When AI writes code at scale, FDEs are paid to own the outcome
An FDE Times analysis of Salesforce’s Agentforce FDE listing and OpenAI FDE head Colin Jarvis argues that accountability, not typing speed, is the job: Jarvis puts model capability at about 20% or less of the deployment gap.
Salesforce’s Forward Deployed Engineer listing for Agentforce and Data Cloud says the hire is accountable for deliverable quality whether the code came from the engineer or from an AI tool. FDE Times reads that as a shift in what the role is paid for: not who typed the code, but who answers for it in production.
OpenAI’s head of forward deployed engineering, Colin Jarvis, estimates that model capability is only about 20% or less of the gap between a demo and a live system. The rest is organisation, data, access and evaluation — work that coding assistants shorten only at the edges. Anthropic’s Applied AI FDE posting points the same way: MCP servers, sub-agents and agent skills encode permission and access decisions the model will not make on its own.
The piece also flags a risk first raised by Constellation Research’s Larry Dignan: FDEs can become a stopgap for immature products. Jarvis and Anthropic both push the other model — leave capability with the customer, and carry field lessons back into the product — so the role compounds instead of becoming permanent consulting headcount.
For candidates, FDE Times advises rewriting CVs around owned production outcomes and risks blocked in AI-generated code, and looking for phrases like “owns the outcome” and “architectural judgment” in job ads.
Source: fdetimes.net