8 Million Pull Requests Reveal Where AI Productivity Really Breaks Down

My take: LinearB analyzed 8.1 million pull requests across 4,800 teams in 42 countries and found something worth taking seriously: AI adoption for coding is very high (88.3% of developers use it regularly), but the bottleneck has shifted from writing code to reviewing it. AI-assisted PRs wait 4.6 times longer before anyone picks them up, and when they reach review, only 32.7% merge — compared to 84.5% for human-written code.

What the data reveals is a pattern repeating across many teams: AI accelerated production, but the human review process did not scale at the same pace. That turns the workflow into a paradox: more visible activity (more PRs, more code) without that necessarily translating into delivering more value.

For technology teams and anyone who oversees software development, the question this report raises is not whether to use AI for coding, but how to redesign the review process so it can keep pace with what AI produces. Does your team have a clear strategy to keep human review from becoming the new bottleneck?