Running Product Development in the Age of AI

Over the last year, I've built and scaled a new product engineering team from scratch: hiring, training, and shipping a real product while AI-assisted development rapidly changed how software gets built.

What surprised me most wasn't the tooling. It was how often our ways of working stopped being effective.

Roughly every quarter, we had to rethink how the team operated: how changes reach production, how engineers work day-to-day, and even how the system itself must be designed to support dramatically faster development.

This talk shares three practical pillars we've learned for running product development in the age of AI.

When engineers — and AI — can generate change faster than ever, safety can no longer rely on manual review alone. This section covers the verification layers we introduced to prevent costly mistakes before production: automated architectural checks, API and schema protection, stronger static validation, and testing strategies designed for high change velocity.

AI tools don't automatically make teams effective. We had to explicitly train engineers in new workflows and rethink how we evaluate engineering performance. This section shares how we coach engineers to work with AI systems productively, and what signals actually indicate effective AI-assisted development.

Our current product workflow assumes many concurrent changes happening safely at once. This section walks through how we use preview environments, isolated branches, and architectural boundaries that allow teams to move quickly without constant coordination or instability.