The FDE Model: How AI Startups Scale Bespoke Solutions

Summary

An in-depth walkthrough of the Forward Deployed Engineer (FDE) model, the strategy Palantir pioneered to solve heterogeneous customer problems and which is now being adopted at scale by AI startups. McGrew explains that AI agents, unlike existing software categories, lack an incumbent product, forcing startups to do extensive on-site product discovery. The playbook covers team structure (Echo teams for domain understanding and account management; Delta teams for rapid prototyping), the tension between custom work and product generalisation, outcome-based pricing, and why the model works best when driven by demo-ready solutions. Watch the full talk.

Key Claims

• AI agents are a new market category with no incumbent product; FDE-led on-site discovery is essential because the market itself is not yet defined.
• The core FDE mechanism is embedding engineers at customer sites to fill the gap between what the product does and what the customer needs, then generalising those site-specific solutions into platform features that serve the next ten customers.
• Outcome pricing, not per-seat or installation pricing, aligns incentives: startups sell the successful result of solving a problem, not just software, so early risk sits with the startup and enterprise buy-in focuses on executive leadership rather than IT gatekeeping.
• Product leverage, not minimisation of custom work, is the success metric: the goal is to increase the contract value and valuable outcomes per customer while product features do more of the lifting, measured by whether FDEs increasingly choose to use the product over one-off solutions.
• Echo and Delta teams have distinct profiles: Echo needs domain rebels—people from the customer's industry who recognise existing processes are broken; Delta needs rapid prototypers who accept throwaway code rather than architects optimising for long-term maintenance.
• The Palantir ontology illustrates a failure mode of FDE work: if each customer's solution is too specialised, the product becomes worthless for the next site. Success means identifying the abstraction level that fits many customers, not implementing every customer's exact request.
• Demo-driven development, where every feature is tested for relevance to a real customer problem (e.g. "How does this help the analyst stop a terrorist plot?"), forces product thinking from the customer's perspective and prevents feature bloat.
• Twenty-plus YC startups now hire FDE roles, up from nearly zero three years ago, reflecting a shift in how AI agent companies organise and an expectation that FDE capability is becoming a hiring and execution moat.

Quotes

• "The FDE model effectively is doing things that don't scale at scale."
• "You are solving one particular problem, and the simplest approach makes sense. That is the gravel road. But the paved road has to go by more than just one customer."
• "You aren't selling the installation of software; you are selling an outcome. You are selling that you have solved the problem."
• "In the FDE strategy, you want to drive the contract size up. You are doing more and more valuable work for this customer and for future customers."
• "It is almost like OpenAI is the home product team and the startups are the FDEs out figuring out how to get adoption of the research."