This session at Mercury Fund Day at The Ion brings together product and engineering leaders building at late-stage companies, and startups: Mustafa Ali (Twilio), Zafir Khan (WhatsApp/Meta), and Salim Gheewalla (Utilitize). Moderated by Heath Butler, the conversation covers what separates a durable product from a thin wrapper on a foundational model, why vertical AI companies are harder to replicate than horizontal ones, and how the speed of AI-assisted prototyping is changing what founders look for when they hire. The panel also gets into pricing, from consumption-based models to the risks of passing token costs straight through to customers, and closes with a candid discussion of how AI tools affect IP and patent risk.
Key Takeaways
- Vertical AI is harder to copy than horizontal AI. Foundational models can write good code fast, but they can't replicate the domain-specific workflows and nuanced data that only come from operating inside an industry.
- Trust separates a real product from a feature. Whether it's HIPAA compliance, brand design, or IT access controls, customers need proof that a solution works, not just a demo that looks good.
- AI has compressed the idea-to-prototype cycle, and hiring is shifting to match. Panelists are prioritizing curiosity, ownership, and first-principles thinking over credentials, since anyone can now build and show a working prototype instead of pitching a deck.
