Amazon bins an internal AI leaderboard for its Kiro employees, because they were burning through too many costly tokens
Amazon has deprecated an internal leaderboard that tracked employee usage of its Kiro agentic AI development platform. The competitive ranking system, which displayed how much individual staff members were using the company’s proprietary AI tool, has been removed due to unexpectedly high token consumption costs.
According to reporting from the Financial Times, the leaderboard produced the opposite effect from its intended purpose. Rather than providing useful performance metrics, the competitive ranking system incentivized heavy usage of the platform, causing employees to burn through costly API tokens at an unsustainable rate. Since token-based AI services charge per usage unit, the combination of workplace competition and unrestricted resource consumption created a significant financial drain.
The situation illustrates a broader challenge facing enterprises rapidly adopting AI tools: managing costs while encouraging employee engagement with new technology. Amazon’s experience demonstrates how seemingly innocuous workplace gamification—in this case a simple usage leaderboard—can create perverse incentives that drive up operational expenses.
Amazon has positioned itself as one of the most enthusiastic enterprise adopters of AI, integrating it across numerous internal operations and customer-facing services. The Kiro platform represents one such initiative, offering developers an agentic AI tool for building applications. However, the company appears to have learned a valuable lesson about balancing innovation encouragement with cost controls when dealing with token-based API pricing models. The deprecation of the leaderboard marks a tactical retreat from one specific implementation, though Amazon’s broader commitment to AI development continues.