Every engineering team got Claude Code, Cursor or Codex this year. Engineers write more code than ever. And the time from a PM's idea to a change in production barely moved.
Over the last two years we helped product teams at hundreds of companies, from public enterprises to 30-person startups, move work upstream: product and design producing merge-ready changes on the real codebase, and engineering shifting from building to verifying. Along the way we read 326 buyer calls across 131 companies to understand exactly where the old workflow breaks. 78% of them say the same thing: the idea still travels PRD to Figma to Jira to engineer to QA, and loses intent at every handoff. 68% say the AI tools only work for engineers. 53% say engineering now has more AI-written code than it can trust.
This talk is what the workflow actually looks like inside the companies that made the shift. In five minutes: why the bottleneck moved from code to intent and trust, the three things AI-native teams stopped doing, the new division of labor between PM and engineering, and the one metric that tells you whether your team is AI-native or just has copilots.