AI is approving our pull requests: Here's how we made it safe

Kesha Mykhailov and Niamh Young make the case that review became the bottleneck once agents started writing most of the code, and that the answer is to automate the gate rather than slow the line. Over 93% of Intercom’s PRs across their two main codebases are now agent-driven, and over 19% are auto-approved with no human reviewer in the loop. In the first four weeks of broader rollout, 497 PRs went fully autonomous, with Claude writing the code and their AI approval system reviewing, approving and shipping to production.

The review agent splits the job into sub-tasks handled by independent agents, each looking at a different thing: whether the change matches the problem description, safety concerns, logical correctness, best practices. Strict size and complexity limits keep large changesets out of auto-approval, and every decision is logged, labelled and auditable. Engineers can ask for a human at any point.

The safety claim rests on revert rates: AI-authored backend code was reverted at 0.53% against 5.39% for human-authored, and on the frontend 0.22% against 2.00%. They also report a 6 to 16x improvement in time-to-approval at the 75th percentile. Their framing is deliberately provocative, and the line that does the work is Human review is not a guarantee of safety. It never was.

Worth reading next to their piece on 2x engineering velocity and the good skills, bad skills write-up. The usual caveat applies: these are self-reported numbers from a company with a strong interest in the conclusion, and a revert rate measures what got noticed and rolled back, not everything that went wrong.