Descriptions:
In this 20VC interview with Harry Stebbings, Gokul Rajaram — veteran operator across Google, Facebook, and Square, and now an active investor — lays out a comprehensive framework for evaluating whether AI companies can achieve durable, defensible competitive positions. The centerpiece is what he calls the eight modes of defensibility: data mode, workflow mode, regulatory mode, distribution mode, ecosystem mode, network mode, physical infrastructure, and scale mode. His rule of thumb: four or more modes and a company is ‘pretty damn secure.’
Rajaram draws on direct experience to ground each principle. At Google he learned that remarkable product is the non-negotiable core; at Facebook, that distribution and multiplayer dynamics compound defensibility (which informed his early conviction on Figma); at Square, that a multi-product portfolio — where each successive product emanates naturally from the last — is the path to a $10B+ company. He applies these lenses directly to AI-native businesses, arguing that vertical agents must own the full software stack to remain defensible as AI replaces digital labor.
The conversation also tackles pricing model evolution with clarity: seat-based pricing survives for access products where predictability matters (he cites ChatGPT Enterprise), but breaks down for work products where the user is no longer the constraint. Outcome-based pricing — paying per contract processed rather than per user — is the natural model for tools like Harvey. For founders building in competitive AI verticals, this interview offers one of the more structured analytical frameworks currently available for stress-testing long-term business viability.
📺 Source: 20VC with Harry Stebbings · Published March 16, 2026
🏷️ Format: Interview
