The FDE Playbook for AI Startups with Bob McGrew

The Forward Deployed Engineer (FDE) Model: Origins and Mechanics

• Definition: An FDE is a technical engineer stationed at a customer site to bridge the gap between a product's current capabilities and the specific needs of a user, effectively performing "product discovery" rather than traditional implementation services.
• Palantir Origin: The model was invented at Palantir to solve the "spies' problem": intelligence agencies would not disclose their workflows, necessitating that engineers build custom solutions on-site to understand actual needs before generalizing them.
• Strategic Shift: Unlike standard SaaS models that prioritize scaling after product-market fit, the FDE model intentionally maintains high-touch, "non-scalable" work to drive up contract size and value for each customer over time.
• Team Structure: Palantir utilizes two distinct roles:
  • Echo Team: Embedded analysts with deep domain expertise (e.g., former military or healthcare) who act as account managers and identify high-impact problems.
  • Delta Team: Rapid-prototyping engineers who build rough, functional code to solve immediate customer problems, accepting that early code may be discarded.
• Product Feedback Loop: FDEs build "gravel roads" (custom prototypes) at customer sites; the central product team then identifies generalizable patterns to convert these into "paved superhighways" (scalable platform features).
• The Ontology Innovation: Palantir's shift from rigid database schemas to a flexible "ontology" (defining objects, properties, and links generically) was a direct result of FDE feedback, allowing customization per customer while maintaining a unified data structure.

Adoption in the AI Agent Sector

• Market Heterogeneity: Unlike SaaS where incumbents exist, the AI agent market lacks incumbent products, requiring extensive, on-site discovery to define what an agent actually does in a specific context.
• Hiring Trends: The FDE model has become dominant in AI; YC job boards show over 100 startups hiring for "Forward Deployed Engineer" roles, up from near zero three years ago.
• Contract Structure: AI startups using this model target large, flexible contracts based on "outcomes" rather than usage or seats, often starting with negative margins that become positive as product leverage increases.
• Risk Asymmetry: Startups often accept high risk early on, agreeing to pay models where revenue is tied to successful outcomes (e.g., number of calls handled) to bypass enterprise skepticism regarding startup execution capabilities.
• Obstacles: Success requires navigating internal IT departments and securing executive buy-in from the CEO's "top five" priorities to bypass bureaucratic hurdles.

Strategic Implementation and Best Practices

• Pricing Strategy: Value is priced based on the outcome delivered to the customer, not the cost of delivery, allowing contracts to grow in size as the FDE team deepens their impact on the client.
• Success Metric: The key internal KPI is "contract size growth" and "value of the outcome," not the reduction of custom work per customer, as customization is viewed as necessary for value creation.
• Product Leverage: A critical measure of success is whether the product team can provide FDEs with tools that allow them to deliver more value without increasing headcount at the customer site.
• Demo-Driven Development: The necessity of repeatedly demonstrating value to new customers forces product development to focus on the end-user workflow and desired outcomes rather than isolated technical features.
• Common Pitfalls: Companies often fail by becoming pure consulting firms (selling installation rather than outcomes) or by building features requested by lower-level sponsors rather than those aligned with the CEO's strategic priorities.
• Talent Acquisition: The most successful AI startups adopting this model have often recruited personnel with prior Palantir FDE experience to navigate the specific mechanics of outcome-selling and on-site discovery.

Bob McGrew's Perspective and Future Outlook

• Current Role: Bob McGrew, former Chief Research Officer at OpenAI (lead on GPT-4 and o1), has joined the U.S. Army Reserve as a Lieutenant Colonel advising on technology transformation for large-scale combat operations.
• Analogy: McGrew describes the current AI landscape as OpenAI acting as a "home product team" while startups act as the FDEs executing on-site to drive adoption of new capabilities.
• Capability vs. Adoption Gap: While AI capabilities (e.g., GPT-4 to o1) are improving rapidly, global adoption is lagging, creating a massive opportunity for FDEs to fill the gap between technical potential and practical utility.
• Organizational Culture: McGrew notes that successful FDE organizations must remain "learning companies" that avoid the complacency of coasting on established strategies, similar to the environment required for startup founders.
• Forward-Looking Statement: McGrew predicts that over the next five years, AI capabilities will continue to race ahead while the world feels "increasingly banal," emphasizing that human ingenuity and on-site adaptation are the primary drivers of value extraction.