we explore revenue opportunities, new complementary
business model, development agenda, Company B already provides several
transformation story, scaling up. commercial data services to its customers,
e.g. to real estate agents, construction
Company description: We held a co- companies, financial institutions, recruiters
creation session with five representatives and news agencies. More of these groups
from the digital unit of this media company. are proactively identified, and the company
The unit develops online classified ads also benefits from continuous inbound data
marketplaces for several sectors such as requests. However, these buying parties are
cars, jobs and real estate, and employs a all other companies rather than consumers.
100+ people. The nature of the business is
fully digital. Company B’s customers are Company B applies a monthly subscription
both consumers and other companies. fee for data sales, and customers can select
between a data dump or continuous daily
Company B (meaning the unit of a large or monthly data delivery – in all cases,
firm, as previously outlined) is advanced the data is provided in CSV or JSON files.
in its data-enabled operations. It has large Revenue share models are also used in
repositories of data from its marketplaces, some transactions, i.e. Company B gets
including both supply and demand of item a share of the value that the customer
that are being traded, and sold-for prices generates via purchased data. However,
and sales time frames from various user Company B’s customers tend to prefer a
transactions. In addition, Company B has fixed monthly fee as it offers predictability
built correlations of supply and demand and and stability. There is already evidence
completed transactions in order to identify that these customers have been able to
patterns and anomalies in trade, allowing it make smarter decisions and improve their
to deliver interesting insights into e.g. local operations using Company B’s data.
housing and job markets.
While Company B works to build
Company B sees data-enabled operations dashboards and enrich its data prior to
as a strategic priority and an area for sale, many customers ask to receive it
rapid revenue growth. Its traditional raw, in order to visualise and enrich it
online classifieds advertising business themselves. An exception is a payable data
model (selling online ads) faces continuous dashboard that Company B has developed
pressure and selling data would temper that builds on location information and
this risk in a way that complements the reveals insights on supply and demand of
model. Selling data also helps Company B objects in a given local market.
to stand out from the competition, leverage
its position in the value chain, reach new Buying data from Company B requires a
customers and get closer to its existing certain level of maturity on the customers’
customers. side, e.g. data science competence, so
that the customers can derive benefit
accordingly.
In order to reduce risk, Company B only Company B faces challenges in getting its
sells data to trusted parties with a good entire sales force onboard with selling data.
reputation and does not allow purchased For traditional media sales teams, data
data to be forwarded to third parties. sales means addressing a completely new
group of customers, with new products,
Opportunities and challenges: While new sales models and new revenue models.
selling data is a well-established practice The overall shift from ads product sales
and a strategic priority for Company B, towards consultative sales of data solutions
our co-creation session identified several is significant, and certainly not easy.
opportunities for further development.
Company B’s customers also lack
So far, Company B’s data services have understanding of data value to a degree:
focused on low-hanging fruit: what it can do they are not always clear how data can help
with existing resources, and in response to them to improve their operations or decision-
proven and immediate customer demand. making. This might be an opportunity for
Going forward, there is room to develop a Company B to build a new consulting arm in
more comprehensive and structured data which it helps customers to use data more
agenda, including prioritisation of new data effectively. Alternatively, its data products
products and refinement of existing ones. could be refined and simplified so that no
specific skills are needed at the customer
Our session together identified an end to reap value. Either way, Company
opportunity for Company B to meet various B could benefit from more storytelling
customer needs by building more granular to define the multiple ways in which data
and sophisticated analyses, and better data can be used to extract organisational and
presentations. Also, there is an opportunity operational value for its customers.
for it to deliver insights not only on what has
happened or what is currently happening, Key learnings:
but what will likely happen in the future • Even a data-literate business such as
(e.g. trends in items being traded). These Company B could benefit from building
additions could strengthen Company B’s a comprehensive transformation story
data services and increase its added value. for its data and AI-enabled business
models (see Chapter 7.3 for more on this
We also saw an opportunity to fine-tune approach).
Company B’s revenue models and pricing • A transformation story can a) help a
to cater to the demands of potential company to justify and prioritise new
customers. It seemed to us that this process business development work, b) better
may be subject to constant iteration by engage its sales team in new business
Company B, together with its various models from the outset and c) encourage
customer groups. customers to start thinking about the
value that these models can bring.
