The Forward Deployed Engineer Playbook
Get AI past the pilot.
A free 42-page field guide from Qexa Technology: how forward deployed engineers find, prove and scale enterprise AI value from inside the client, with every template you need.
Most enterprise AI never leaves the pilot.
Models stopped being the bottleneck. The hard part is engineering AI into how one specific business actually works: its data, its systems, its controls and its people. That is the forward deployed engineer's job, and the market is reorganizing around it.
- 95%
- of enterprise generative AI pilots showed no measurable P&L impact. MIT NANDA, The GenAI Divide, 2025
- 643 → 5,330
- US forward deployed engineer job postings, April 2025 to April 2026. Indeed data via Business Insider
- $5.5B+
- committed in 2026 to new deployment arms: OpenAI's Deployment Company and Anthropic's enterprise services venture. Company announcements, May 2026
Week 0 to Day 120, gate by gate.
The guide walks one engagement from the first qualification call to a clean handoff. Phases overlap; gates don't. Here is the spine of the method, and what each part of the guide gives you.
-
Week 0
Qualify before you deploy
Six checks that decide whether an engagement deserves a scarce FDE: sponsor, access, value, ambiguity, success criteria and your own prep.
Gate: sponsor, access and go/no-go agreed
-
Weeks 1 to 4
Discover and map the real work
The eight layers of any company, interview guides for seven stakeholders, two workshops, the day-one data request and the process card.
Gate: one falsifiable success sentence
-
Weeks 3 to 6
Prioritize and build the business case
A weighted value-and-feasibility score, the five-part lighthouse test, MVP scoping rules and payback math with a conservative case finance will trust.
Gate: lighthouse picked, baseline agreed
-
Weeks 4 to 6
Architect with governance built in
A seven-layer reference architecture for agentic AI, the autonomy ladder, and the identity, audit and evaluation controls risk teams sign off on.
Gate: security and risk sign-off
-
Weeks 3 to 10
Prototype on real data, evaluate first
An eight-week pilot plan, the evaluation harness you build before the build is done, and the unhappy paths production will hit.
Gate: KPI threshold met
-
Weeks 8 to 13
Go live and stay close
Production readiness, two to three weeks of hypercare, and the operating rhythm from weekly KPI checks to quarterly executive readouts.
Gate: live, monitored, on-call
-
Day 120
Hand off and productize
A handoff checklist that makes the client self-sufficient, value realization signed by finance, and the review that turns field work into reusable product.
Gate: ownership transferred, next use case funded
Write the sentence that ends discovery.
The guide's definition of done for discovery is one falsifiable sentence the sponsor agrees to. If you can't write it, discovery isn't finished. Try it on a project of yours.
20 chapters, four parts, one method.
Written for practitioners and for the enterprises that deploy them. Every chapter ends in something you can use: a checklist, a table, a template or a talk track.
Part I · The landscape
- What a forward deployed engineer is
- Why now: the deployment gap
- When the FDE model is the right answer
Part II · The engagement playbook
- Operating principles
- The engagement map and qualification
- Discovery and workflow mapping
- The workflow intelligence architecture
- Prioritization and MVP scoping
- Building the business case
- Architecture and governance for agentic AI
- Prototype, pilot and the evaluation harness
- Production, hypercare and handoff
- Value realization and productization
Part III · Building the function
- The operating model
- Economics and the commercial model
- Metrics that matter
- Hiring and developing FDEs
- Anti-patterns and how to spot a real FDE program
Part IV · Field kit
- Templates: weekly status, eval rubric, executive readout, qualification scorecard
- Your first 90 days and the talk track
Plus a glossary and 16 cited sources.
Written for whoever has to make AI work.
| If you are… | Start with | You'll leave with |
|---|---|---|
| An FDE, or preparing for an FDE interview | Part II, then the field kit | A phase-by-phase method, templates and a 60-second talk track |
| A solutions architect, SE or consultant moving into FDE work | Chapter 1, then Part II | The habits that separate FDEs from presales and services roles |
| An enterprise leader deploying FDEs | Chapters 2 and 3, then Part III | How to tell a real FDE engagement from slideware, and what to demand |
| A vendor or services firm building an FDE team | Part III | The structure, economics and metrics that make the model compound |
Written by Qexa Technology's Chief AI Architect.
Day 120
Your copy is ready to hand off.
Enter your name and email and the PDF is yours on the next screen. Free, 42 pages, October 2026 edition.
Questions
Is it really free?
Yes. No trial, no paywall. You get the full 42-page PDF right after you enter your name and email.
What will you do with my email?
Qexa keeps it to know who downloaded the guide. You'll only get occasional notes on deploying AI if you tick the box, and you can unsubscribe at any time.
Can I share the guide with my team?
Please do. Send them to fde.qexa.com so they get the latest edition.
Is it tied to one AI vendor?
No. The method is stack-agnostic. Chapter 7 maps each layer to IBM watsonx and to common alternatives, so you can apply it on whatever platform your client runs.
Can Qexa help with a deployment?
Yes. Qexa Technology builds private AI infrastructure and AI agents inside businesses using this method. Book a conversation with Alex.