AI amplifies whatever is already true of a team.
Amazon reportedly pulled engineers into a deep review this week after a run of outages where AI-assisted changes were cited as a factor. The company has since tightened approvals, documentation and safeguards around critical systems.
Customers do not experience each individual change. They experience the total number of things that go wrong. So when AI increases output, but review, testing and rollout discipline stay the same, more mistakes reach the customer.
Engineering sees this first because failures are obvious. But the same pattern applies anywhere AI increases output without increasing control.
AI makes things faster. That only helps if you can keep up.
© 2026 Thomas Wainwright