The FDE Playbook for AI Startups with Bob McGrew
Bob McGrew, an early executive at Palantir and former Chief Research Officer at OpenAI, explains the Forward Deployed Engineer (FDE) model he helped pioneer. He details its origins, operational structure, and why it has become the dominant strategy for AI agent startups navigating a new market with immense product discovery needs.
The Forward Deployed Engineer (FDE) model, once a unique strategy pioneered at Palantir, is experiencing a resurgence and has become the dominant go-to-market and product development strategy for modern AI agent companies. This is not a coincidence, but a necessary adaptation to a market defined by a completely new product category where no incumbent solution exists. The core challenge is one of immense product discovery, which can only be effectively done from inside a customer’s organization. The FDE model is, in effect, a method for “doing things that don’t scale, at scale.”
The FDE Model Explained
An FDE is a technical engineer who resides at the customer site to fill the gap between what a product currently does and what the customer truly needs. When a new customer presents a problem the product hasn’t solved before, the FDE takes the existing platform and, with support from the core product team, builds out the specific use case to deliver a valuable outcome.
This approach contrasts sharply with the traditional “product-market fit” playbook. Once a typical SaaS company finds product-market fit, it embraces distance from the customer to focus on scaling a standardized solution. The FDE model, however, is for situations where customer needs are heterogeneous and a one-size-fits-all approach is impossible.
Origins: Building Software for Spies at Palantir
The FDE model was born out of necessity at Palantir. The initial goal was to build software for intelligence analysts, but the founders didn’t know any spies and couldn’t simply ask them about their workflows.
The process started with demo-driven development. They would show a demo to potential users, who would invariably say it was terrible. By asking, “how would you like it to be different?” and iterating, they slowly got closer. They discovered that every customer’s problem was slightly different. This led to the creation of a flexible platform rather than a rigid product, which required someone on-site to perform the necessary customization. Palantir’s early leader, Shyam Sankar, realized this necessity could be turned into a strategic advantage by making the FDEs a core part of the product discovery process.
The FDE goes and builds a gravel road to where the product needs to go. And then the role of my team, the product and engineering team, was to look at that and basically figure out how that should generalize to the next five customers or the next 10 customers and then turn that gravel road into like a paved superhighway.
The FDE Operating Manual: Echo and Delta Teams
The model was structured around two key roles that function much like a startup’s founding team:
- Echo Team (Embedded Analysts): These individuals are domain experts, often hired from the customer’s industry (e.g., former army officers). Crucially, they must be “heretics”—people who understand the domain deeply but recognize its current methods are insufficient. They act as account managers and are responsible for identifying the key problems and use cases that will deliver a step-function change in value for the customer.
- Delta Team (Deployed Engineers): These are software engineers skilled at rapid prototyping. Their job is not to build perfectly abstracted, maintainable code, but to deliver a working solution on a tight timeline. They are comfortable “eating pain” to get a functional prototype deployed that solves the immediate problem.
Avoiding the Consulting Trap
A common criticism of the FDE model is that it’s just consulting disguised as a software business. This is a real risk that requires significant organizational discipline to avoid. The key difference lies in the business model and product leverage.
A true software business using the FDE model will see its profit margins at a customer increase over time. This happens for two reasons:
- Improved Product: As the core product team generalizes the solutions built in the field, the product becomes better suited to the customer’s needs, reducing the amount of custom work required for subsequent tasks.
- Expanded Scope: By successfully solving initial problems, the team earns the right to tackle more important and valuable problems within the organization, driving the contract value up.
This requires a product team that can think at a higher level of abstraction. A prime example from Palantir is the creation of its Ontology. Instead of building specific database tables for “people” or “money,” they created a generalized system of objects, properties, and links that could be defined by the FDE team for each customer’s specific domain.
Why AI Companies are Adopting the FDE Model
The reason the FDE model is so relevant for AI agent startups is that they are creating a completely new market category. Unlike a new SaaS tool that replaces an existing, well-understood workflow (like paying bills), AI agents have no direct incumbent. The landscape of what’s possible and valuable is unknown.
This creates an environment of extreme heterogeneity, similar to what Palantir faced. The product discovery required to find valuable, repeatable use cases cannot be done from the outside; it must be done by embedding engineers within the enterprise to understand the problems and workflows from the inside out.
Metrics and Strategy: From Cost Reduction to Value Expansion
The KPIs for an FDE-driven company are counterintuitive compared to a traditional SaaS business.
- Traditional SaaS: The goal is to decrease the cost and work required per customer while keeping contract size stable.
- FDE Model: The goal is to constantly increase the value of the outcome delivered, which in turn drives the contract size up. The amount of customization per customer can remain constant or even grow, as long as the value being delivered is growing faster.
The key internal metric is product leverage. The core platform should make it progressively easier for an FDE to deliver more value to the customer over time, whether that’s deploying a known solution faster at a new customer or using existing platform capabilities to solve a completely novel problem.
The Biggest Opportunity for AI Founders
The current AI landscape is defined by a massive gap between capability and adoption. Foundational models are advancing at an incredible pace, but the world is not changing nearly as fast.
The world is going to look like over the next 5 years is that the capabilities just race ahead and race ahead and race ahead and somehow the world feels increasingly banal.
The greatest opportunity for founders lies in bridging this gap. It requires the human ingenuity, exploration, and willingness to deal with pain that is characteristic of the FDE model. In this analogy, foundational model companies like OpenAI are the central product team, and AI startups are the FDEs in the field, figuring out how to apply these powerful new capabilities to solve real-world problems and drive adoption.