FDE: Forward Deployed Engineer

About this book

Third Edition, September 2026. Previously published as Shipping Enterprise AI: The Forward Deployed Engineer’s Playbook.

The Third Edition substantially expands the book to reflect how Forward Deployed Engineering has evolved with production GenAI and agentic AI: bounded coding-agent delegation, sub-agent and skill boundaries, retries and recovery, tool and API interoperability including MCP, identity and authorisation, AI unit economics, evaluation, observability and evidence-based deployment gates.

Two continuous worked cases

CASE-4832, a financial-services review workflow, starts with a deceptively simple requirement and progresses through service-contract definition, source authority, durable workflow state, delegated execution, partial-failure recovery, client-led handover and implementation review.

A shipment-exception service follows the field-delivery side of the role: a client’s broad request for an “autonomous agent” is narrowed into a governed production service through workflow discovery, first-30-days planning, decision rights, organisational readiness, handover and outcome measurement.

From client ambiguity to production

You will learn to define bounded workflows before choosing architecture, establish which systems have operational authority, build retrieval around admissible evidence rather than search relevance alone, design durable agentic workflows that survive retries and partial failure, and control which actions may actually be executed. The book also covers environment separation, contracts, testing, security boundaries, release evidence, rollback, cost per accepted unit of work, evaluation and observability.

Beyond architecture, it examines the professional practice of the FDE (discovery, the first 30 days, stakeholder power and decision rights, handover, proving business value) and the FDE career path: technical readiness, interviews, role boundaries and how organisations build and scale an FDE function.

Topics include

• The FDE operating model and enterprise AI production readiness
• Enterprise RAG, source authority and retrieval evidence
• Agentic AI, state, retries, idempotency and durable execution
• Coding-agent delegation and bounded engineering authority
• MCP, APIs, tools, identity and enterprise integration
• Security and trust boundaries, AI cost and unit economics
• Evaluation, observability and release evidence
• Financial services, healthcare and high-stakes deployments