AI testing is shifting from test generation to maintenance and governed intent

Independent evidence that AI or codeless tools reduce test upkeep remains scarce. QA Wolf reports triage falling from 2.3 hours to 21 minutes and 75% autonomous resolution, without sufficient methodology; a Cypress migration case reports flakiness falling from 15% to under 0.5% across 200+ nightly runs, also self-reported. A LocalStack-sponsored survey highlights staging-validation bottlenecks, but its estimates are not independent evidence of reduced maintenance.