Generative AI and its Impact on
Business:
1. What is Generative AI?
Definition: Uses AI models to create content such as text, images, music, or code, based on massive
datasets.
Example: In a marketing agency, generative AI could help create personalized advertisements or
content by analyzing customer data, making it easier to engage each customer with relevant
messaging.
________________________________________
2. Models used in Generative AI:
GANs (Generative Adversarial Networks):
Components: Generator and Discriminator
Functioning: The generator creates content (e.g., dishes) and the discriminator evaluates
whether the content is realistic.
Example: In art generation, the generator might create digital art, and the discriminator compares it
to real-world art to see if it’s realistic enough. If the discriminator can’t tell the difference, the art is
considered high-quality.
Transformers:
Known Models: GPT (Generative Pretrained Transformer) and BERT.
Mechanism: Use attention mechanisms to focus on important parts of the input data.
Example: For email marketing, if you want to send personalized emails, the model can focus on key
details from customer data (like their past purchases) to create a personalized message that makes
sense and feels relevant to the recipient.
________________________________________
3. Applications of Generative AI:
Marketing and Customer Engagement:
- Creation of personalized content at scale: emails, dynamic web pages, etc.
- Example: A fashion brand can use AI to send customized newsletters to customers based
on their browsing history or favorite products.
Operations and Logistics:
- Forecasting demand trends and optimizing the supply chain.
- Example: An e-commerce platform uses AI to predict when popular products will run out
of stock, so they can reorder in time to meet customer demand and avoid overstocking.
Human Resources:
- Generating job descriptions and personalized training simulations for employees.
- Example: A HR department uses AI to automatically create job descriptions that attract
diverse candidates, ensuring a broader applicant pool and reducing bias in the hiring
process.
Document Management and Technical Writing:
- Analyzing documents to replicate the necessary format, style, and technical language.
- Example: Law firms use AI to draft contracts or analyze legal documents to ensure they
comply with current regulations.
Adapting Writing Styles:
- Maintaining a consistent brand voice across different types of content.
- Example: A social media manager uses AI to generate posts that maintain the same tone
and style as the company’s previous posts, ensuring consistency across all platforms.
Customer Service:
- Intelligent chatbots that can handle customer queries in real-time.
- Example: E-commerce websites use AI-powered chatbots to answer customer questions
instantly, like "Where is my order?" or "Can I return this item?"
________________________________________
4. Limitations and Ethical Considerations:
Limitations:
- What it means: AI may not always be accurate. It depends on how much and what kind
of data it has been trained on.
- Real-World Example: If a shopping website's AI is trained only on high-end product
descriptions, it may struggle to write effective descriptions for budget-friendly items.
- Output quality can vary depending on the training data.
- Inaccuracies may occur if the data is insufficient or outside the model's domain.
Biases:
- What it means: AI can inherit biases from the data it was trained on, leading to unfair or
discriminatory results.
- Real-World Example: If a recruitment AI is trained mostly on resumes from male
candidates, it may favor resumes that appear similar to the training data, unintentionally
disadvantaging female applicants.
- Models may amplify biases present in training data, leading to unfair outcomes.
Ethical Issues:
- Data Privacy: How to protect personal data used by AI.
- Intellectual Property: Who owns the content generated by AI?
- Misuse: Could AI be used to create harmful content like fake news or deepfakes?
- Real-World Example: A news organization could use AI to generate articles quickly, but if
the model is trained on biased or incorrect information, it might produce misleading or
harmful content
- Data Privacy: How to protect the data used to train AI?
- Intellectual Property: Who owns the content generated by AI?
- Misuse: Risks of generating harmful content (e.g., fake news, deepfakes).
________________________________________
5. Conclusion:
Generative AI is revolutionizing various business functions, but its successful integration depends on:
- A solid understanding of the technical foundations.
- Responsible management, with strong ethical standards.
- Optimizing repetitive tasks while preserving human creativity.
________________________________________
Key Takeaways:
GANs = Generator (creation) + Discriminator (evaluation).
Transformers = Selective attention to generate coherent text.
Applications: Marketing, logistics, HR, writing, customer service.
