What does a typical FDE engagement lifecycle look like for an AI project?

30-second answerSay your answer out loud first, then reveal.

Discovery (understand the business, users, data and constraints) → use case selection and success criteria → a time-boxed prototype or POC on real data with an eval set → pilot with real users in a limited scope, with production basics (security, monitoring, HITL) → production rollout and integration → adoption, measurement and iteration → handover to the customer's team, plus product feedback. Each stage has explicit exit criteria agreed with the customer.

The seven stages of an FDE engagement from discovery to handover, with a loop back from the POC to use-case selection when exit criteria are not met and learnings flowing to the product team.

Exit criteria examples

Stage	Exit criteria
POC	≥ 85% accuracy on 200 SME-labelled cases; latency < 5s; security review started
Pilot	30 users × 4 weeks; 25% handle-time reduction; no critical incidents; > 60% weekly active
Production	SLAs met; runbooks; customer team trained; monitoring and alerts live

Key insight: most AI projects die between POC and production. The work there is security reviews, integrations, edge cases, change management and trust. Strong FDEs plan for that from day one.

PreviousWhat is a Forward Deployed Engineer (FDE), and how is the role different from a software engineer, solutions engineer, or consultant?

You understood something today that you didn't yesterday.

What does a typical FDE engagement lifecycle look like for an AI project?

Exit criteria examples

Related