• Above all, a comprehensive for its operations. It currently uses a
transformation story can help an freemium revenue model: both B2C and B2B
organisation like Company B scale up its customers are given basic features for free
data and AI-enabled business models. and pay extra to receive premium services
This would allow it to translate its great online.
early results into something even more
impressive and enduring. At the time of our co-creation session,
Company C faced an important strategic
choice that would set the guidelines for its
growth, organisational setup and company
Case study #3: A digital culture for years to come. The decision
involved three alternatives:
platform provider company 1. Company C focuses on aggressively
(Company C) growing its international user base,
giving its core platform service for free,
Key themes: Digital platform, startup, but implementing data and algorithms
selling data, strategic alternatives, business monetisation (the business model
model overturn, uniqueness of data, overturn) as part of a future revenue
customer validation, willingness to pay. model.
2. Company C develops its existing digital
Company description: Company C is platform and current revenue model
a startup with operations in multiple (freemium) along with some additional
countries across two continents. It data-enabled revenue models on the
employs 30+ people in three locations. side.
Company C develops and runs a digital 3. Company C builds new face-to-face
platform on which consumers (B2C) and consulting services on top of its existing
companies (B2B) carry out transactions. digital platform, in order to complement
Company C’s CEO and Head of Engineering its current business model and open
joined our co-creation session. new sources of revenue through
professional services delivery.
Company C is already quite advanced
when it comes to its software, data and Option three was the least likely from the
algorithms. It has already experimented outset. So, the decision really came down to
with various business model options. There how much Company C wanted to focus on
is rich and valuable data on both the supply business model overturn and data-enabled
and demand end of transactions, as well revenues (the more radical option 1) versus
as differences as a result of the platform its existing online platform and freemium
being used by both B2C and B2B users. revenue model (the more conservative
Company C is still trying to find an optimal option 2).
product-market fit that would deliver the
best growth avenues and profitability
This strategic decision will impact Company and 2) who are the customers with the
C’s valuation once it approaches a new most incentive to pay for Company C’s data
financing round in coming years. At that and so represent the strongest revenue
point, will it be perceived as a data and AI potential?
commercialising firm, or as a transaction
handling firm? In terms of the quality of its consumer
market data (B2C), Company C compared
It was evident that Company C could run itself against global giants such as Google,
both options 1 and 2 in parallel for the next Facebook and LinkedIn. Commercialisation
six to 12 months. But eventually a clear of “just any data” wouldn’t do. Instead,
strategic choice seems inevitable, in order it needed a model that would create
to avoid splitting its scarce resource across outstanding value to very well-chosen
multiple businesses. target groups; an offering that large giants
would not be capable of matching.
Opportunities and challenges: Company
C had already identified several potential From a privacy point of view, it was
customer groups (e.g. consumers, evident that data on individual consumers
companies, institutions, governmental generated in the platform could not be
organisations) for its new data-enabled shared with customers directly. Instead, it
services, as well as innovating several would have to be aggregated and refined
potential format options (e.g. data prior to sale. On the other hand, the
dumps, reports, dashboards). Our co- consumers in Company C’s platform could
creation session focused on mapping and be considered data customers, too. But
consolidating various customer groups and again, they would likely need consolidated
anticipated value propositions for them, as data as opposed to data on individual
well as identifying the service and revenue companies from the other side of the
model options that seemed most feasible. platform.

While there were several opportunities As a next step, Company C proposed

mapped, the key questions we needed starting a validation round, during which
to answer were neatly summarised by potential customer groups would be
Company C’s CEO: interviewed and presented with mock-
ups of products/services. Revenue model
“Where is the money? Who perceives most value for options would also be tested. The aim of
our assets? Can we create something unique that this exercise would be to measure and
Google or Facebook or LinkedIn cannot?” prove added value for potential customer
groups, thanks to data and algorithms.
In other words: 1) how can Company C
create data that is unique, reliable and Another important next step we discussed
abundant enough that customers would in the co-creation session was to involve
be willing to pay for it in the given market, Company C’s local sales teams in the
validation work. This would ensure that they Case study #4: A telecoms
understood the new data and AI-enabled
business models early enough to be able to
operator company
successfully implement them later on, i.e. (Company D)
once the new products were developed.
