Y CombinatorThe FDE Playbook for AI Startups with Bob McGrew
Diana Hu and Bob McGrew on palantir’s Forward Deployed Model Becomes Blueprint For AI Agent Startups.
In this episode of Y Combinator, featuring Bob McGrew and Diana Hu, The FDE Playbook for AI Startups with Bob McGrew explores palantir’s Forward Deployed Model Becomes Blueprint For AI Agent Startups Bob McGrew explains how Palantir’s forward deployed engineer (FDE) model was created to bridge a constantly shifting gap between a flexible platform and highly heterogeneous, mission‑critical customer needs, especially in government and defense. Instead of classic SaaS product‑market fit followed by distance from customers, Palantir institutionalized “doing things that don’t scale” through embedded technical teams who discover, prototype, and validate outcomes on‑site. Product and engineering then generalize these bespoke “gravel roads” into reusable platform capabilities—the “paved highways”—while growing account value over time. McGrew argues this approach is now spreading rapidly to AI agent startups because agents are a new, undefined product category with no incumbents, demanding deep in‑enterprise product discovery and outcome‑based pricing.
At a glance
WHAT IT’S REALLY ABOUT
Palantir’s Forward Deployed Model Becomes Blueprint For AI Agent Startups
- Bob McGrew explains how Palantir’s forward deployed engineer (FDE) model was created to bridge a constantly shifting gap between a flexible platform and highly heterogeneous, mission‑critical customer needs, especially in government and defense. Instead of classic SaaS product‑market fit followed by distance from customers, Palantir institutionalized “doing things that don’t scale” through embedded technical teams who discover, prototype, and validate outcomes on‑site. Product and engineering then generalize these bespoke “gravel roads” into reusable platform capabilities—the “paved highways”—while growing account value over time. McGrew argues this approach is now spreading rapidly to AI agent startups because agents are a new, undefined product category with no incumbents, demanding deep in‑enterprise product discovery and outcome‑based pricing.
IDEAS WORTH REMEMBERING
7 ideasTreat forward deployment as embedded product discovery, not implementation services.
FDEs sit inside customer organizations to identify high‑value problems, rapidly prototype solutions, and validate outcomes; their work is explicitly used to inform what the core product team should generalize for the next 5–10 customers.
Structure field teams into complementary “echo” and “delta” roles.
Echoes are domain‑savvy, rebellious analysts/account leads who find and frame valuable use cases, while Deltas are fast‑moving engineers who absorb “pain” and ship rough but working solutions under tight timelines.
Hire domain heretics and rapid prototypers, not classic enterprise sales and craftsman engineers.
Echoes should deeply know the domain yet believe the status quo is broken; Deltas must prioritize speed and learning over elegant, long‑lived abstractions to deliver concrete wins quickly.
Design the core product as a leverage engine for FDEs, not a finished vertical app.
The home team’s job is to convert one‑off field solutions into generalizable platform capabilities (like Palantir’s ontology), increasing product leverage so future deployments deliver more value without proportionally more headcount.
Measure success by outcome value and contract expansion, not by minimizing per‑customer work.
Unlike classic SaaS, a healthy FDE business often starts unprofitable at an account; over time, product improvements and deeper access to critical problems should lower cost per unit of value and grow deal size.
Use demos as a forcing function to build from the user’s perspective.
Demo‑driven development, when grounded in real workflows, compels teams to integrate features into compelling end‑to‑end stories that create genuine desire in customers and reveal friction before broad rollout.
AI agents need FDE-style adoption because the category is undefined and heterogeneous.
With no clear incumbents and huge capability–adoption gaps, AI agent startups must learn in situ which workflows matter, how to integrate into complex enterprises, and how to price against real business outcomes.
WORDS WORTH SAVING
5 quotesThe FDE model effectively is doing things that don’t scale at scale.
— Bob McGrew
Fundamentally what you’re selling with the FDE model is that you’re not selling the installation of software. You’re selling an outcome.
— Bob McGrew
Your other key customer is the FDE. Your product should be delivering leverage to the FDE who’s delivering that outcome at the customer site.
— Bob McGrew
With AI agents, there is no incumbent product.
— Bob McGrew
This is like exactly the training to become a startup founder.
— Bob McGrew
QUESTIONS ANSWERED IN THIS EPISODE
5 questionsHow can an early‑stage AI startup practically distinguish between high‑value “outcomes” worth pursuing and distracting bespoke requests that will push them into low‑margin consulting?
Bob McGrew explains how Palantir’s forward deployed engineer (FDE) model was created to bridge a constantly shifting gap between a flexible platform and highly heterogeneous, mission‑critical customer needs, especially in government and defense. Instead of classic SaaS product‑market fit followed by distance from customers, Palantir institutionalized “doing things that don’t scale” through embedded technical teams who discover, prototype, and validate outcomes on‑site. Product and engineering then generalize these bespoke “gravel roads” into reusable platform capabilities—the “paved highways”—while growing account value over time. McGrew argues this approach is now spreading rapidly to AI agent startups because agents are a new, undefined product category with no incumbents, demanding deep in‑enterprise product discovery and outcome‑based pricing.
What concrete metrics or leading indicators should founders track to know whether product leverage for FDEs is actually increasing over time?
How do you design hiring, compensation, and career paths so that echo and delta roles remain attractive, sustainable, and tightly aligned with product evolution instead of drifting into siloed services?
In heavily regulated or on‑premise environments, what are the most effective strategies FDE teams can use to secure top‑level executive sponsorship and navigate IT and security roadblocks?
As AI capabilities race ahead but adoption lags, what kinds of “missing middle” companies or products are most needed to translate raw models into reliably deployed, organization‑wide workflows?
EVERY SPOKEN WORD
Install uListen for AI-powered chat & search across the full episode — Get Full Transcript
Get more out of YouTube videos.
High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.
Add to Chrome