Forward Deployed Engineers have become one of the more awkward hiring problems in AI.

The demand is real enough. Research shared with TechCrunch by executive search firm Christian & Timbers found that the share of companies planning to hire FDEs had risen from roughly 5–10% at the start of this year to 70% by the end of Q2. Some large consulting firms were apparently planning to increase their FDE headcount tenfold.

The supply side is less straightforward. Christian & Timbers estimates there are around 17,000 people in the US market who could reasonably be described as FDEs, but only around 2,000 with the particular mix of applied AI experience, industry knowledge, technical depth and seniority that it believes can consistently produce serious enterprise returns. That 2,000 figure is a recruiter’s estimate rather than an objective census, but the distinction it is trying to make is useful. There are plenty of people with some of the ingredients. There are far fewer with all of them.

At first glance this looks like a conventional talent shortage. Lots of demand, not enough candidates, salaries go up, recruiters get busy. I think the problem is stranger than that.

Companies are trying to hire a role before the industry has agreed where that role begins or ends.

The job market exists. The job description doesn’t.

One recent study scanned nearly 240,000 postings and found 5,426 FDE and FDE-adjacent roles across 1,311 employers. Whatever else is true, this is no longer a tiny Palantir curiosity. There is now a substantial market of companies trying to hire people to work somewhere between engineering, product, deployment and the customer.

The problem appears when you look at what those companies actually call the people they want.

Adam Boudjemaa analysed 519 roles across 57 employers that belonged to the broader forward-deployed family. Only 36% actually carried the FDE title. The rest were scattered across Solutions Architect, Solutions Engineer, FDSE, Deployment Engineer, Deployment Strategist, Applied AI Engineer, Field Engineer, Customer Engineer and a handful of others.

Even the roles explicitly called FDE were inconsistent. Under Boudjemaa’s broad test, 28% of FDE or FDSE-labelled postings didn’t contain enough evidence to qualify as forward-deployed work. Under a stricter definition requiring an engineering or product reporting line, only 60.4% passed. Nearly all of the wider role family required production coding, but the organisational homes varied across engineering, product, services, delivery and sales.

This creates an odd situation. Pave now considers Forward Deployed Engineering established enough to deserve its own compensation benchmark rather than being priced as a variation of software engineering or solutions engineering. The compensation market has decided the role exists. The hiring market is still arguing about what it is.

That makes searching for an FDE by title a fairly unreliable way of finding one.

Stop looking for people called FDEs

If you wanted to hire a backend engineer, looking for people who have already been backend engineers would be a reasonable starting point. The previous title gives you useful information about the work they have probably done.

FDE is different because most of the interesting characteristics of the role sit between established functions.

You need someone technical enough to build production software, but interested enough in the customer to spend time understanding why it should exist. They need to be comfortable with ambiguity, but not so comfortable with it that nothing ever gets shipped. They have to form opinions quickly while being willing to throw those opinions away when the evidence changes. They need enough commercial sense to understand which problems matter, without turning every technical decision into an account-management exercise.

There aren’t many conventional career ladders designed to produce that combination.

A very good candidate might currently be a software engineer at a startup who keeps getting dragged into customer calls because they are the only engineer who can translate a messy complaint into something buildable. Another might be a solutions engineer who has gradually crossed the line from configuring the product into changing the product itself. Another might be an applied AI engineer who has spent the last two years inside one industry and understands its workflows better than most product managers.

None of them necessarily have FDE on their CV.

This is why the market can simultaneously contain thousands of FDE roles and apparently very few people who feel like obvious FDE hires. Employers are searching for a mature job category when much of the talent is still hiding inside adjacent ones.

The Palantir problem

There is an obvious shortcut: hire people who have already worked somewhere that knows how to produce FDEs.

Palantir sits at the centre of that market for a reason. It spent years selecting technically strong generalists, putting them inside difficult customer environments and giving them unusually broad responsibility. If you want someone who already understands forward deployment, an ex-Palantir FDE is an unusually legible candidate.

But that shortcut does not scale particularly well.

If every AI company decides that the safest way to hire FDEs is to recruit from the handful of companies that already employ them, the industry hasn’t solved the talent problem. It has created an increasingly expensive game of musical chairs.

