INSIGHTS
What a forward deployed engineer does
A forward deployed engineer is useful when they sit with a job that already exists, build in the live stack, and leave a draft a named person can approve. They are not useful as a hiring buzzword or a platform demo.
The title is new. Palantir coined it. Other AI vendors copied it. The work is older: put an engineer next to the problem, not in a backlog two buildings away.
Most pages on this job are written for people applying at those firms. This page is for the owner who would pay for the work, and for anyone who wants a plain description of what a forward deployed engineer does. We work as an AI consultant for South African SMEs. The same last-mile shape shows up on a factory quoting desk.
What is a forward deployed engineer
A forward deployed engineer (FDE) is an engineer who embeds with one customer. They map the real workflow, write software against the systems that already run, and stay until the thing works in production. They are not a demo specialist. They are not a remote team taking tickets.
The name is military. Forward deployed means close to the action. In software it means close to the customer’s data, people, and constraints. One customer, many capabilities: invoices today, a quote draft tomorrow, a job-status digest if that is where the hours sit.
What a forward deployed engineer does
Day to day the job is four loops.
- Sit with the people who already do the job. Watch the invoice pile, the quote rebuild, the “where is the job” calls.
- Write a first version in the stack that already exists: spreadsheet, email, job card, ledger. Do not start by replacing those.
- Put a person on the approval. The engineer does not become the new owner of the process.
- Leave something timed. If they work for a platform company, they also send what they learned back to a product team. If they work with an SME, the test is whether hours came off the job.
That last loop is the one most job ads skip. A prototype that dies when the visitor leaves is a longer demo. A workflow that still runs next month is the job.
What this looks like on a factory floor
This is the part most explainers skip. The role is not a slide about “enterprise AI.” It is the same factory admin jobs we already write about.
Invoice capture
PDFs arrive, someone retypes them, month-end waits. The engineer reads the file, drafts the capture line, and flags a mismatch. A person still posts. See invoice automation.
Quoting
The estimator rebuilds last month’s bill of materials. The engineer drafts from prior jobs and the price list. The estimator still sets margin.
“Where is the job?”
The owner still calls because the answer lives in WhatsApp, a whiteboard, and a spreadsheet. The engineer assembles job, stage, and blocker from the same places the team already types them. The test is whether those calls stop.
That list is the same as the AI automation examples and the rule for AI agents for small businesses: one repeating job, a named owner, a draft they can approve. The forward deployed engineer is the person who times that job and builds the draft. The software brand is chosen after.
Forward deployed engineer vs software engineer
A software engineer builds a feature for many customers. A forward deployed engineer builds until one customer can run the job. Scale is a later question. First contact with the messy stack is the job.
A solutions engineer or sales engineer shows what a product could do, often on sample data. An FDE ships in the live environment: the actual ledger, the actual job cards, the actual inbox.
A consultant can look similar. The difference that matters for an SME is whether someone still owns approval after they leave, and whether hours were counted. If the work dies when the visitor leaves, it was a demo with a longer stay.
When the role is worth paying for
Pay for this shape of work when all four are true.
- One job repeats every week or month.
- You can name the person who will approve the draft.
- The hours are real, not a guess from a vendor slide.
- The stack is messy, and a new portal would make the first month worse.
Do not pay for it when you want a chatbot on the homepage, a new finance system, or “AI transformation” with no timed job. That is the same path as the offer on the home page: time one job, then take a scoped step.
Take the free scorecard, or book a 20-minute fit call. We will talk hours, not a job description.