Key themes: Telecoms, B2B customers,
Finally, Company C would need to realign its offering extension, prioritisation,
organisation to reflect whatever strategic resourcing, transformation story, scaling
choice (option 1 or 2, as discussed) it up.
decided to run with. It would likely form a
separate growth team to experiment with Company description: Company D has
the new service and revenue models. But operations in several countries and employs
this team would also need to align and more than 20,000 people. Company
share learnings with Company C’s existing D serves both consumers (B2C) and
platform-developing organisation, to corporate customers (B2B), but our co-
maintain consistency in the work. creation session focused on its B2B side.
We met three people from Company D: its
Key learnings: Innovation Manager, AI Lead and Head of
• A key learning in this co-creation Unit.
session was that a business model
overturn, as introduced in our playbook Company D wants to find ways to help its
in Chapter 5, can be a realistic corporate customers get more out of their
alternative revenue model for a digital data that was created and collected in
platform provider such as Company C. various IoT systems. So far, Company D’s
But if Company C chose that path it may role has mostly been to deliver connectivity
have to abandon its existing revenue products and services. Now it wants to
model: it could be problematic to drive expand its services towards data analytics,
several business models in parallel recommendations and AI applications – all
(see Chapter 7.3 for more discussion aimed at creating added value for its vast
on business model innovation in small customer base.
firms).
• Company C is very familiar with the Implementing data and AI-enabled business
process of testing new business models, models seems like a natural progression
as well as creating new competences for Company D, as a traditional telecoms
and digital concepts. So, from the operator. And a move in this direction could
perspective of company transformation, help it win more telecoms business, since
its foundation for business model its corporate customers will appreciate
renewal is solid. data and AI being part of the total telecoms
package.
Company D has access to abundant data 3) getting the people in Company D’s
and continuously collects more from its organisation excited about the new data
corporate customers. However, typically and AI-related business potential, 4)
this data is owned by the customers, dedicating adequate human resources
allowing Company D only limited rights to to the prioritised new business projects,
use it. Data processing also requires a lot of 5) enabling Company D’s sales teams to
work, with plenty of time and effort needed approach the right type of customers with
to convert it into a format suitable for right messages, services, revenue models
analytics and AI applications. and sales models, and finally 6) engaging
customers to view Company D as a reliable
Company D already has several proofs partner in data and AI-related matters.
of concept in place around data and AI,
and more are likely to come. Overall, its To tackle these obstacles, Company D
foundation for data and AI-enabled business needs more “quick wins” and evidence that
models seems strong. the new businesses are lucrative, as well
as a more coherent transformation story
Opportunities and challenges: In our around data and AI models to tell within
co-creation session we considered the and outside of the company. Company D
varying needs of Company D’s corporate also needs to make a few strategic choices,
customers e.g. small companies, large going forward: who exactly are the focal
industrial corporations and large consumer corporate customers for data and AI? What
business companies. We also ideated precisely is the product/service being
new value propositions and products/ provided to them? How is the business
services (following our playbook laid out carried out for maximum revenue impact?
in Chapter 5). These included stronger Therefore, Company D’s next challenge is
insights, savings in energy consumption, to take its data and AI models from nice
data platforms, dashboards, predictive pilots and early wins towards a full-blown
analytics, cloud services and consulting workable business.
services that could be provided to Company
D’s corporate customers. Finally, we also Key learnings:
came up with revenue model options that • In large innovative companies like
could be applied to these new services, e.g. Company D, it is relatively easy to start
monthly subscriptions, consulting fees and data and AI- related pilot projects and
performance-based fees. create proofs of concepts.
• But once initial efforts have proved
The key obstacles to success were related successful, the real challenge lies in
to 1) prioritising the many data and AI- prioritising initiatives and scaling up the
enabled ideas and emerging business business (see also Chapter 7.3).
opportunities, 2) raising Company D’s • There is also a need to create a coherent
internal maturity and awareness around narrative – a shared “north star” that
data and AI-enabled business models, defines direction – to attract both
 internal company support and target E is considering various partnerships that
customers (see Chapter 7.3 for more). would enable it to access more health-
related data to complement and enrich its
own archive.

Case study #5: A healthcare The aim is to consolidate these data and
AI assets into a continuously calibrating
service provider company predictive solution. Company E already has
(Company E) version 1.0 of this plan in place, in the form
of a unique, solid and valuable data model.
