What Is a Forward Deployed Engineer? (And Why We Built Something Different)

A forward deployed engineer is a software engineer who works inside a customer's business instead of at their own company's office. They sit with the customer's team. They learn the workflows. Then they build and ship software that solves that customer's specific problems, using their employer's platform.

Palantir invented the role. OpenAI and Anthropic made it famous again. Every AI company now wants FDEs because AI products don't work off the shelf. Someone has to close the gap between the model and the messy real business.

That's the definition. Now let me tell you what the job actually looks like. Because I think the FDE model is half right. And the half that's wrong is the reason we built something different.

What FDEs actually do all week

One FDE at an AI startup described his week in public. It stuck with me.

Ten clients at once. Fifty hours a week. Half the time in meetings. The other half building, testing, and fixing things under pressure. Dozens of Slack workspaces. Two builds running at the same time. Permanent firefighting.

His sharpest line: the job is less about building AI systems and more about understanding businesses well enough to know what to build.

He's right about that last part. Dead right.

But look at the model. One human, ten clients, fifty hours. That's not engineering. That's triage. Every client gets a slice of a tired person. And the person burns out.

Why companies want one anyway

Because the alternative is worse.

Most companies bought AI tools. Some ran pilots. Very few have systems running in production that their own teams use every day. The gap between "we tested it" and "we run on it" is where all the value dies.

An FDE closes that gap by force. They sit inside your business until the thing works.

Here's the problem nobody says out loud. A classic FDE walks in cold. They spend their first months learning your business from zero. You pay for that learning curve. Every time.

The last mile problem

We have trained more than 2,500 people. Our programs end with teams building real tools. Not slideware. Working prototypes that won a competition in front of their executives.

Then the program ends. And I'll be honest about what happens next.

At a 450-person software company, the cohorts built dozens of tools. One cut training recaps from 30 minutes to 30 seconds. At an 89-person energy company, the team built 5 tools worth at least $220,000 a year. The president built one himself, next to his team. At Source Capital, 123 leaders went through three cohorts, from the firm's own team out to its portfolio and then to peer PE firms.

Some of those tools made it into daily use. Some didn't. The ones that didn't had nothing wrong with them. They just never got the last mile: security review, integrations, error handling, a home in the company's systems.

That last mile is exactly the work an FDE does. But hiring a cold FDE to ship tools your own team already built makes no sense. The FDE would spend months learning what your team already knows.

The Operator in Residence

So we flipped the model. We call it an Operator in Residence.

The difference comes down to three things.

They don't start cold. An Operator in Residence only goes into companies that trained with us first. The team already knows the method. The tools already exist as prototypes. Day one is build day, not discovery day.

Agents do the volume. The FDE burning fifty hours across ten clients is doing the building by hand. Our operator runs a stack of AI agents that do the heavy lifting. The human is there for trust, judgment, and the conversations that matter. Three clients, not ten. Build days, not firefights.

Your team runs it after. A classic FDE creates dependency on the vendor's platform. Our rule is the opposite. Everything we ship is documented and runs without us. If only we can operate it, we failed. You keep us because the backlog keeps producing value, not because you're trapped.

An operator inside your business on a fixed scope. Agents building in between. Your people operating everything we ship.

FDE vs Operator in Residence

• Starting point. Forward Deployed Engineer: Cold. Months of discovery. Operator in Residence: Trained team, existing prototypes.
• Load. Forward Deployed Engineer: Up to 10 clients, 50-hour weeks. Operator in Residence: A few clients, agent-leveraged.
• Builds with. Forward Deployed Engineer: Their hands. Operator in Residence: An agent stack plus their hands.
• Goal. Forward Deployed Engineer: Deploy the vendor's platform. Operator in Residence: Ship your team's backlog.
• After they leave. Forward Deployed Engineer: Dependency on the vendor. Operator in Residence: Your team runs everything.
• Sold as. Forward Deployed Engineer: A person's time. Operator in Residence: Shipped tools.

Who this is for

If your team went through an AI training program and built things that never shipped, that backlog is money sitting still. The energy company had at least $220,000 a year in it. The software company had dozens of tools and no owner for the last mile.

If you're hiring an FDE to "figure out AI" from zero, stop. Train the team first. It's cheaper, it spreads, and the prototypes your own people build will surprise you. At Source Capital, trained teams cut deal screening from a couple of days to about 7 hours. Their people built that. Nobody deployed it to them.

Forward deployed engineer FAQs

Is a forward deployed engineer the same as a consultant?

No. Consultants advise and leave a deck. FDEs build and ship working software inside your business.

How much does a forward deployed engineer cost?

It depends on the vendor and the load. What matters more is what you pay for. An FDE is sold as a person's time. An Operator in Residence is priced against the tools that ship, because agents carry the build volume and the discovery phase does not exist.

Do I need to be technical to work with one?

No. The whole point is that your existing team, trained on the method, operates everything after it ships.

What does "forward deployed" mean?

It's a military term. Forward deployed units are stationed where the action is, not back at base. A forward deployed engineer works at the customer, not at headquarters.

Where to start

Software ships when someone sits inside the business and refuses to leave until it works. The FDE model understood that.

That someone does not have to be a stranger doing it by hand.

Train the team. Keep the prototypes. Put an Operator in Residence on the last mile.

If your team trained with us, the first step is free. Send me the list of what your cohort built and I will tell you which three or four are worth shipping and what is sitting frozen in them. That is the Backlog Audit, and it is the way in.