The Skills That Make a Great Forward Deployed Engineer (Only Half Are Technical)

Ask people what a forward deployed engineer needs and you'll hear "strong engineer who can talk to customers". True, and uselessly vague — like saying a great chef needs "cooking plus taste". Having watched the role explode across AI hiring (background here), and knowing the terrain from the deployment side of the industry, here's the actual stack — six skills, three technical, three not, ranked by how often their absence sinks an engagement.

The Technical Half

1. Glue engineering

Not algorithms — integration. APIs, webhooks, data pipelines, authentication, the customer's ancient database, the AI platform's SDK, and the unglamorous code that makes them hold hands reliably. FDE codebases are mostly glue, and glue quality decides whether the deployment survives contact with Monday. Related and underrated: data hygiene, because customer data is always worse than promised (clean data in, better AI out is practically the FDE motto).

2. AI-layer fluency

Knowing what models genuinely do — next-word prediction and its consequences (the ten-minute version) — so you design around hallucination instead of being surprised by it. Structured prompting as engineering discipline, not vibes (the anatomy). Evaluation: proving to a sceptical customer that the system is right often enough, with numbers.

3. Scrappy shipping

The FDE ships a working version in weeks inside someone else's infrastructure, then iterates. That's a distinct skill from building beautifully at your own pace — it's ruthless scoping, sensible corner-cutting, and knowing which corners are load-bearing. Portfolio projects that demonstrate this beat any credential (how to build that portfolio).

The Non-Technical Half (Where Engagements Actually Die)

4. Workflow discovery

The defining FDE skill. Before anything gets built, someone has to extract how work actually flows — the real triggers, handoffs, exceptions and blockers, which never match the official process. It's interviewing, observing, and drawing the map that makes the mess legible. You can practise the mechanics today: run your own role (or a friend's business) through our free Workflow Mapper, then verify the flowchart against reality — the gap between the two is exactly what FDE discovery hunts for. The deeper craft is in our guides on mapping before automating and finding hidden time drains — both are FDE thinking in civilian clothes.

5. Translation

An FDE spends the day code-switching: model limitations to a COO, business constraints to the product team back home, timeline reality to an excited sponsor, change management to a wary end user whose job just changed shape. Every failed AI deployment we hear about includes a translation failure somewhere. The test: can you explain the same system, honestly, in three registers without slides?

6. Judgement under ambiguity

Whose data can touch which system. What gets verified before the customer sees it. When to say "AI is the wrong tool for this step" — the credibility-defining moment of most engagements. This is the layer we call the missing AI skill, and the protect → verify → trust sequence is its pocket version. FDEs live it because they operate inside someone else's risk envelope, usually without adult supervision.

The Honest Ratio

Interviews overweight the technical half because it's easy to test. Outcomes overweight the other half: a middling engineer with excellent discovery and translation ships successful deployments; a brilliant engineer without them ships elegant systems nobody adopts. If you're choosing what to strengthen next, strengthen whichever half you'd rather not — the market is full of your mirror image, and the combination is what's scarce (skills stack multiplicatively).

Building the Stack Deliberately

Self-serve: the free tools linked above cover practice reps for skills 2, 4 and 6, and the AI Jobs Board shows you weekly which skills employers name. Structured: Cocoon's AI for Professionals programme drills the workflow, integration and judgement layers with real work, and Become the AI Champion is, in effect, FDE training aimed inward at your own company — same stack, deployed where you already work. Compare the adjacent roles before you commit the effort: FDE vs Solutions Engineer vs Consultant.