Skip to main content
Aimeice Tech
AI AgentsAI AdoptionEngineering TeamsForward Deployed Engineer

Forward Deployed Engineer: When You Need One and What You Get

A forward deployed engineer builds AI inside your business, not beside it. When the model pays off, when it doesn't, and what our clients got from it.

Andrii Nasadchuk, Founder & CEO · November 16, 2026 · 7 min read

Until 2016, Palantir employed more forward deployed engineers than regular software engineers. The company that invented the role bet that the hard part of software was not writing it, but making it work inside one specific customer's mess. Ten years later that bet looks obvious: the forward deployed engineer is one of the most requested roles in AI, OpenAI built its own forward deployed team, and a16z calls it "the hottest job in startups".

The reason is simple. Models are now good enough for most business tasks. What fails is the last mile: your data lives in six systems, your process has exceptions nobody wrote down, and the people who know them are busy. This post explains what a forward deployed engineer does, when the model is worth paying for, when it is not, and what it produced on our own client projects.

What a forward deployed engineer actually does

A forward deployed engineer (FDE) is a senior engineer who works inside the customer's business instead of behind a ticket queue. They sit with the people doing the work, find where the time and money leak, and write production code against the customer's real systems until a number moves.

Three things separate the role from an ordinary developer:

  • They own the outcome, not the task. The brief is "cut lead response time" or "launch more ad tests", not "build endpoint X".
  • They do their own discovery. No analyst hands them a spec. They watch the process, ask the awkward questions, and write the spec themselves.
  • They ship into your environment. Your CRM, your ad accounts, your permissions, your compliance rules. Not a demo on clean sample data.

That last point is why the role exploded with AI. In our field report on AI adoption in software teams we found that most companies are far closer to a proof of concept than to anything running in production. The gap between the two is almost never the model. It is integration, edge cases and trust, which is exactly the work an FDE does.

Forward deployed engineer vs consultant vs staff augmentation

These three get confused because all of them put outside people next to your team. They sell different things.

Model What you buy Who defines the work What is left at the end
Consultant Advice The consultant A roadmap someone else must build
Staff augmentation Hours You Whatever you managed to specify
Forward deployed engineer A result Both, around an agreed number A working system in your stack

Staff augmentation works when you already know exactly what to build and only lack hands. Consulting works when you need a decision, not software. The forward deployed model fits the case in between, which is where most AI projects live: you know the business problem, but nobody can write the technical spec until someone has dug through the real data.

When you need a forward deployed engineer (and when you don't)

From our projects, the model pays off in five situations:

  1. Your AI pilot works in the demo and nowhere else. The prototype handles the happy path. Production needs the other 40 cases.
  2. The knowledge is in people's heads. If the process is not documented, an engineer has to sit next to the people who run it. No spec will capture it.
  3. The data is spread across tools that disagree. Ad platforms, CRM and analytics each define a "conversion" differently. Someone has to reconcile them before any agent can act on them.
  4. You have no in-house AI team and no time to hire one. A senior AI engineer takes months to find. An embedded team starts this month.
  5. The cost of being wrong is real. Money, medical data, client ad budgets. You need approval steps, spend caps and an audit log designed in from day one, by someone who understands the business rules.

And when you don't need one:

  • The task is standard and an off-the-shelf tool already does it. Buy the tool.
  • You have a clear spec and a capable team. Add hands, not an embedded engineer.
  • Nobody on your side can spend a few hours a week answering questions and approving decisions. The model depends on access to the people who know the process. Without it, an FDE is just an expensive remote developer.

What this looked like on our projects

We did not call it "forward deployed" when we started. We just noticed that the projects that worked all began the same way: an engineer next to the operator, not a requirements document.

50 → 1,000
ads launched per day, same team (Dolead)
8 → 1
marketing tools per merchant (Agency AI)
30 sec
from insight to live change, down from a week

Dolead. A performance marketing agency serving 200+ lead-driven companies. A media buyer could build about 50 ads a day by hand, so the number of tests was capped by team hours, not by budget. We built the layer that assembles, launches and tunes ads through the platform APIs. The same team now launches 1,000 a day. The part that only came from working inside their operation: spend caps, approval steps and a full audit log on every automated change, because an agency paid per accepted lead cannot afford an unexplained invoice. Full story in the Dolead case study.

Agency AI. Before writing code we sat alongside two Shopify operators for a week. Every "why did this happen?" they asked meant logging into three dashboards. That observation became the product brief: merchants did not need more data, they needed one surface that had already reconciled it and could explain its recommendation. The result replaced eight tools with one engine and cut the path from insight to a live change from a week to 30 seconds. See the Agency AI case study.

Clevergig. We shadowed three operators: a restaurant group, a stadium events team and a festival organizer. The finding no spec mentioned: 20% of shifts change at the last minute, and if software cannot handle a cancellation in under five minutes, people go back to WhatsApp and spreadsheets. That became the design constraint for the whole scheduling engine. Details in the Clevergig case study.

How we work with clients in the forward deployed model

Our engagement is built around this role. The full version is on the how it works page. The short one:

  1. A 30-minute call, then a written plan within one business day. Which agent first, what number it moves, what it costs.
  2. First agent live in 14 days. We start where money shows fastest, usually marketing: SEO, AI search or ads. The first month runs as a pilot.
  3. One agent at a time, each with its own number. Lead follow-up, booking, reporting, quality control. The next one starts when the previous one pays.
  4. A monthly report. Hours worked, accuracy, uptime, revenue per channel. We answer for the first three.
  5. Everything stays yours. Code, prompts, data and dashboards, with a documented handover. You can stop with 30 days notice.

On your side we need one person who can approve decisions and answer questions. No IT department required. Retainers start from $3k a month, and the tiers are on the pricing page. If you want to see what we would build, start with our AI agent development overview.

What to ask before you hire a forward deployed team

The title is popular now, so plenty of ordinary outsourcing is being relabeled. Four questions filter most of it out:

  • What system did you put into production, and what number did it move? A real answer names the system and the before and after.
  • Who did the discovery? If a separate analyst wrote the spec and engineers never met the users, it is not forward deployed.
  • What happens when the model is wrong? Look for approval steps, limits, logs and monitoring. "It is very accurate" is not an answer.
  • What do we keep if we stop? Code, prompts and data should be yours, with a handover, in writing.

Have an AI pilot stuck before production?

Tell us what should grow next. Within one business day you get a written plan: which agent first, what number it moves, what it costs.

Book a 30-min call

Not ready to talk? Run the free growth audit first.

Written by

Andrii Nasadchuk

Founder & CEO

Product engineer. Leads strategy, sales and delivery.

Start here

Tell us what should grow next

Send what you sell and where clients come from now. Within one business day: which agent first, what number it moves, what it costs.

Or run the growth audit first. No email needed to see your score.