Key themes: Healthcare, predictions, But it is constantly being developed.
competence development, resourcing,
customer validation, technology licensing, Overall, Company E sees data and
spin-off, business plan. AI-enabled business as a strategic
matter, enabling significant new growth
Company description: Company E is a opportunities. Data and AI models are
healthcare service provider that employs seen to impact company valuation in a
500+ people. There were three people very positive way, allowing Company E to
present in our co-creation session: the Chief reposition itself as a leading player in the
Strategy Officer, responsible for new growth field.
opportunities; the Programme Director,
responsible for data-related matters; and Opportunities and challenges: Company E
the Development Manager, responsible for considers its current corporate customers
preventive coaching services. as a focal target group for new data and
AI services in the local market. There is an
Company E has already undertaken opportunity for it to enhance its current
extensive work on its huge archive of patient healthcare services provision offering
healthcare data collected to the cloud. For with data and AI-related new products/
example, it has worked intensively to link services (e.g. health predictions and
various test results and genetic information, recommendations). Doing this could help
in order to credibly and reliably identify Company E win more corporate customers,
early health issue signs/risks. e.g. in local competitive biddings.

Company E already provides its patients Several local customer groups such as
with personalised recommendations for health equipment manufacturers, doctors,
proactively improving their own health. insurance companies, drug manufacturers,
Soon it intends to use its anonymised and drug distributors and governmental bodies
aggregated data to predict health issues are potentially willing to pay for Company
for collective groups of patients (see also E’s health-related data. The customer
Tanner87), e.g. to look at its corporate groups’ ability to harness data and AI in
customers’ employees as an entity. Company this situation varies. For example, whereas
health equipment manufacturers have e.g. finalising products, revenue models,
advanced data skills, insurance companies price points and commercial terms. It also
and doctors aren’t yet at the same level of needs to evaluate and prioritise between
literacy. several different customer groups and
lucrative opportunities, both locally and in
One radically new business opportunity the international market.
that came up in our co-creation session is
that Company E could productise its health A logical next step For Company E is to
prevention data and AI assets and license start validation discussions with potential
them abroad. This would provide healthcare customer groups, to understand their
companies that are lagging behind Company needs and preferences in more detail and
E with the technologies they need to catch to assess their ability to pay for data and
up. Company E saw this idea as a major AI-enabled offerings.
opportunity: it would divert significantly
from its core business, yet it would also For now, we assume that Company E is
offer significant new growth projections likely to apply models such as selling
that could reach well beyond the size of data, analyses & reports and selling data
the current, and limited, local market. platforms – i.e. the product-oriented
This new business opportunity would also models (as in Chapter 5) – to complement
enable Company E to take a completely new its existing healthcare services. In our
position in the value chain of the healthcare co-creation session, we also considered
industry (moving from a service provider to other models such as selling outcomes
a developer of enabling technology). (e.g. pricing is tied to health improvement
results) or selling capacity (e.g. providing
For Company E, the challenges of data and doctor appointment capacity against a fixed
AI commercialisation relate to applying recurring fee). But both these options felt
the technology successfully. It’s currently a little out of reach from the day-to-day
hard to be sure which data points or reality of Company E. There are several
combinations of data truly predict health good options available for revenue models
problems, and how to identify systemic for Company E, for example monthly or
issues in health. Company E does not yet annual technology licensing fees. But the
have in-house data science capabilities, business model overturn model could also
nor the required technical capability to potentially work if Company E decided to
build systems around data and AI assets. target its consumer market. In other words,
So, there are still several competence it would give consumers health services for
development and resourcing issues to be free in order to collect valuable anonymised
resolved before Company E’s data and AI and aggregated health data that would then
ambitions can be scaled up. be monetised in the corporate customer
market.
