What Is a Forward Deployed Engineer, and Do You Need One?

A year ago almost nobody outside Palantir used the phrase. Now "forward deployed engineer" is one of the most searched job titles in technology, OpenAI has set up a company built around the role, Anthropic has a venture staffed with them, and an executive search firm told TechCrunch it expects demand to rise 2,100% by the end of 2026.

So what is a forward deployed engineer, why did the AI industry suddenly decide it needs thousands of them, and does a mid-sized firm that is trying to get real value from AI need one?

I run engineering at Chronexa, and most of what my team does every week is, in practice, forward deployed engineering for companies too small to hire it in-house. Here is my explanation of the role, what the hype gets right and wrong, and how to think about it as a buyer.

Here's what matters most:

• A forward deployed engineer (FDE) is a software engineer who works inside a customer's business, not at a desk back at headquarters, building systems on the customer's real data and processes.
• The role was invented at Palantir and is now central to how OpenAI, Anthropic and others get their AI into large companies.
• Demand is exploding because AI models are easy to buy and hard to make useful. The bottleneck has moved from the technology to the last mile inside each business.
• The supply is tiny. Executive search firm Christian & Timbers estimates only about 2,000 US engineers combine the industry knowledge, credibility and hands-on AI skill to deliver consistent results.
• Most mid-sized firms do not need a full-time FDE. They need FDE-style work for a defined project and a team accountable for running it afterwards.

What is a forward deployed engineer?

A forward deployed engineer is a software engineer who is embedded with a customer, sometimes physically on site, to understand how the customer actually works and then build, configure and adapt technology until it solves a real problem in that environment.

A traditional software engineer builds one product for many customers and rarely meets them. A traditional consultant advises customers but often does not build. An FDE does both: sits with the people who do the work, finds out where the process really breaks, and then writes the code, connects the systems and stays until the thing is used.

The term "forward deployed" comes from military language, where forward deployed units operate close to the front line rather than at a central base. That is not an accident. Palantir, the data analytics company that invented the role, built its early business with government and defense customers. Shyam Sankar, who joined Palantir in 2006 as roughly its thirteenth employee and is now its chief technology officer, is widely credited with conceiving the forward deployed engineer model. Accounts of the company's early years describe engineers spending long stretches away from headquarters, embedded with customers and adapting software on the spot.

For years, critics described Palantir's model as "a consulting firm with good margins." Today the rest of the AI industry is copying it.

What does a forward deployed engineer actually do?

The job varies by company, but in practice an FDE's week usually includes four kinds of work.

Sitting with the people who do the job. Watching how an accounts team actually processes invoices, how a claims handler actually triages a file, how a paralegal actually assembles a document set. The goal is to find the real process, which is almost never the one written down.

Connecting the messy parts. Most of the effort goes into getting data out of old systems, cleaning it, and making the AI's output land back in the tools people already use. In my experience this is 70% of the work on any AI project, and it is the part a demo never shows.

Building and adapting the AI system. Writing the instructions the AI follows, choosing which model handles which task, building the review steps where a person approves the AI's work, and testing it on real cases, especially the strange ones.

Staying until it is used. Training staff, fixing what breaks in the first weeks, adjusting when people route around the system, and measuring whether it delivers what was promised.

That is why the role pays well and is hard to fill. It needs a working engineer, a patient listener, someone comfortable in a boardroom and someone who understands the customer's industry well enough to be trusted by its experts.

Why are AI companies suddenly hiring so many forward deployed engineers?

Because the AI models themselves have become easy to buy, and getting value from them has not.

Companies have spent heavily on AI pilots that never reached production. A widely cited 2025 study from MIT's NANDA initiative reported that about 95% of corporate generative AI pilots produced no measurable profit-and-loss impact, blaming a "learning gap" between the tools and how organizations actually work. Whatever you think of that exact figure, the direction matches what every AI vendor sees: the bottleneck is the last mile inside each business.

The AI labs have responded by building their own forward deployed armies:

• OpenAI launched the OpenAI Deployment Company in May 2026, a majority-owned subsidiary backed by more than $4 billion in initial capital from 19 investment firms, consultancies and systems integrators, led by TPG. It agreed to acquire Tomoro, a UK applied AI consultancy, which was expected to bring about 150 experienced forward deployed engineers and deployment specialists. Its engagements reportedly begin with a diagnostic, after which FDEs design, build, test and deploy production systems connecting OpenAI's models to the customer's systems.
• Anthropic has a services venture called Ode, led by chief executive Chris Taylor and staffed with FDEs, according to TechCrunch. Decrypt reported that Anthropic unveiled a $1.5 billion venture with Blackstone, Goldman Sachs and Hellman & Friedman shortly before OpenAI's announcement.
• Large enterprises and consultancies are building their own teams. In a study shared with TechCrunch in July 2026, Christian & Timbers reported that at the start of 2026 only 5% to 10% of companies planned to hire FDEs, mostly for small pilots, but by the end of the second quarter that share had reached 70%. The largest consulting and services firms told the search firm they needed to grow FDE headcount tenfold, into teams of 20 to 100 people.

The economics are simple. An AI lab earns money when its models are used at scale. Models sitting unused in a stalled pilot earn nothing. Embedding engineers who make the models useful is how the labs turn a pilot into a long-term, high-usage customer.

How much does a forward deployed engineer earn?

Pay data for a title this new is messy, and most published figures come from recruiters and job sites rather than audited data, so treat these as ranges.

