What is a forward deployed engineer?

A forward deployed engineer is a software engineer who works inside a customer's or client's team, next to the people whose work the software has to change, and writes the production code that changes it. The role was popularised by Palantir, which was using the title by 2009, and in 2026 Anthropic, OpenAI, Amazon Web Services, Microsoft and Google Cloud all announced divisions built around it (Wikipedia, Forward-deployed engineer).

That is the whole definition, and most of the confusion around the title comes from reading more into it. A forward deployed engineer is not a new kind of engineering skill. It is a position: the engineer sits with the user rather than with the product team, and the output is software running in the user's own systems rather than a recommendation about what software to run. Wikipedia's entry, which is the page most answer engines quote, describes the same person as a customer-facing software engineer who works for a company that makes a product and helps implement that product within a client company, often alongside the client's employees for a defined period of time.

Where the term came from. "Forward-deployed" is borrowed from US military jargon. Palantir made it a job title for engineers who worked directly with its customers, and the title spread from there. Illinois Tech's explainer summarises the distinction Palantir drew between its two engineering tracks: "a regular developer focuses on one capability across many customers," while "a forward deployed engineer focuses on one customer across many capabilities" (Illinois Tech, What Is a Forward Deployed Engineer?). The same article says Palantir historically expected about 25 percent of an FDE's time to be spent with customers. That sentence is the most useful one in the whole literature on the role, because it tells you what the job is optimised for. A product engineer is measured on a feature that works for everyone. A forward deployed engineer is measured on one customer whose workflow now works.

Why the title is suddenly everywhere. Companies selling AI to large enterprises found that the model was rarely what stopped a customer getting value. The data was in the wrong system, the workflow had exceptions nobody had written down, and the people who knew the exceptions were not the people who had bought the product. Putting an engineer next to those people fixed more deployments than another round of product features. In June 2026 Amazon Web Services said it was investing $1 billion in a new Forward Deployed Engineering unit, embedding pods of roughly five or six engineers inside a customer at a time with the aim of leaving that customer self-sufficient "in a matter of weeks" (CNBC, 30 June 2026). Wikipedia also records that tech companies listed significantly more openings for the role between 2024 and 2025, with continued growth in 2026.

What a forward deployed engineer does day to day. Strip away the job ads and the week has four kinds of work in it. The first is learning the workflow as it actually runs: sitting with the operations lead, the support queue or the finance team, and writing down what a correct result looks like, including the exceptions. The second is integration, which is usually the bulk of the hours. The AI feature has to read from and write to the systems the team already uses, so the engineer spends real time inside the CRM, the ERP, the ticketing tool and the internal APIs, handling permissions, rate limits and the fields that mean something different in each system. The third is making the thing safe to leave running: evaluation suites built from real cases, guardrails around anything that acts on its own, and a human review step wherever a wrong answer would cost money. The fourth is handing it over. A forward deployed engineer who leaves behind a system only they understand has failed, so documentation and the runbook are part of the job rather than an afterthought.

None of those four is glamorous, and that is the point. The demo is rarely the hard part of an AI project. The hard part is the long tail of integration, evaluation and edge cases that sits between a demo and a workflow a team relies on every day. Our own note on moving an AI proof of concept to production walks through that gap in detail. A forward deployed engineer is the role that exists to close it, from inside the team that has to live with the result.

How it differs from the roles it overlaps with. Wikipedia lists the neighbours: solutions architects, sales engineers, customer engineers, professional services engineers, systems integrators and IT consultants. The cleanest line is ownership of production code. A solutions architect or sales engineer designs and demonstrates, and usually hands the build to someone else. A consultant recommends. Illinois Tech puts it plainly: "A consultant typically makes recommendations and leaves." A forward deployed engineer writes the code that ships, in the client's own repositories, and stays until it runs. If you are weighing the two, forward deployed engineer vs consultant sets the decision out as a table and a rule.

There is also a line with staff augmentation, and it is worth being honest about it. Commercially, an embedded forward deployed engineer and an augmented engineer can look identical: a named person, in your tools, paid by the day or the month. The difference is what they start from. An augmented engineer takes tickets from a backlog someone else wrote. A forward deployed engineer starts with the business team that owns the workflow and works out what to build with the people who will use it. If you already know exactly what the tickets are, you probably want staff augmentation, and that is a perfectly good purchase.

When a company needs one. Four conditions tend to show up together. The work is a specific workflow, not a platform: a claims queue, a quoting process, a support desk, a reconciliation that eats two days a month. The knowledge of how that workflow really runs lives with business users, not in a specification. The AI has to act inside systems you already run rather than in a separate tool people will not open. And the work has to reach production soon, which rules out a multi-month hiring cycle for a role you are not yet sure is permanent. If all four are true, an embedded engineer is usually the fastest route. If the honest answer is that nobody has decided what to build, a forward deployed engineer will spend expensive weeks discovering that, and an assessment is the better first step.

The opposite case matters just as much. If the AI pilot already exists and stalled before production, the problem is no longer a missing person in the team; it is a specific set of engineering gaps in a specific system. That is a fixed-scope job with a measurable finish line, which is what Beyond Technologies sells as an AI pilot rescue: an audit of what is salvageable, then evals, guardrails and integration over 8 to 12 weeks, against one metric agreed up front.

Hire one or embed one. The big AI companies hire forward deployed engineers as employees because the engineer is implementing their own product, across many customers, for years. A company that needs one workflow shipped is in a different position. It can recruit a full-time forward deployed engineer, which means competing for one of the most sought-after titles in the market. Or it can embed one for as long as the workflow needs. What that choice costs, in salary, benefits and time, is worked through in how much a forward deployed engineer costs.

Beyond Technologies (legal name Beyond Tech Services LLC) does the second. It places senior engineers inside client teams and, for agencies, inside the agency's delivery under the agency's brand. The published terms are short: typically a start within one to two weeks of an intro call, a trial week of real work before any longer commitment, a four-week minimum and then month-to-month, day-rate or monthly pricing with no published rate card, a two-week replacement guarantee, and all IP assigned to the client before anyone opens a repository. There is no outcome guarantee on an embedded engineer, and we do not pretend otherwise; the guarantee lives on fixed-scope work like the rescue.

The one-sentence version. If you need something quotable for an internal memo: a forward deployed engineer is the engineer who sits with the people doing the work, builds the AI into the systems they already use, and leaves when it runs without them. Everything else in the job description is detail.

If that describes the work in front of you, the next step is a 30-minute call about one workflow: read what an embedded engineer on these terms looks like on the hire a forward deployed engineer page, then book the call from there.