Company E still has many decisions to make
around the commercialisation of its data,
At the time of our co-creation session, Case study #6: Two car
Company E was about to consolidate a
business plan around its data and AI efforts.
manufacturing companies
That document will give answers to many (Companies F and G)
of the prioritisation and commercialisation
questions we’ve discussed here. The plan Key themes: Automotive, idea generation,
will also need to draw a fine balance in data provision, customer validation,
weighing up incremental new opportunities breaking silos, communities, culture change.
in the local home market versus the
radical new opportunities of international Company description: We held two
technology licensing. consecutive co-creation sessions with two
large automotive companies – Companies F
Spinning off the international technology and G. Both companies have already worked
licensing business to a different unit or extensively on their digital services and
legal entity is another viable alternative strategies but are interested in exploring
here. This would allow Company E to easily new data and AI business opportunities to
separate its current core business from one complement their existing digitalisation
that may require different competences, plans. Specialists from the respective digital
resources and ways of working.70 innovation units of the firms were present
in both firms (in Chapter 3.4 we discuss
Key learnings: the fact that large companies often set up
• Before our co-creation session with separate units for business model innovation
Company E, we anticipated that the – this was exactly the case for Companies F
privacy and legal issues in healthcare and G). We discussed the current situation
would prevent companies in this and the changes that they wanted to make in
industry from being able to do much the automotive industry overall.
with their data and algorithms.
• Company E’s case study proved that Incumbents in the European car
assumption wrong. If a healthcare manufacturing industry are currently facing
company manages to collect pressure from firms such as Uber, Waymo
anonymised and aggregate health data and Tesla, whose technology-enabled new
from its patients to predict health concepts are forcing traditional players
issues, there are good opportunities to to defend their positions. Typically, the
commercialise that knowhow. There are incumbents’ challenge is to define where
no ethical or legal issues preventing it they stand with their software stacks,
from doing so, provided that patients’ which will empower self-driving vehicles,
personal information is kept strictly on-journey entertainment, electrification,
undisclosed (see Chapter 8.2 for more and the drivers’ user experience. Their
on the ethics of this topic). dilemma is also how to avoid becoming
commodified hardware at the bottom of
the stack in an economy where data and
software are creating the greatest value for Therefore, it’s hard to establish a shared
customers (this challenge is also common to point of view on data and algorithms.
various manufacturing firms, as discussed
in Chapter 3.3). The most dangerous To make things more complicated, even the
competitors here actually come from Silicon car dealerships only have access to data
Valley, not (as one might assume), Japan or on their own operations. This is stored in
China.133 various different local IT systems. So, the
dealers face difficulty in petitioning for the
Opportunities and challenges: The kind of universal, large-scale data repository
co-creation sessions with automotive that would inform personalised suggestions
Companies F and G revealed that they were for their customers.
still very much in the starting phase when
thinking of harnessing data and AI for Finally, we also considered legal matters in
business model renewal. While innovation our co-creation sessions. Who owns the data
was instinctive for both companies, they from cars: the driver, the dealer or the car
lacked inspiration and ideas on how to use manufacturer? Which consents are needed
data for business purposes (other than for the car manufacturer to start using data
automating and improving the existing for business purposes? Companies F and G
processes and operations, a process that were not always clear what was acceptable
was already underway). Car firms’ data practice here: there was a tendency to avoid
and AI services often look quite similar, any risk-taking with data.
e.g. they are formed around car location
information, or they aim to improve in- It’s clear that the traditional business model
car user experience. In this context it’s of Companies F and G still works well within
difficult to think outside of the box and the classic car industry framework. This
identify opportunities that reach beyond reduces their sense of urgency for any
the obvious. Overall, the sentiment in both business model innovations and stands
Companies F and G is that, even though in the way of the kind of mindset change
data and AI are strategically important needed for new data and AI concepts.
assets, they are not yet adequately used for
business purposes. Yet, at the same time, it was clear from our
co-creation sessions that Companies F and
In addition, it was unclear in our co- G had an imperative to move ahead. One
creation sessions what data and algorithms comment illustrates this need well:
Companies F and G possess, given the scale
of their companies and the various data “Car manufacturing is not a future-proof business,
collection methods/ algorithm development so we need to invent new business models, and use
already in play. These large companies data as an enabler for them. There’s so much room
tend to be quite siloed, with different for improvement. If we don’t do it, our competitors
departments working on their own projects. will.”
