You might already be a forward deployed engineer.

Forward deployed engineer, or FDE, is now one of the most in-demand job titles in tech. Salaries start around $150,000 and some roles are reportedly going for as much as $1 million a year. Most of the explanations floating around are aimed at software engineers, but the more interesting story is for everyone else. In this deep dive we'll explain what an FDE actually is, and why people who can't code, but know a business inside out, may be surprisingly well positioned for this kind of work.

ELI5: What's a forward deployed engineer?

A forward deployed engineer (FDE) is a software builder who works inside a customer's business instead of back at the software company's office. They sit with the people doing the job, learn how the work really happens, then build software that fits. Like a tailor who comes to measure you at home, rather than selling you a suit off the rack.

Where the role comes from

Palantir popularised the role. Its platform pulls a company's data into one place, and Palantir sends engineers on site to customise it for each client. That model is a big part of why Palantir's average contract value sits around $4 million, compared with roughly $1.2 million at ServiceNow and $600,000 at Workday. What customers are really paying for is an outcome, rather than software licences or consulting hours.

Why the role is suddenly everywhere

Every company can now buy the same intelligence. The same frontier models are available to anyone with a credit card, so having access to AI is no longer an advantage. The advantage has moved to deployment, meaning where, how and why a company applies it.

Most companies are bad at this. A widely cited MIT study found that despite $30 to 40 billion of enterprise investment in generative AI, 95% of organisations are seeing no measurable return from their pilots. One executive reportedly burned through a $10 million AI budget in three months by simply giving everyone access, and it barely moved the needle for the business. The FDE is the person who decides where AI belongs, and just as importantly, where it doesn't.

What the job actually involves

The work splits into three phases.

• Understand the real workflow. The documented process is rarely the real one. A step like "an email arrives" turns out to mean emails from 40+ senders, none formatted alike, with data buried in PDFs, screenshots and forwarded threads. Half of them are exceptions, and the rules for handling them usually live in one person's head.

• Apply judgment. In a ten-step workflow, maybe three steps genuinely need AI. The rest can be handled with ordinary automation, which is cheaper and more reliable. A large part of the job is telling the two apart.

• Build and deploy. Construct the system on top of the tools the company already uses, keep a human approving the important steps, and log everything the AI does so people can trust it.

The part that's genuinely hard

The rare and expensive thing about FDEs is the combination of skills. They need to be strong communicators who can draw messy reality out of people, and strong enough builders to turn that into working software. People who are excellent at both are scarce, which is why the salaries are so high.

Notice where the time actually goes, though. The bulk of the job is understanding the business, handling exceptions and bringing people along. The code is the smaller half, and AI has made that half dramatically cheaper.

Why non-coders are already halfway there

If you've spent years in a finance team, an ops role or a sales function, you already have the half that can't be shortcut. You know how the work really happens, where it breaks, which exceptions matter and whose head the unwritten rules live in. In a real sense, you are already forward deployed.

The traditional approach embeds an engineer in your department for months so they can learn what you already know. There's an added friction there too, because many developers have only ever worked in engineering teams, so adapting to the messy reality of a finance or ops department takes them longer than you'd expect.

With no-code platforms and AI tools like Claude Code, the building can now go the other way. The domain expert describes workflows in plain English and lets the tools handle the technical work.

The reality check

The engineering half hasn't disappeared. Production systems need guardrails, testing, exception handling and audit trails, and a company won't trust an agent it can't inspect. Where what you build touches customer data or live business processes, a solid technical understanding is still needed to get those safeguards right.

There's also an open debate about whether the role is permanent or a temporary bridge that fades as AI tools become more self-serving. Either way, the underlying skill, translating messy business reality into working systems, seems likely to hold its value.

How to start

Start with your own job. Map one workflow in painful detail, including everything that goes wrong. Decide which steps need AI judgment and which are just rules. Then build a small agent for it and measure what it saves in time, cost or risk. That loop of audit, evaluate and deploy is the whole FDE job in miniature, and doing it once on your own work will teach you more than any explanation.