AI Merged 98% More PRs. Review Time Went Up 91%.
Faros.ai measured high-AI-adoption teams and found the speed didn't disappear — it moved. Code got written faster and piled up at the one stage nobody automated: human review.
NeuroX AI · June 24, 2026
The most honest AI-productivity number of 2026 isn't about how fast code gets written. Faros.ai instrumented teams with high AI adoption and found they merge 98% more pull requests — while PR review time climbs 91%. The work didn't vanish. It moved downstream to the one stage nobody automated.
The mechanism is simple. Agents draft code at machine speed, so PRs get both more numerous and bigger — Faros clocks a 154% jump in average PR size. A reviewer who used to read a 40-line diff now faces a 100-line one, twice as often. Throughput at the keyboard became a queue at the review gate. And the diffs are riskier: the same data shows a 9% increase in bugs per developer as adoption rises.
Here's the part that should stop a VP: at the company level, Faros found no significant correlation between AI adoption and delivery improvement. Individual developers complete 21% more tasks; the org ships the same. The gain evaporates in the review backlog before it reaches a customer.
This is the prototype-to-production gap wearing a new hat. Generating code was never the bottleneck. Verifying it under real constraints is — and that's exactly the part a demo skips and a production system can't.