What Is Forward Deployed Engineering (FDE)?

Why AI pilots fail at the integration stage — and how Forward Deployed Engineering solves the problem

AI pilots are running in almost every organisation. The demo works, the model delivers compelling results, and everyone is excited. Then the project stops right there: at the pilot stage. The reason is rarely the model itself. It is the legacy databases, the lack of authentication integration, and the processes that no one outside the IT department fully understands.

The bottleneck between pilot and production is not a model problem, it is an integration problem. And that is exactly the problem Forward Deployed Engineering solves.

What is a Forward Deployed Engineer?

A Forward Deployed Engineer (FDE) works directly within the customer’s system environment to make a complex software product work in practice. Unlike a traditional consultant, an FDE writes code rather than simply providing advice: they build connections to existing data sources, adapt workflows to established processes, and uncover undocumented constraints that would otherwise block a project after kick-off.

The role combines three skill sets that have traditionally been separate: software engineering, product thinking, and customer consulting. An FDE is therefore neither purely an engineer nor purely a consultant, but a combination of both.

Where does the term come from?

The term was coined by Palantir. The reason was straightforward: from an early stage, Palantir worked with large customers such as government agencies, airlines, and banks. Their problem was never a lack of features, but fragmented data systems, long-established legacy workflows, and high-stakes operational risks. The answer was to embed engineers directly within customer teams rather than delivering features from a distance.

The fact that this model has since become standard practice in the enterprise market is evident from the number of providers adopting it. OpenAI has launched its own FDE initiative, “The Deployment Company”, Google is expanding its FDE organisation, and even established software providers such as Adobe are hiring “Forward Deployed AI Engineers”. When the companies with some of the world’s best models send people directly to customers to make those models work, the message is clear: value is created through integration, not through the model alone.

What FDEs actually do

The ideal candidate has a “T-shaped” profile: technically deep enough to build systems in practice, while broad enough to break down a complex customer problem into solvable parts, deliver a lean initial approach, and refine it iteratively.

The difference from traditional software development lies in the starting point. Most AI projects do not fail because of the model, but because it cannot communicate with the customer’s legacy databases or support their authentication systems (OIDC, SAML). This is exactly where FDE work begins: integrating AI into existing, historically grown software environments rather than selecting the model itself.

Why building an in-house FDE team is unrealistic for most companies

The obvious response for many business leaders is to hire their own FDE. In practice, however, this usually runs into two challenges.

Cost. Experienced FDEs are in high demand and command correspondingly high salaries. For a mid-sized company in the DACH region, a single position at this level is rarely a realistic budget item.

Availability. The talent pool is small, and job titles are inconsistent. The same role may be advertised as anything from “Deployment Engineer” to “Applied AI Engineer”, while the actual responsibilities vary considerably. Making the wrong hire at this level is costly and difficult to correct.

FDE as a Service: the pod approach for mid-sized companies

For most mid-sized companies in the DACH region, the choice between “no FDE at all” and “an in-house FDE team on enterprise-level salaries” is not really a choice. demicon’s FDE pod approach bridges precisely this gap: instead of permanently hiring a single, hard-to-find specialist, demicon provides embedded engineering capacity as a service — project-based, scalable, and with the T-shaped skill set that complex integration projects genuinely require.

This is particularly relevant for industries with long-established, heterogeneous system landscapes: automotive, manufacturing, finance, the public sector, and aerospace & defence — precisely the environments in which AI pilots tend to fail due to legacy integration, compliance requirements, and a lack of internal capacity.

FAQs