One number from the Faros 2026 Report stands out beyond the headline: code merged without review rose 31%.
That's not a minor change in process. It's a structural shift in how software reaches production — and it has direct consequences for QA teams.
Why Review Rates Are Declining
AI coding assistants create pressure on the review process in two ways:
1. Volume outpaces capacity. When individual engineers ship 30-50% more code, the total PR volume increases without a proportional increase in reviewers. Reviewers become a bottleneck. Teams relieve the pressure by merging with lighter review — or none at all.
2. AI code looks clean. AI-generated code is syntactically correct, well-structured, and follows patterns. Reviewers scan it quickly because it looks right. The semantic errors — API assumptions, missing null guards, state dependencies — aren't visible in a quick scan. They only appear in production with real data.
The result: more code reaching production, with less scrutiny, with a higher subtle-bug rate.
What This Means for QA
For QA teams, the math is straightforward: more bugs will reach production. The question is whether your bug reporting infrastructure is equipped to handle them efficiently.
The traditional QA model — find the bug, file the bug, wait for the developer to investigate, wait for the fix, verify it — assumes bugs are rare and each one gets significant attention. With bugs per developer up 54%, that model creates a permanent backlog.
The adapted model requires:
1. Bug Reports That Front-Load the Investigation
Every bug report should arrive with enough context that the developer can go straight to the fix. No investigation phase. The page state, API payloads, and error trace should be in the report, automatically.
2. Auto-Triage That Routes Immediately
With more bugs entering the system, the triage step becomes a bottleneck. AI-powered auto-triage — priority assignment, smart routing to the right developer — removes the human triage step from the critical path.
3. Duplicate Detection That Clears the Backlog
More bugs at higher volume means more duplicates. Semantic duplicate detection — matching by meaning, not just keywords — clears duplicate reports automatically before they clutter the backlog and waste developer time.
The QA Opportunity
There's an opportunity hidden in the Faros data. Teams that build the right infrastructure now — capture tools with full context, auto-triage, duplicate detection — will turn the AI era's higher bug rate into a competitive advantage.
Those teams will ship faster because each bug costs less to fix. Their developers will spend time coding, not investigating. Their QA teams will file bugs that get fixed on the same day, not after a week of back-and-forth.
The 31% decline in code review is a signal that review as the primary quality gate is weakening. The replacement quality gate is production bug capture with full context — so that when bugs do reach users, the path from report to fix is measured in minutes.
Further reading:
