AI Forward Deployed Engineering — Research Memo

A premium practitioner memo on the rise, evolution, and strategy of AI Forward Deployed Engineering.
Covering Palantir, C3 AI, OpenAI, Anthropic, and Google Cloud — with a framework for the ideal FDE team and operating model.

---

Overview

This repository contains the full research and drafting pipeline for a comprehensive memo on AI Forward Deployed Engineering (FDE) — one of the fastest-growing and least-understood functions in enterprise AI.

The memo analyzes the FDE landscape across five market leaders, surfaces the unresolved organizational and economic tensions these teams face, and proposes a framework for the ideal FDE pod structure and operating model.

Primary Audience: Enterprise AI leaders, GTM executives, and technical founders evaluating or building FDE capabilities.

---

The Memo

📄 Read the full memo →

The memo covers six sections:

#	Section	Description
1	Origins & Definition	The integration-and-outcomes gap; Palantir's historical blueprint; the Build-Prove-Generalize loop
2	Competitive Landscape	Five deployment models compared: Palantir, C3 AI, OpenAI, Anthropic, Google Cloud
3	Foreseeable Challenges	The pricing/profitability dilemma; the failure of engineering-only pods
4	The Ideal FDE Pod	The three-role model: Strategist, Engineer, Data Scientist
5	The Ideal Operating Model	Three operating imperatives: model-agnosticism, open-source portability, output-based pricing

---

Repository Structure

fde-memo/
├── AGENTS.md                        # AI agent coordination & architectural guidelines
├── README.md                        # This file
│
├── docs/
│   └── writing_style.md             # Tone, structure, and formatting guidelines
│
├── inputs/                          # Drop source PDFs or text files here for research
│
├── output/
│   ├── outline.md                   # Structured memo outline
│   └── memo.md                      # ✅ Final memo draft (6 sections, ~5,000 words)
│
├── reference/
│   ├── 2026 AI Observations.pdf     # Recent 2026 observations on AI engineering
│   └── Adobe - Field PM memo.pdf    # Structural reference: Adobe's Field PM memo
│
├── research/                        # Per-company research notes and job description analyses
│   ├── palantir_fde.md
│   ├── c3_ai_fde.md
│   ├── openai_fde.md
│   ├── anthropic_fde.md
│   ├── google_cloud_fde.md
│   ├── palantir_delta_jd.md
│   ├── palantir_echo_jd.md
│   ├── c3_ai_fde_jd.md
│   ├── c3_ai_fdds_jd.md
│   ├── c3_ai_ai_solution_manager_jd.md
│   ├── openai_fde_jd.md
│   ├── openai_ai_deployment_manager_jd.md
│   ├── anthropic_fde_jd.md
│   ├── anthropic_solutions_architect_jd.md
│   ├── google_cloud_fde_jd.md
│   └── google_cloud_ai_consultant_jd.md
│
├── skills/
│   ├── research_memo/
│   │   └── SKILL.md                 # End-to-end memo research & drafting workflow
│   └── scripts/
│       ├── export_docx.py           # Export memo to DOCX dynamically using markdown filename
│       ├── export_pdf.py            # Export memo to PDF dynamically using markdown filename
│       ├── read_docx.py             # Read and parse text from DOCX
│       └── read_pdf.py              # Read and parse text from PDF
│
└── tmp/                             # Temporary scripts and logs (not committed)

---

Exporting the Memo

The memo can be exported to DOCX or PDF using the included scripts. Ensure dependencies are installed first:

uv sync

By default, running the scripts without any arguments will read from output/memo.md and dynamically output to output/memo.docx or output/memo.pdf in the same folder.

Export to DOCX

uv run python skills/scripts/export_docx.py [input.md] [output.docx]

Export to PDF

uv run python skills/scripts/export_pdf.py [input.md] [output.pdf]

---

Reading/Parsing Documents

You can extract text from compiled .docx and .pdf documents directly to stdout or a text file using the read utilities:

Read DOCX

uv run python skills/scripts/read_docx.py [input.docx] [output.txt]

Read PDF

uv run python skills/scripts/read_pdf.py [input.pdf] [output.txt]

---

Workflow

inputs/          →   research/        →   output/outline.md   →   output/memo.md
(source PDFs)        (synthesized          (structured plan)       (final draft)
                      notes per company)

1. Research & Synthesis — Source materials in reference/ and inputs/ are analyzed; findings are documented in research/ as per-company notes and job description breakdowns.
2. Outlining — Core themes and section structure are organized in output/outline.md.
3. Drafting — The full memo is written to output/memo.md following the style guide in docs/writing_style.md.
4. Exporting — The memo is compiled into a distributable format (DOCX/PDF) via skills/scripts/.

---

Key Findings (TL;DR)

• The 2026 AI market is a deployment competition, not a model competition.
• FDE teams comprise 30–40% of total headcount at deployment-native companies.
• No company has credibly solved the FDE pricing and profitability dilemma.
• The minimum viable FDE pod requires three roles: Strategist, Engineer, and Data Scientist.
• The ideal operating model is open-source, model-agnostic, and priced on outputs — not headcount.

---

Research current as of May 2026.