Amazon Discontinues Internal AI Leaderboard Due to Rising Costs and Abuse

Amazon has decided to shut down its internal Kirorank tool, an employee-built system designed to monitor artificial intelligence usage within the company. Launched as a beta project, Kirorank lasted only a few weeks before becoming a financial liability. The primary issue stemmed from 'tokenmaxxing,' where employees used AI to automate routine tasks solely to increase token consumption and climb the leaderboard rankings.

According to Financial Times, the decision to discontinue Kirorank was driven by two key factors: a sharp rise in AI-related costs and inefficient resource allocation. Amazon incurred substantial expenses on tokens that delivered no tangible business value. The company emphasized that the tool was never officially approved and was intended solely to raise awareness about AI capabilities—not to encourage its reckless application.

Amazon is not alone in facing such challenges. Earlier, Meta* also scrapped a similar internal leaderboard after employees began abusing AI to achieve 'Token Legend' status. Uber, meanwhile, admitted it had already exhausted its entire 2026 AI budget within just three months. These incidents highlight how many tech companies still struggle to balance AI adoption with cost control.

Despite some firms scaling back AI spending, overall adoption of generative models continues to surge. For instance, Google reported that token processing for its Gemini system skyrocketed from 480 trillion in May 2025 to 3.2 quadrillion by May 2026. Experts note that companies are still refining AI deployment strategies, and cost-cutting measures do not signal a retreat from the technology.