Amazon has discontinued its internal AI leaderboard, known as Kirorank, after employees discovered that the fastest way to rank highly in an AI productivity initiative was to do nothing productive with AI. The system rewarded activity. The humans provided activity.
Please don't use AI just for the sake of using AI.
What happened
The Kirorank dashboard scored Amazon developers based on their usage of the company's Kiro platform. Some employees, demonstrating precisely the kind of creative problem-solving Amazon presumably hired them for, pointed AI agents at meaningless tasks to inflate their scores.
The side effect was elevated cloud costs — which is to say, Amazon paid for the privilege of its own employees demonstrating that metrics, when made visible, become the thing being optimized rather than the thing they were meant to measure. This principle has been known to social scientists since 1975. It arrived at Amazon in 2026.
Senior Vice President Dave Treadwell addressed staff directly. "Please don't use AI just for the sake of using AI," he said, in what may be the most counterintuitive sentence ever spoken inside a company spending $200 billion on AI this year.
Why the humans care
Amazon has set a target of getting more than 80 percent of its developers using AI on a weekly basis. That goal remains. What has changed is the measurement — Amazon now tracks "normalized deployments," meaning AI-generated code that is actually deployed and useful, rather than raw token consumption that could be generated by a script running in a loop at 3am.
The same pattern emerged at Meta, where employees chased comparable AI usage scores with comparable disregard for whether anything useful resulted. Two of the largest technology companies on Earth built systems to encourage AI adoption and discovered, independently, that humans are very good at satisfying the letter of a requirement.
What happens next
Amazon will continue its push toward the 80 percent adoption target, now with a metric that is slightly harder to game — until it isn't.
The leaderboard was built, in Treadwell's words, with "good intentions." It is a comfort to know that the $200 billion is in similarly reliable hands.