Summary
Will Gaybrick, Stripe’s President of Product and Technology, joins Andreessen Horowitz to detail how the payments giant is embedding AI throughout its product stack and internal engineering culture. Gaybrick describes Stripe’s evolution from a payments processor into a 25-to-30-product financial infrastructure platform, and explains how agentic AI workflows are letting a single engineer accomplish what required two full teams just two years ago — a productivity shift he says is now stressing every internal system as the volume of merged code surges.
On the customer-facing side, Gaybrick shares concrete outcomes: Stripe’s AI fraud pipeline, built on foundation model embeddings and a reasoning layer, is enabling 11 Labs to block 2,000 fraudulent free-trial signups per day using Stripe Signals. He also reveals that Stripe built global tax filing — a problem of far greater complexity than US-only filing — in roughly one-third the time it took to build the domestic version, crediting AI-assisted engineering. Internally, Stripe created a tool called Stripe Minions to handle agentic development tasks with minimal back-and-forth.
Gaybrick argues that the “Cambrian explosion” moment for AI-driven commerce hasn’t arrived yet, and that companies currently using AI primarily to optimize cost structure are missing the larger opportunity. The conversation also covers Stripe’s approach to resourcing in the agentic era, why he believes founder-like agency inside large companies is now achievable, and how the average AI company now uses 11 distinct Stripe products.
📺 Source: a16z · Published August 17, 2026
🏷️ Format: Interview