As a next step, car manufacturer F is about • Instead, their edge might come from
to start a series of innovation workshops using design thinking to dig deeper into
to generate more ideas for data and AI the needs of customers and partners,
commercialisation. The plan is to break addressing their real problems worth
internal silos by inviting representatives of solving (as discussed in Chapter 3.2).
various departments to the room, as well as This would also allow them to form
potential external customers and partners. strategic partnerships (see Chapter
Several potential external customers 5.9) with players outside the automotive
and partner groups were identified in industry, thereby unlocking value on the
our sessions, ranging from cities and margins of industry boundaries.
municipalities to car insurance providers, • In addition, car manufacturing
navigation companies and smart home companies that are able to become
providers. Car owners as the customers and more collaborative, more open and more
end-users of products, as well as car dealers proactive in their approach to data and
as the middlemen, would be invited to the AI will have competitive advantage over
workshops, too. their rivals in terms of culture and ways
of working (this is discussed further in
Car manufacturer G, on the other hand, is Chapter 7.7).
striving to build a common data platform for • Innovation workshops, in which various
various internal departments to use. The aim people get together to innovate valuable
of this is to create an internal community of solutions, might be an important next
people who are playing with data, as well as step towards building new data and
setting up company-wide revenue targets AI-enabled businesses in these firms.
and KPIs for data and AI initiatives. Company Internal data communities and shared
G also wants to create 360-degree views of cross-company platforms and targets
car drivers based on various data points, as will also be key to development.
a prototype for developing a stronger range
of personalised services. This company also
aims to break silos and move towards a
common data and AI plan that is shared by Case study #7: A
their entire firm.
construction services
Key learnings: company (Company H)
• When it comes to data and AI-enabled
business models, the competitive Key themes: Construction, data-
advantage of Companies F and G is not enabled leadership, open data platform,
derived from unique data (as various transparency and monitoring, data
car companies are able to collect similar maturity, culture change.
data), nor does it come from new ideas
related to data and AI usage. Company description: Company H runs
various construction service operations
that are targeted at both consumers and the design, construction and maintenance
companies. It operates locally and employs phases of a building project.
300+ construction professionals. In our co-
creation session, we had a good sparring Overall, Company H has a big opportunity to
discussion with the Chief Financial Officer plan construction projects more effectively
who is experienced in leveraging data. and cost-efficiently for the benefit of their
During his earlier career he had designed customers. This relies on some parts of the
various reports, predictions and data construction work becoming standardised,
models for companies to use. as well as better data harnessing to
understand the issues and bottlenecks in
Company H has already worked intensively key construction processes.
on its data. Data-enabled leadership is a
critical strategic theme across Company In addition, Company H has a nascent
H: there is a major opportunity for it to opportunity to run remote control building
improve its business, using data. Company maintenance services using data. But these
H has a data warehouse in place to services are still in early development and
consolidate data from various systems are not widely adopted by customers.
which enables the creation of insights from
construction operations. One obvious challenge in the construction
industry is that many activities, including
In addition, Company H recently created a hour reports or drawings, are still
new Head of Data role to promote and drive produced in paper format. Employees in
all its data-related activities. It also has a the industry are also used to working with
data-enabled control room in place which drawings and qualitative information,
provides a real-time view of the company’s but not yet with numbers or quantitative
construction sites. data. And, as the lifecycle of construction
projects cover several steps from planning
Opportunities and challenges: Company to building and maintenance, plans often
H has an ambition to collaborate with its change over time. A coherent view of the
partners on an open data platform it has entire lifecycle is often difficult to establish
created. This would enable it to orchestrate from data.
the value chain, share important
information, co-ordinate construction work There is also ongoing debate in the
more effectively and eliminate waste in construction industry and its labour unions
construction projects. Some partners have about how transparent, monitored and
already joined the platform, so the future of data-enabled construction work can be in
this initiative looks promising. the future.

Company H also has multiple pilot projects Key learnings:

in place with selected customers in which a • When it comes to building new business
digital twin has been leveraged throughout models on data and AI, the low maturity
 of the construction industry is clearly an
issue (see also Chapter 4.2).
• Current data and AI-related activities
are mostly focused on improving
internal efficiency and adding more
transparency in operations.
• If maturity in these areas increases,
the next step could be to build new
paid-for services (e.g. data dashboards)
for various stakeholders.
• But this model is not feasible yet, and
there are currently no examples of
construction firms that have achieved
such an approach. Company H’s first
step is to drive a culture change
towards data and AI centricity (as
discussed in Chapter 7.7), and in the
construction industry overall.