What a Forward Deployed Engineer is and why a small business should care

A Forward Deployed Engineer, or FDE, works close to a customer and gets a technical solution into use within the company. I work in a similar format with small and mid-size businesses. This article explains the work involved and how to arrange it without a separate staff position.

Where the role came from

Palantir invented the format. Instead of selling software and waiting for the customer to roll it out, the company sent engineers to the customer: to sit in their office, understand their data and build the solution on site. That engineer was called forward deployed, borrowing the military term.

In 2024 the AI labs adopted the model. OpenAI, Anthropic, then Microsoft, Meta, Google, Nvidia and Scale AI started hiring FDEs because they ran into the same problem. The model can do a lot. The customer buys access and six months later discovers that nothing has changed in actual operations: the data is in the wrong place, the process is undocumented, nobody knows how to check quality.

That gap between what a model can do and what a company needs in production is called the deployment gap. The FDE exists to close it.

The market and pay

I covered recent Lightcast data on the role and an example of a published salary range in my AI job market overview for 2026. It includes data periods and the difference between base salary and a package with equity. For an implementation budget, the scope of your task matters more than compensation at a large lab.

For the distinction from building a shared AI feature, see the separate AI Engineer vs FDE comparison.

What an FDE actually does on site

Behind the loud title is fairly down-to-earth work. The sequence goes roughly like this.

1. 01Maps the process. Sits next to the people who do the work and writes down how it really runs, not how the policy says it runs. Where the inputs come from, who makes decisions, what counts as an error.
2. 02Connects the data. CRM, email, spreadsheets, documents, telephony. Usually it turns out that half of what is needed lives in personal files and message threads.
3. 03Builds the first version with the team. Not a slide deck but a working tool on real data, used by the same people who took part in the mapping.
4. 04Measures. Before and after: time per task, error rate, how many cases went to a person for review.
5. 05Hands over. Documentation, training, a plain answer to "what do we do if it breaks".

The main difference from a regular developer is that an FDE is accountable for the result inside the customer's process, not for the code.

Why this matters for a small business

A small company can also have scattered data and a process that is difficult to transfer to a new system. Enquiries might arrive through several channels while a manager keeps the rules for the next step in their head. Someone first needs to understand that work and agree on what can be automated.

A limited task may not justify a separate staff position. An engineer and consultant can take on the work in stages, with responsibility for implementation and support after launch.

How to get FDE-style work without paying lab prices

Do not hire. Buy it stage by stage. The format I use with small and mid-size businesses looks like this.

Audit. From 450 EUR. One or two days of review: which processes exist, where the data is, what will pay off first. The output is a list of two or three candidates with an estimate, and an honest "AI is not needed here yet" if that is the case.

One process. From 1,500 EUR. An AI agent for a single task: triaging incoming requests, drafting proposals, first-pass lead qualification, answering from documents. First working version in two weeks, then measurement on real data.

Handover. Documentation, training for the people who will use it, an agreement on support. The goal is a system that runs without me.

I do for small and mid-size businesses what an FDE does for the labs' corporate customers: I come into the process rather than sell a license. All three services are described on the home page.

What to ask a candidate or a contractor

Whether you are hiring an FDE on staff or engaging a contractor, the questions are the same.

• How will you measure quality? A good answer contains the words "evaluation set" and specific metrics. In the 2026 KORE1 survey, employers name building evaluation systems as the number one skill for agentic AI engineers.
• What guardrails will the system have? What it may not do without human confirmation: send, pay, delete, change access.
• What documentation do I get at handover? If the answer is "you will figure it out from the code", that is not an FDE.
• Who owns the data and the prompts? The answer must be "you", with access to everything produced during the project.

If a contractor cannot answer these four questions in the first conversation, I would not start.

Discuss evaluation, limits and handover before signing a contract. That gives the project concrete acceptance criteria and helps establish the scope of work.

If you have a process that eats your team's time and where you suspect AI could help, send me a short brief: what the process is, who is involved, where the data lives. I will tell you where to start.

FAQ: Forward Deployed Engineer

How is an FDE different from a solutions engineer?

Boundaries depend on the company. A solutions engineer often works on applying an existing product. In the FDE format, the engineer also investigates the customer process, writes the code needed and owns implementation. Read the responsibilities in the particular role description.

How much does a Forward Deployed Engineer earn?

Pay depends on location, employer and seniority. A checked vacancy example and an explanation of base salary and total compensation are in the AI job market overview. Project implementation for a small business is priced separately.

Does a small business need its own FDE?

On staff, almost never. But a small business does need FDE work: map the process, connect the data, build the first version, measure and hand over. It makes sense to buy it in stages from a contractor: an audit, then one process, then handover.

What is the deployment gap?

The distance between what a model can do in a demo and what a company needs in real operations: data in the right place, a documented process, quality checks, limits on actions. That gap is why the AI labs started hiring FDEs on the Palantir model in 2024.

What does it cost to put AI into one process?

In my format an AI readiness audit starts from 450 EUR, an AI agent for one process from 1,500 EUR, with the first working version in two weeks. The final price depends on the number of data sources and the evaluation requirements.