Amazon removed an internal AI ranking dashboard after employees inflated their scores by running trivial AI tasks, a practice that increased the company’s internal cloud costs.
Amazon dismantled an internal AI leaderboard after staff inflated their rankings by running trivial or pointless AI tasks, a behavior that pushed up internal cloud spending. The removal responds to concerns that the leaderboard encouraged usage for scoring rather than productive engineering, creating costly activity with little business value.
The dashboard, known inside the company as Kirorank, ranked employees based on activity on Amazon’s Kiro developer platform. Some workers reportedly directed AI agents at meaningless jobs to climb the leaderboard. Senior Vice President Dave Treadwell warned employees, “Please don't use AI just for the sake of using AI,” saying the leaderboard had been created with “good intentions” but had added extra costs.
The change comes amid a broad internal push to expand AI usage: Amazon aims to have more than 80% of its developers use AI weekly and is planning roughly $200 billion in 2026 spending focused largely on AI infrastructure. To better measure value, the company said it will replace raw token metrics with “normalized deployments,” a metric meant to reflect AI-generated code and deployments that deliver real results. A similar scoring chase has been observed at other large tech firms.
By scrapping Kirorank and moving to normalized deployments, Amazon intends to curb wasteful token consumption and realign incentives toward practical outputs. The episode highlights a familiar tension in enterprise AI rollouts: large infrastructure commitments and usage targets can create perverse incentives unless measurement emphasizes usefulness rather than activity alone.
Sources
Replies (0)
No replies in this topic yet.
