What does a forward deployed engineer do?
Short answer
A forward deployed engineer sits with the people who will use the software, watches the real workflow, and writes production code against what they observe instead of a written specification. The job covers discovery, scoping, system design, the build and the production rollout, and it continues after launch. Unlike a consultant, a forward deployed engineer still commits and deploys code.
#Where the title comes from
Palantir coined the label, and its careers page still heads the posting "The Original Forward Deployed Software Engineer". The role also appears in Palantir's first annual report. The FY2020 Form 10-K says its forward deployed engineers "have travelled to bases in Afghanistan and factories in the industrial Midwest to deploy our platforms" and that, as they help customers, they "observe users' challenges firsthand".
Current postings describe it in operational terms. OpenAI's San Francisco listing gives the engineer ownership of "discovery, technical scoping, system design, build, and production rollout" and measures success by production adoption. Anthropic's listing asks the engineer to "work within customer systems to build production applications with Claude models". Both define the job by whether people use what was built, not by whether a specification was delivered.
#How it differs from a product engineer and a consultant
| Product engineer | Consultant | Forward deployed engineer | |
|---|---|---|---|
| Where the requirement comes from | A ticket that passed research, a PRD and design review | Interviews and workshops | Watching someone do the task |
| Deliverable | A feature in a shared product | A recommendation, process map or prototype | Deployed software people log into |
| After launch | Moves to the next ticket | Engagement ends | Owns it, and hears about it from the user |
| Writes production code | Yes | Rarely | Yes |
#Why companies hire for it
AI projects fail between the demo and daily use, and several independent sources point at the same gap. Treat each figure with its caveat:
| Source | Figure | Caveat |
|---|---|---|
| Gartner, June 2025 | Over 40% of agentic AI projects predicted to be canceled by end of 2027 | A forecast, not a measurement |
| 451 Research, 2025 (704 IT decision-makers) | 24% of AI/ML projects abandoned before production, down from 31% in 2024 | Survey of US, UK and India respondents; the figure fell year on year |
| RAND, 2024 (65 interviews) | Top root cause: stakeholders misunderstand or miscommunicate the problem to be solved | Qualitative finding; the 80%+ failure rate RAND mentions is an outside estimate it cites, not its own measurement |
RAND's finding is the one that matters here. Models get built that are optimized for the wrong metric or do not fit the business workflow. Embedding an engineer next to the user attacks that cause directly.
#What it looks like in practice
At Trilliant Digital the agency has client pods that own accounts end to end and practice pods that cut across every account. Nobody could state a rule for who sees what, because the rule changed with each new account. Instead of shipping fixed roles, the permission matrix became data that a super admin can edit live, every account-scoped query goes through one function, there is a View-As mode so a lead can see the app the way their team does, and every change is audit logged.
That design came from sitting with the teams. A written specification for "build permissions" would reasonably have produced roles and a static mapping.
- The requirement is an observation, not a document.
- The unit of shipping is days, so a wrong guess is cheap and asking is cheaper.
- The engineer owns the result afterwards, and the person who complained sits two desks away.
- Part of the roadmap is a by-product of watching workarounds.
The full argument, with the SEC filing, the job postings and the failure data, is in Forward Deployed Engineer: What the Job Actually Is.
Sources
- Palantir Technologies, Forward Deployed Software Engineer posting (Lever)Fetched 2026-08-16.
- Palantir Technologies Inc., Form 10-K for FY2020 (SEC EDGAR)Filed 2021-02-26. Describes forward deployed engineers travelling to customer sites.
- OpenAI, Forward Deployed Engineer (FDE) SF posting (Ashby)Posted 2025-08-06, fetched 2026-08-16.
- Anthropic, Forward Deployed Engineer posting (Greenhouse)Updated 2026-08-12, fetched 2026-08-16.
- First Round Review, So You Want to Hire a Forward Deployed EngineerFetched 2026-08-16.
- Gartner, Over 40% of Agentic AI Projects Will Be Canceled by End of 20272025-06-25.
- 451 Research / S&P Global, AI infrastructure strategies evolve amid widespread data challenges2025-07-18. Survey of 704 IT decision-makers.
- RAND Corporation, The Root Causes of Failure for Artificial Intelligence Projects (RR-A2680-1)2024-08-13. Interviews with 65 data scientists and engineers.
Read the full write-up
Have a question that is not answered here?
Ask me directly