One mid-2026 analysis of 135 US forward deployed engineer job postings found a median base salary of about $190,000. Palantir's own postings for its equivalent role have advertised base salaries of roughly $125,000 to $200,000. Roles at the frontier AI labs are reported to pay considerably more in total compensation, with most of the difference coming from equity, though the published ranges for those roles vary widely.

For a mid-sized company, the more important number is not the salary but the scarcity. Christian & Timbers estimates there are about 17,000 people working as FDEs in the US, many of them at Palantir, but only about 2,000 with the combination of sector knowledge, credibility and hands-on applied AI experience to consistently deliver a return. The firm's founder, Jeff Christian, put it bluntly: "Not 2,000 available. 2,000 total."

Does your company need a forward deployed engineer?

Here is my honest view as someone who does this work: most mid-sized firms do not need to hire one, but almost all of them need the work an FDE does.

Consider what you would be hiring. A great FDE is a scarce, expensive person who is being recruited by OpenAI, Anthropic, Palantir and every large consultancy at the same time. If you hire one, you have one person, with one set of skills, who will need projects to stay engaged and will be a single point of failure for every system they build. When they leave (and in this market they will be approached constantly), the knowledge leaves with them.

I learned this lesson before Chronexa. At a fintech company I helped build, we had our own development team and plenty of connected tools, and we still needed someone whose whole job was connecting the tools, the data, the approvals and the AI into something the business could run. The engineering talent was there. The missing piece was dedicated ownership of that connecting layer. Most companies without an engineering team have the same need with none of the in-house capacity.

So there are three realistic options.

Hire an in-house FDE if you have a large, continuous pipeline of AI work, the budget to pay frontier-lab rates, and enough engineering around them that they are not alone. This is the right answer for larger companies treating AI as a permanent capability.

Buy FDE time from a platform vendor (Palantir, the OpenAI Deployment Company, Anthropic's services venture, or a large consultancy) if you are a large enterprise committing to one vendor's platform at scale. Expect seven-figure engagements and a strong pull toward that vendor's technology.

Use a specialist implementation partner if you are a mid-sized firm with a few high-value workflows to fix and no appetite for a permanent hire. You get FDE-style work (embedded discovery, building on your real data, staying until it is used) for a defined project, plus an ongoing team accountable for running and fixing the system afterwards. This is the gap firms like Chronexa exist to fill, and it is the model I would choose in a mid-sized firm's position.

Whichever you choose, insist on the parts of the FDE model that actually matter: the person building the system must spend real time with the people who do the work, must build on your real data rather than samples, and must stay accountable after launch.

What should you watch out for?

The FDE boom has a downside buyers should know about.

Title inflation. Because the title is hot, some firms are relabelling ordinary consultants or solutions engineers as forward deployed engineers. Ask what they have personally built and put into production, on what kind of data, and who still runs it.

Vendor gravity. An FDE employed by a model provider or platform company will naturally build on that company's technology. That may be fine, but understand the incentive. Anthropic's Chris Taylor made a related point to TechCrunch: many FDEs are well equipped to help roll out an AI coding tool to a workforce, but very few can build a flagship AI product feature.

Data and confidentiality. An embedded engineer sees your processes, data and clients. Make sure the contract covers confidentiality, where data is processed, whether your data can be used to improve anyone's models, and what happens to everything they built when the engagement ends. Christian told TechCrunch that some companies are building internal FDE teams precisely because they worry that sharing proprietary processes with AI firms could help those firms compete with them.

A role that may not last. Even the recruiters are cautious. Christian told TechCrunch the FDE role could "go away" entirely within five to ten years as AI agents automate more of the integration work. Build systems your organization can own and run, not systems only one embedded engineer understands.

Frequently Asked Questions

What does a forward deployed engineer do?

A forward deployed engineer works inside a customer's business to understand how work is really done, then builds, connects and adapts software (today usually AI systems) on the customer's real data until it is in daily use. The role combines software engineering, consulting and support, and was pioneered by Palantir.

How much is a forward deployed engineer paid?

Published figures vary. One mid-2026 analysis of 135 US job postings found a median base salary of about $190,000, and Palantir's postings have advertised roughly $125,000 to $200,000 base. Roles at frontier AI labs are reported to pay more in total, mostly through equity.

Why is forward deployed engineering in such high demand?

Because AI models are easy to buy but hard to make useful inside a specific business. Christian & Timbers reported that the share of companies planning to hire FDEs jumped from 5% to 10% at the start of 2026 to 70% by the end of the second quarter, and OpenAI launched a dedicated deployment company in May 2026.

Does a small or mid-sized company need a forward deployed engineer?

Usually not as a full-time hire. Most mid-sized firms have a handful of high-value workflows to fix, which is better served by a specialist partner doing FDE-style embedded work for a defined project and then running the system, than by recruiting one scarce engineer who becomes a single point of failure.

The bottom line

The forward deployed engineer is the AI industry admitting something it spent two years avoiding: the model is the easy part. Value comes from someone sitting with your team, working on your real data, and staying until the system is actually used. Large companies will hire that capability or buy it from the AI labs. Mid-sized firms should buy the work, not the headcount, and insist on the same discipline: embedded, real data, accountable after launch.

If you have a workflow that needs that kind of attention, our AI agent development and AI integration services pages explain how we work, and you can book a 30-minute call to talk it through. Before you commit to any partner, read our guides on whether to pay for an AI proof of concept and on fixed price vs time and materials for AI projects.