Review bottleneck
The point at which AI-accelerated code generation outpaces the team's capacity to review it, so cycle time stops improving.
If AI doubles PR volume and PRs are 2.5× larger, roughly 5× as much code enters review. Faros measured a 91% rise in review time on high-AI teams. Responses include AI-assisted review (72.6% of Copilot code review users say it helped, per Octoverse), smaller PRs, stronger tests, and making review capacity — not generation — the planning constraint.