The Forward Deployed Engineer, Chapter 1: What Is a Forward Deployed Engineer?

This is the first post in a new series walking through my new book The Forward Deployed Engineer: Architecting the Last Mile of Enterprise AI. The previous series, The Engineering of Intent, closed yesterday. Where that book was about the practice of writing software inside the AI-native era, this one is about the role that decides whether enterprise AI actually works in production at all.

For most of my time as the CTO of a large business process outsourcing firm, I watched the same contradiction play out month after month. The technology was working. The pilots passed every benchmark we set. The procurement got signed. And then the deployments would stall — not in the model, not in the platform, but in the gap between the demo and the desk where the work was actually done.

It took me longer than it should have to name what was missing. The companies selling us the technology had no shortage of brilliant engineers, and no shortage of polished demos. What they didn’t have was anyone whose job it was to live inside our operations long enough to design the deployment around our reality, rather than the other way around. That role has a name now. It came out of Palantir, of all places — a defense-intelligence consultancy in 2005 that needed an operating model nobody else in software was building. Most of the press still traces the Forward Deployed Engineer back to that single company, and the mythology matters. The structural details matter more.

The Anatomy of the Role

Strip away the Palantir story and the FDE is a hybrid of four things at once: a production engineer, a data integrator, a product manager for one customer, and a field strategist. Each of those four skills is already rare. The combination is rarer still — which is why most companies’ first attempt to build the function ends up with senior solutions engineers in a new uniform, doing about a third of the job and discovering the other two-thirds the hard way.

Several technology cycles have argued for a role like this, and none has needed it as urgently as AI. The reasons are structural. AI is non-deterministic where SaaS was deterministic, so the workflow around it has to be redesigned, not just configured. AI changes workflows where SaaS replaced tools, so the deployment has to negotiate with the operator instead of merely installing. AI’s value lives in the long tail of customer-specific operations, where the demo never goes. And AI’s failure mode is silent rather than loud — the model doesn’t crash; it gives a slightly wrong answer that compounds across a thousand transactions before anyone notices. Each is a different reason the demo-to-deployment gap has widened, and each is a different reason the FDE is the role that closes it.

This is also why the most common mistake organizations make is treating the FDE as a relabel of an existing role. The Solutions Engineer sells before the contract; the FDE owns after it. The Sales Engineer answers questions; the FDE writes the code that answers them. The Customer Engineer maintains; the FDE deploys. The Implementation Consultant configures; the FDE re-architects. The differences are not cosmetic. They show up in compensation, in reporting line, in success metrics, and — most consequentially — in the kind of person you should be hiring.

The New Wave

For most of the 2010s, the Palantir operating model was a curiosity. Few outside the company understood the function; fewer still tried to copy it. In 2026, that has changed completely. The model is the template. OpenAI, Anthropic, Runway, Greptile, and the next generation of AI labs have rebuilt the function for their own customer bases — sometimes copying Palantir directly, sometimes adapting hard, but always landing on something recognizable as the same shape.

What this means depends on who you are. If you’re a founder building an AI platform, the question is no longer whether you need a Forward Deployed Engineer; it’s when, how, and how seriously. If you’re an executive evaluating enterprise AI, the question is which vendors actually have one and which only have a sales team that uses the title. And if you’re an engineer considering the move, the question is whether you want to spend the next five years on the most leveraged, most exposed, and — by quite a wide margin — most highly compensated engineering work in the industry.

This series walks through all of it. Tomorrow, the last-mile problem the role exists to solve. Then the skillset, the playbook, the case studies, the economics, and the future. Nineteen days, one chapter at a time.