Generative AI Terminology Revision
Sheet
1. Artificial Intelligence (AI)
- Definition: AI refers to systems that can perform tasks that usually require human
intelligence, like learning, problem-solving, and language understanding. It ranges from
simple rule-based systems to complex neural networks that mimic human thinking.
- Real-life Example: Imagine a smart assistant like Siri or Google Assistant. It understands
your questions and provides answers or executes commands by learning from your past
interactions.
2. Machine Learning (ML)
- Definition: ML is a subset of AI where machines learn from data to make decisions
without explicit instructions. The system improves as it processes more data.
- Supervised learning: Trains on labeled data to predict outcomes (e.g., predicting house
prices based on past data).
- Unsupervised learning: Finds patterns in unlabeled data (e.g., clustering customers
based on buying behavior).
- Reinforcement learning: Learns by trial and error, receiving feedback from its actions
(e.g., teaching a robot to walk).
- Real-life Example: Google Maps uses supervised learning to predict traffic patterns based
on historical data. It can suggest the fastest route by learning from millions of trips.
3. Deep Learning (DL)
- Definition: A branch of ML using neural networks with many layers. It's especially good at
handling complex data like images and speech.
- Real-life Example: DL is behind facial recognition technology. When you unlock your
phone with your face, a deep learning model is identifying your face from thousands of
possible images.
4. Neural Networks (NN)
- Definition: Neural networks are computational models inspired by the human brain.
They have layers of nodes (neurons) that process and analyze data.
- Real-life Example: Spam filters in email use neural networks to identify unwanted emails
by analyzing patterns from millions of emails they've been trained on.
5. Generative Adversarial Networks (GANs)
- Definition: GANs consist of two networks: a generator that creates fake data and a
discriminator that tries to distinguish between real and fake data. The two networks
"compete" to improve each other, generating highly realistic outputs.
- Real-life Example: Deepfake videos use GANs to generate hyper-realistic videos, where a
person’s face is swapped with someone else’s. It's like having a fake but convincing video
of a celebrity saying something they never said.
6. Natural Language Processing (NLP)
- Definition: NLP allows computers to understand and generate human language. It's used
in translation, chatbots, and voice assistants.
- Real-life Example: When you chat with a customer service bot, NLP is working behind the
scenes to understand your messages and respond with useful information.
7. Transformers
- Definition: Transformers are models that process all the words in a sentence
simultaneously rather than one at a time. This helps them understand context better,
especially in long texts.
- Real-life Example: Google Translate uses transformers to translate entire sentences by
understanding the context of each word within the sentence, not just the word-by-word
translation.
8. Generative Pre-trained Transformers (GPT)
- Definition: GPT models (like GPT-4) use the transformer architecture to generate human-
like text. They're pre-trained on massive amounts of text data to understand language
and context.
- Real-life Example: When you use AI-powered writing assistants, such as ChatGPT, it
generates coherent responses based on your input. The model was trained on a vast
amount of text to write or assist in writing.
9. Tokenization, Word2vec, and BERT
- Tokenization: Splitting text into smaller chunks (tokens) like words or subwords, which
makes it easier for models to process.
- Word2vec: A technique that turns words into numbers (vectors) to capture word
relationships. Words with similar meanings have similar vectors.
- BERT: A transformer model that looks at words in context (both left and right of a word)
to understand deeper meanings.
- Real-life Example: Word2Vec can help in recommendation systems. For instance, if you're
reading an article about "coffee," the system might suggest articles about "tea" based on
their similar word vectors.
Conclusion
Generative AI is not just about making AI "smart" – it’s about creating systems that can generate new
content. By understanding these terms, you can start to see how AI is shaping industries from
healthcare to entertainment.
Generative AI - Under the hood
Generative AI refers to systems that not only understand language but create it—producing content
like text, images, and even music. Unlike traditional AI that mainly analyzes data, generative AI
proactively generates new content.
Key Technologies:
- Machine Learning: The core of generative AI, broken into two main types:
o Supervised Learning: Models are trained using labeled data (e.g., a dataset of articles
with their summaries) to learn specific tasks.
o Unsupervised Learning: Models work with data without labels, learning patterns and
relationships to create new data.
- Neural Networks: A key technology in AI, especially Generative Adversarial Networks
(GANs) and Variational Autoencoders (VAEs). These help AI generate new data by
learning from previous patterns.
