The Bottleneck Moved
AI coding tools promised faster development. They delivered — sort of. Developers using AI complete 21% more tasks and merge 98% more pull requests. But PR review time has increased 91%. The bottleneck didn't disappear. It relocated from writing code to verifying it.
LinearB's 2026 analysis of 8.1 million pull requests across 4,800+ organizations found that developers feel 20% faster but are actually 19% slower. That's a 39-point gap between perceived and actual productivity.
The Numbers Are Stark
Sonar's 2026 State of Code Developer Survey of 1,100+ developers confirmed that AI accounts for 42% of all committed code — a number developers expect to reach 65% by 2027. Yet 96% of developers say they don't fully trust AI-generated code to be functionally correct, and only 48% report that they always verify it before committing.
The Pragmatic Engineer's 2026 survey described a specific archetype — the "Builder" — who is most overwhelmed by reviewing AI-generated code from colleagues. Some teams are seeing 30 PRs per day with only six reviewers. That ratio is unsustainable no matter how you look at it.
Meanwhile, Lightrun's 2026 report found that 43% of AI-generated code changes require manual debugging in production even after passing QA and staging. Zero percent of surveyed engineering leaders described themselves as "very confident" that AI-generated code will behave correctly once deployed.
What Actually Helps
The teams handling this well share a few traits:
- They enforce PR size limits regardless of generation speed. AI makes it trivially easy to produce massive diffs. That doesn't mean reviewers can absorb them.
- They use AI-assisted review as a first filter. The emerging "review sandwich" approach — AI catches surface-level issues first, humans focus on architecture and business logic — reduces human review time by 30-50% according to GitHub's internal data.
- They track review metrics as seriously as shipping metrics. If you measure deployment frequency but not time-to-review or review queue depth, you're flying blind on half the pipeline.
- They centralize visibility across repos. When PRs are scattered across dozens of repositories and providers, stale reviews become invisible. Tools like Code Board exist specifically to aggregate PRs into a single view with risk scoring, so teams can spot bottlenecks before they compound.
The Real Question
The teams that are shipping well in 2026 aren't the ones generating the most code. They're the ones whose review culture and tooling adapted to match their new output velocity. If your PR volume doubled this year but your review process didn't change at all, that's where your next problem is hiding.
Top comments (0)