There is also a danger in confusing pedigree with capability. The whole attraction of an FDE is supposed to be an ability to reason from the problem in front of them rather than relying on a neat playbook. Hiring only from a tiny list of recognised FDE organisations would be a strange way to build a profession whose defining characteristic is adaptability.

The better question is not where did this person learn to be an FDE?

It is what have they already done that looks like forward deployment, even if nobody called it that?

Look for the scar tissue

The strongest signal may be a history of owning problems that didn’t fit comfortably inside one function.

Has this person ever walked into a situation where the customer didn’t really know what they needed? Did they have to work that out themselves? Could they then build enough of the solution to test the idea rather than handing a requirements document to somebody else?

What happened when their first interpretation was wrong?

Did they stay around for the ugly part after the demo, when the integration broke, the data was worse than expected, security objected, the user behaved differently from the person who bought the system, or the supposedly universal workflow turned out to have twelve exceptions?

Those experiences matter because forward deployment is full of situations where the problem changes as you understand it.

A candidate who has spent three years shipping technically excellent work against clearly specified tickets may be a superb engineer and still hate FDE. Someone with slightly less pristine engineering credentials who has repeatedly taken ambiguous projects from a confused stakeholder through to something running in production may be much closer to what you actually need.

The CV tends to capture technologies and titles. FDE ability often lives in the stories between them.

Hiring for a combination is harder than hiring for a skill

This is also why the usual technical interview can miss the point.

Boudjemaa’s dataset suggests production coding really is a near-universal expectation across the role family: 98.3% of the postings he studied required it. This is not a euphemism for sending engineers into sales meetings. The engineering bar matters.

But passing a hard coding interview only verifies one component of the job.

An FDE can be a brilliant programmer and still repeatedly build the wrong thing. They can be charismatic with customers and lack the depth to know when a technically convenient answer is dangerous. They can understand the domain beautifully and fail to move quickly enough when the information is incomplete.

The skill is in the combination.

That suggests a different type of hiring process. Give someone a messy situation rather than a clean specification. Let the customer be wrong about what the problem is. Introduce a technical constraint halfway through. See whether the candidate asks sensible questions before building anything. Let them use AI. Let them write code if it helps. What matters is whether they can keep forming and testing a view of what is actually going on.

A good FDE interview should probably feel a little like doing FDE work.

There may never be enough ready-made FDEs

The current talent shortage has led to some extraordinary responses. Anthropic is committing $100 million to train 10,000 Frontier Deployed Engineers. AWS is putting billions behind forward deployment and building the capability into its partners. OpenAI has acquired experienced deployment talent to seed an entire Deployment Company.

That investment makes sense if you assume there is a finite pool of people called FDEs and everyone simply needs more of them.

I suspect the larger opportunity is to get better at recognising people who could become one.

There is a useful comparison with product management twenty years ago. Companies eventually stopped expecting every good product manager to arrive from another company with five years of product management experience, partly because there weren’t enough of them and partly because they noticed good candidates coming from engineering, design, consulting, operations and founding companies themselves.

FDE may go through something similar.

The first generation is easy to identify because it largely comes from organisations that already use the title. The next generation will probably be less tidy. Some will come from software engineering. Some from AI engineering. Some from solutions roles. Some will be founders who discover they prefer the customer problem to running the company. Some will be people whose current organisation has no name at all for the work they are already doing.

That is not a flaw in the category. It may be the only way it gets big enough to satisfy the demand now appearing around it.

Maybe the shortage is partly a search problem

Christian & Timbers may be right that only a couple of thousand people in the US currently have the complete package required to operate at the very top end of enterprise AI deployment. The combination is genuinely rare, and calling everybody with customer exposure and a GitHub account an FDE won’t change that.

But the number of people capable of growing into the role is almost certainly larger than the number whose LinkedIn headline already contains the words Forward Deployed Engineer.

That distinction matters.

If companies respond to the FDE boom by competing for the same tiny group of obvious candidates, salaries will rise and the same people will move between the same companies. If they get better at identifying the underlying behaviours — technical depth, curiosity, customer judgement, ownership and a habit of following problems past organisational boundaries — the talent pool begins to look different.

FDE is becoming a recognised profession unusually quickly. Pave can benchmark its compensation. Thousands of jobs are being advertised. AI companies are building entire organisations around it.

The hiring machinery is catching up more slowly.

Everyone knows they want an FDE.

The companies that win the talent race may be the ones that learn to recognise one before the title appears on the CV.