- GANs: Work by having two models—one generates data, and the other evaluates it.
- VAEs: Encode data into a compact form and then decode it to recreate the original or
generate new samples.
- Transformers: The Backbone of NLP. Transformers revolutionized Natural Language
Processing (NLP), helping machines understand the context of words in sentences.
Example: BERT (Bidirectional Encoder Representations from Transformers) and GPT
(Generative Pre-trained Transformer) are key models that can generate coherent,
contextually appropriate text. Real-life Example: When you use Google search, BERT
helps understand your query better by considering the entire context, not just individual
words.
- Self-attention mechanism: This lets models look at different parts of a text
simultaneously, improving learning efficiency and context understanding.
Word2Vec and Semantic Understanding
Word2Vec maps words into a geometric space, where words with similar meanings are close
together. Example: The words "king" and "queen" might be close in this space because they share
similar contexts.
Generative AI in Business
Generative AI has practical applications in various business areas:
- Content Generation: AI models like GPT can generate articles, blogs, and social media
posts. These models are trained on diverse data sources and can be fine-tuned for
specific topics.
- Personalization: Using data pipelines (e.g., Apache Kafka), AI processes real-time user
data (from websites, apps, social media) to generate personalized content.
- Automation: Robotic Process Automation (RPA) combined with AI models helps
automate repetitive tasks, saving time and reducing errors.
- Innovation: Generative AI can help businesses predict future trends, simulate scenarios,
and assess risks, aiding in the creation of new products or strategies.
How Generative AI Works Technically
Neural Networks (especially GANs and VAEs) are central to generative AI. They learn from data and
generate new content, such as:
- Text (e.g., articles), Images (e.g., synthetic images generated from descriptions), Music (e.g.,
AI-composed songs), Real-life Applications and Examples
- Content Creation: GPT-3 generates articles, product descriptions, and creative content.
- Personalized Marketing: AI personalizes user experiences based on their online behavior,
generating product recommendations or advertisements tailored to their interests.
- Synthetic Image Generation: AI can generate entirely new images or alter existing ones based
on parameters.
The Program Life Cycle
Program initiation
- Defining program objectives and scope
- Key activities:
o Define program objectives
o Define program scope
o Define program benefits
o Develop program charter
o Identify and analyze stakeholders
o Establish program governance (role, responsabilities..)
Program planning
- Developing detail plan that guide program execution
- Key activities:
o Develop program management plan that includes schedule budget resource
allocation risk management strategy expected benefits and communication plan
o Establish milestones and deliverables to track progress and keep the program on
course
o Align the project plans with the overall program objectives and timeline
o Risk management by identifying risks and developing mitigatin strategies to minimize
their impacts
Program delivery
- Coordinate project activities to ensure that the projects’ progress contributes to overall
objectives
- Monitor program performance: Track the progress again the benefits realization plan using
KPIs and make adjustments when needed
- Optimize efficiency by managing resources
- Carry out stakeholder communication by providing regular updates on progress and
addressing any concerns
Program closure
- Finalizing activities, meeting objectives and documenting lessons learned
o Deliver final outcomes
o Conduct a program review by evaluating success, document lessons learned and
identifying improvement areas
o Integrate the outcomes into operational teams and processes
o Close program governance: close accounts and release resources
Governance structures and processes
Program governance is a structured approach that ensures the alignment of a program’s objectives
with organizational goals providing oversight and control throughout its lifecycle to achieve
successful outcomes.
It’s a framework based on program definition, planning, benefits realization and closure. It ensures
programs align with an organization’s strategy are adequately resourced and are effectively managed
to deliver value.
- Decision making: Informed decisions based on accurate data, comprehensive analysis and
strategic alignment.
o Informed decisions
o Timely decisions
o Clear criteria
o Decision quality
o Enhancing transparency
- Performance monitoring ensures that programs stay on track to meet their goals. With
governance metrics are regularly reviewed, adjustments are made and the program aligns
with the objectives.
- Continuous monitoring Track progress across multiple projects and make data-driven
adjustments to ensure success.
