The FDE Playbook for AI Startups with Bob McGrew
Timestamps are as accurate as they can be but may be slightly off. We encourage you to listen to the full context.
Podcast Summary
In this episode, Bob McGrew, former Chief Research Officer at OpenAI and early Palantir executive, reveals why the Forward Deployed Engineer model has become the dominant strategy for AI agent startups. McGrew explains how this unconventional approach—embedding engineers directly at customer sites to bridge the gap between product capabilities and customer needs—evolved from Palantir's early days (03:21) and why it's experiencing a renaissance in Silicon Valley. He shares tactical insights on structuring FDE teams, avoiding the consulting trap, and the critical importance of outcome-based pricing over traditional SaaS metrics (30:31). The discussion explores how AI agents represent a completely new market category with no incumbent products, making traditional scaling strategies insufficient and necessitating this hands-on discovery approach for ambitious founders building the next generation of enterprise AI.
Speakers
Bob McGrew
Early engineer at PayPal, early executive at Palantir, and former Chief Research Officer at OpenAI where he led development of ChatGPT, GPT-4, and the o1 reasoning model. Recently joined the US Army Reserve as a Lieutenant Colonel in detachment 201, advising on military technology transformation.
Jared (Host)
Host of Lightcomb podcast, discussing topics with high-achieving startup founders and industry experts on product strategy and scaling technology companies.
Key Takeaways
Target the CEO's Top 5 Problems
Build solutions for executive-level priorities, not department requests. You need CEO-level sponsorship to overcome IT resistance and organizational inertia—anything lower lacks the authority to push through friction. (10:00)
Graduate from Gravel Roads to Superhighways
FDEs build rough prototypes (gravel roads) at customer sites; product teams then generalize these into scalable features (paved superhighways). The key is abstracting the right level—solving for classes of problems, not individual customer quirks. (07:11)
Hire Heretics, Not Experts
Your embedded analysts need deep domain knowledge but must be rebels who see current processes as insufficient. Craftsmen who love perfect abstractions make terrible forward deployed engineers—you need builders who can deliver outcomes on tight timelines. (12:06)
Scale Contract Size, Not Customer Efficiency
In traditional SaaS, you optimize for less work per customer. With FDEs, optimize for bigger contracts per customer—more valuable outcomes justify higher touch. Measure contract value growth, not operational efficiency. (36:33)
Make Your Product Serve Two Masters
Build for end users and FDEs simultaneously. Your platform should deliver great user outcomes after FDE customization while providing leverage to FDEs. If FDEs bypass your product for custom hacks, you're not abstracting correctly. (38:12)
Statistics & Facts
- Over 100 YC startups are currently hiring forward deployed engineers according to job board data, up from basically zero three years ago. (02:09)
- In the early days of Palantir, even when dealing with users who didn't perform the same workflows, if you can't make a 3x or 10x change within an organization, there was no reason to go through all the effort of the FDE deployment process. (02:28)
- The Palantir forward deployed engineering strategy requires solving one of the CEO's top five priorities, as organizations won't have the energy to persist through the challenging route of getting a new product built without high-level executive buy-in. (09:00)
