17 PRs a Day, One QA: How We Automated E2E Failure Triage

Short summary

A QA engineer at pdf.net describes how their team of 18 AI-augmented developers merging 17 PRs/day automated E2E failure triage using a GitHub Action powered by Claude. The bot collects failed tests, logs, deploy ranges, and PR diffs, then classifies failures as PR-related, non-PR-related, or insufficient-data, creating Linear tickets and Slack threads automatically. The system costs less per month than one hour of engineering time and handles flake detection with auto-reruns.

• •Automated E2E triage bot handles 17+ PRs/day with one QA engineer
• •Claude classifies failures as PR-related, non-PR-related, or insufficient-data, creating tickets and Slack threads
• •Full system costs less per month than one hour of engineering time

Generated with AI, which can make mistakes.