- Risk management critical to program governance
- Stakeholder engagement critical to program governance. Ensure their perspective are
considered in decision making. Stakeholders are informed and involved, fostering trust,
collaboration and valuable insights that guide decision making and program execution
- Strategic alignment: ensures all programs and projects are aligned with the overall
organization’s strategy
Strategic decision-making
Decisions made at a program level:
- Have far reaching implications
- Impacting the success of individual projects
- Impacting the program alignment with the organization’s goals
There are key factors that influence decision making
- Organizational culture
- Stakeholder influence: consider the needs of customers, investors and team members to
impact decisions positively
- Risk tolerance
- Available data : data driven decision are vital.
The importance of decision-making frameworks
The key decision-making frameworks in programs include:
Analytical framework: relies heavily on data, logical reasoning and structured processes.
o Key characteristics:
Ample data available for analysis
Predictable, stable conditions
Time available for thorough evaluation
Intuitive framework: relies on experience and instinct in decision making. This framework is
less structured but highly valuable when time is limited or decisions must be made in
environments of high uncertainty.
o Key characteristics:
Limited or ambiguous data
Rapid decision-making required
High complexity and unpredictability
Collaborative framework: prioritizes the involvement of multiple stakeholders in the
decision-making process. This approach aims to achieve well rounded and widely supported
decisions by harnessing diverse perspectives.
o Key characteristics:
Diverse stakeholder involvement
Need for broad support and buy-in
Complex issues requiring varied experience
Integrated framework: It is the decision-making management framework that can
incorporate the three previous approaches—analytical, intuitive, and collaborative—
depending on the context. It structures how and when to apply analytics, intuition, or
collaboration. It defines roles, levels of authority, tools, timing, and other key elements
involved in the decision-making process.
o Key characteristics:
Combination of structured and unstructured conditions
Need to balance data-driven and intuitive decisions
High stakeholder engagement with uncertain outcomes
The program sponsor
The program sponsor is a senior executive who ensures the alignment with the organization's
strategic goals and adherence to policies and standards.
Key attributes:
Strategic vision
Provides direction
Strategy alignment
Understand the technology being developed and the sector’s specific needs
Ability to communicate effectively (stakeholder communication)
Understanding strategic planning
Strategic planning: define the long term goals and determine the best approach to steer the
organization’s overall direction.
It asks WHERE THE ORGANIZATION WANT TO BE IN THE FUTURE?
It provides an organization with:
A roadmap
It helps that all parts of an organization work towards the same objectives
Program managers must evaluate project proposals which is a formal document outlining the
objectives scrop and plans for a project aiming to secure approval or funding from stakeholders or
decision making.
A project that is no aligned can:
Drain resources
Divert attention
Project managers must assess whether a project is aligned with the organization’s goals by
considering the project feasibility (resources, potential risks, ROI) and prioritizes projects that have
greater strategic values.
They must be proactive.
Strategic plans set the direction while operational plans detail the steps needed.
How the vision, mission, and action plans interconnect
While the vision statement, mission statement, and action plans are distinct components of a
strategic plan, they are deeply interconnected. The vision statement sets the organization's long-
term aspirations, the mission statement defines its current purpose, and the action plans provide the
roadmap for achieving both. Together, these elements ensure that the organization remains focused
on its strategic objectives while being flexible enough to adapt to changing circumstances.
Common pitfalls in strategic planning
While the components of a strategic plan are essential, organizations must avoid common pitfalls to
prevent strategic failure. These pitfalls include:
1. Lack of clarity: If the vision or mission statements are vague or overly complex, they can
confuse rather than guide the organization. All stakeholders must understand these
statements clearly and concisely.
2. Overambition: While a vision statement should be aspirational, it must remain attainable.
Unrealistic goals can lead to frustration and disengagement.
3. Poorly defined action plans: A strategic plan needs clear, specific, and realistic action plans.
Without well-defined steps, even the best vision and mission statements will remain
theoretical ideals with little impact on operations.
4. Failure to measure progress: Organizations often fail to track progress toward strategic
objectives. A strategic plan can quickly become outdated or irrelevant without regular
updates and relevant metrics.
5. Lack of flexibility: A strategic plan must be adaptable. As the business environment changes,
organizations need to adjust their plans to stay competitive.
Program roadmaps
A program roadmap is a visual representation outlining the progression of projects and initiatives
within a program. When fully developed, it is a strategic plan with program goals, milestones,
deliverables and timelines.