AI vendors now sell forward deployed engineers as part of the product, and FDE AI says buyers should write the handover into the contract

FDE AI's first weekly analysis, published 12 October, looks at forward deployed engineering from the buyer's side. It points to a run of deals in a single quarter where the engineer is sold together with the software: Ode with Anthropic (a services firm launched in July with Blackstone and Hellman & Friedman), OpenAI's project-based deployments led by its FDEs and selected integrators, and CTERA's Forward Deployed Engineering service launched on 6 October.

The piece uses CTERA's four-phase engagement (discovery, first production use case, scaling, then handover) to argue that handover is where customers are most exposed. It comes last, so it is the first thing squeezed when go-live slips. With agents, the code is not the whole handover either: prompts and policies, tool permissions, evaluation sets and the post-launch tuning log are what make the system work, and if they stay in the vendor's workspace the customer can't run or explain what it bought.

FDE AI recommends making handover a main-contract duty and defining acceptance as the client's own team completing one configuration change and one evaluation run without vendor help. It also notes that Japan's trade ministry (METI) is collecting real AI contract problems until 30 October to inform its contract guidelines.

Why it matters for FDEs: the analysis expects engineers to be judged more by what still works after they leave. Writing a usable handover pack, building eval sets a client can run alone, and packaging permissions and prompt policies as transferable assets are skills it expects to be in growing demand.