AI sped up coding but broke QA: 64% of defects caught before prod in January

alex_verem · x · 2026-10-11

After rolling out AI across a three-year-old client platform (coding assistance plus AI review on every PR), implementation throughput rose—but validation capacity didn't, and QA became the bottleneck.
In January, QA caught 64% of defects before they reached production. Coding got faster; everything after the code (review, QA, rework) absorbed the extra time.
The author links this to McKinsey's 2026 finding: 80% of respondents say AI boosted productivity, but only 37% see it show up in EBIT.
The fix wasn't a better model but the surrounding system: a QA retrospective, a tiered quality-gate plan, and integration tests against a real database instead of in-memory stand-ins.

More from coding & agent