Business
Amazon Discontinues AI Leaderboard After Employees Used It for Pointless Tasks
Amazon is ending its internal AI leaderboard, 'Kirorank,' after employees used it to inflate scores with meaningless tasks, according to the Financial Times.
Amazon has decided to stop using its internal AI ranking system, known as 'Kirorank,' after employees exploited it to boost their scores through unnecessary AI activities. The dashboard, which measured employee engagement with Amazon's Kiro developer platform, was reportedly used by some workers to point AI agents at trivial tasks, thereby increasing their rankings without contributing meaningful value. Senior Vice President Dave Treadwell reportedly warned staff, "Please don't use AI just for the sake of using AI," stating that the dashboard was built with "good intentions" but ultimately created additional costs for the company. The move comes as Amazon aims to have more than 80 percent of its developers use AI weekly and plans to spend around $200 billion in 2026, primarily on AI infrastructure. Similar issues were observed at Meta, where employees also pursued AI usage scores. Amazon now tracks "normalized deployments," focusing on AI-generated code that provides actual utility. *Source: [thedecoder](https://the-decoder.com/amazon-kills-internal-ai-leaderboard-after-employees-gamed-it-with-pointless-tasks/)*
Key points
- Amazon is ending its internal AI leaderboard, 'Kirorank,' after employees used it to inflate scores with meaningless tasks.
- The dashboard measured employee engagement with Amazon's Kiro developer platform.
- Senior Vice President Dave Treadwell reportedly warned staff not to use AI just for the sake of using AI.
- The dashboard was built with 'good intentions,' but ended up creating extra costs for the company.
- Amazon aims to have more than 80 percent of its developers use AI weekly and plans to spend around $200 billion in 2026 on AI infrastructure.
- Similar issues were observed at Meta, where employees also pursued AI usage scores.
- Amazon now tracks 'normalized deployments,' focusing on AI-generated code that provides actual utility.