AI Now Writes 27.6% of All Pull Requests, Up From 1% a Year Ago

Fully AI-generated code jumped from 1% to 27.6% of all pull requests over the past year, according to recent industry data, and a 2026 Cortex survey found nearly 90% of engineering leaders say their teams are actively using AI coding tools.

A 27x Increase in a Single Year

Going from 1% to 27.6% is not incremental adoption — it's a fundamental shift in how code enters production codebases at a huge share of engineering organisations, in under 12 months. That pace outstrips almost every other enterprise software adoption curve on record, including the recent Gartner finding that AI agents will be embedded in 40% of enterprise applications by year end.

The Review Bottleneck Nobody Solved Yet

Code generation speed increasing 27x does not mean code review capacity increased at all. Engineering teams that haven't deliberately redesigned their review process for AI-generated volume are likely either rubber-stamping more than they realise, or have quietly become the new bottleneck in their own delivery pipeline — neither is a stable place to be.

What Engineering Leaders Should Do With This Number

Nearly 90% tool adoption with this much AI-generated volume means the question is no longer whether your team uses AI coding tools — it's whether your review standards, test coverage requirements, and security scanning have been deliberately adapted for a world where a quarter of your pull requests started with a model, not a person.