01 Find the problem behind the request
“Start with the outcome, do what it takes to succeed, then follow with scale” (opens the source)
Name the operational outcome before choosing the AI feature.
FDE interviews when AI makes the prototype easy
Turn an unclear customer request into a system people use and can own.
Start with the customerPractice interviews
02 Resolve the data before the demo
A plausible answer built on the wrong data grain is still wrong.
03 Make the first release small and testable
“Treat governance as part of the architecture.” (opens the source)
A small, permitted release beats a broad promise.
04 Measure useful work, not curiosity
“Curiosity drives initial adoption. Value drives sustained usage.” (opens the source)
Count completed work and rework, not just sign-ins.
05 Make the customer less dependent on you
“Identify and codify repeatable deployment patterns” (opens the source)
Prove that someone else can operate and recover the system.
Sources
- Colin Jarvis, Head of Forward Deployed Engineering, at the time. OpenAI, “What FDE means at OpenAI”, Oct 22, 2026. Open the source. Checked against the published text.
- Elisa, Forward Deployed Software Engineer, New York, at the time. Palantir Blog, “Dev versus Delta: Demystifying engineering roles at Palantir”, Apr 8, 2019. Open the source. Checked against the published text.
- Atlassian, FDE field playbook, at the time. Inside Atlassian, “Atlassian’s Forward Deployed Engineering Playbook”, Oct 4, 2026. Open the source. Checked against the published text.
- Atlassian, FDE field playbook, at the time. Inside Atlassian, “Atlassian’s Forward Deployed Engineering Playbook”, Oct 4, 2026. Open the source. Checked against the published text.
- Anthropic, Munich FDE role, at the time. Anthropic Careers, “Forward Deployed Engineer — Munich”, Undated. Open the source. Checked against the published text.
Last reviewed Nov 10, 2026