A Forward Deployed Engineer is a hybrid role: fluent in the business, fluent in AI, able to design systems, able to build with their own hands — and accountable for shipping to production with measurable impact.
An FDE (Forward Deployed Engineer) is a hybrid role that emerged with enterprise AI adoption: diagnose problems in the field (Discovery), scope and design the technical solution (Scoping & Design), build it hands-on (Build), and drive it into production where it is actually adopted (Rollout & Adoption). It is neither an engineer who only writes code, nor a product manager who only writes documents.
Based on OpenAI's published FDE role: five stages covering the full path from discovery to launch.
Work on-site with the customer to diagnose problems truly worth solving with AI, instead of accepting secondhand requirements.
Assess feasibility, data readiness and constraints; turn a vague ask into an executable technical scope.
Design the architecture: models, RAG, tool calling, workflows, and integration with existing systems.
Write the code yourself — prototype through productionized implementation — instead of handing the plan to someone else.
Drive launch, training and adoption until the system runs continuously inside real workflows.
Success is not a flashy demo — it is how many real users rely on the system in daily work.
Did the workflow actually change: time saved, errors reduced, capacity unlocked.
Measure quality continuously with eval sets and data, instead of tuning prompts by feel.
Sources: OpenAI FDE role description, OpenAI Deployment Company announcement
FDE is not a rename of any existing role — it combines the critical parts of several.
| Role | Core focus | Typical output | Relationship to FDE |
|---|---|---|---|
| Product Manager | Requirements, value, priorities | PRDs, roadmaps, reviews | Knows the business but does not own hands-on building or production rollout |
| AI / Backend Engineer | Implementation, system quality | Code, services, model capability | Can build, but usually does not own discovery or delivery |
| Solution Architect | Solution design, client communication | Architecture diagrams, proposals, POCs | Designs solutions but rarely builds and follows through to production |
| Project / Delivery Manager | Schedule, resources, risk, stakeholders | Plans, milestones, acceptance | Keeps delivery on track but does not own technical design or system building |
| FDE | The full chain from problem to production value | A live AI system plus measurable business results | Owns discovery, design, building and delivery at once |
We break FDE capability into eight assessable dimensions — the same scoring framework used by the readiness quiz.
Interviews, scenario breakdown, prioritization — diagnosing problems worth solving.
User value, business value, ROI and workflow redesign.
Practical command of agent loops, global settings, RAG and tool calling — and their limits.
Overall design across APIs, data, model routing, the harness and observability.
Write code independently, integrate APIs and MCP, ship real projects with CLIs and AI tools.
Test sets, rubrics, accuracy measurement and regression protection.
Reliability, security and permissions, cost, latency and audit.
Stakeholder management, pilot-to-rollout, adoption and change management.
Start with self-assessment, then move toward learning roadmaps, architecture design and enterprise delivery practice.
8 dimensions, 16 questions: locate your strengths and gaps, get a 30/60/90-day action plan.
Available nowA 90-day transition path based on your background and current capability: 3 phases, 12 weeks, with deliverables and acceptance criteria.
Available nowDecide whether a need calls for a prompt, a workflow, RAG or an agent — with an architecture proposal.
PlannedA 12-step delivery method for enterprise agents, from Discovery to Production.
PlannedThe questions people ask most about FDE and this quiz.
Not sure where you stand with FDE? 8 dimensions, 16 questions, 2 minutes — get your capability profile and a 90-day action plan.