Amazon has discontinued an internal ranking system designed to encourage the use of artificial intelligence after some employees reportedly began generating unnecessary AI activity to improve their standings, leading to higher computing costs for the company.
The decision, first reported by the _Financial Times_, highlights a growing challenge for organisations seeking to accelerate AI adoption while ensuring that usage delivers meaningful business value. As companies invest heavily in AI tools, many are discovering that measuring activity alone can create unintended incentives and increase operational expenses.
The internal leaderboard, known as Kirorank, tracked employee engagement with Kiro, Amazon’s AI-powered software-development platform. The tool was intended to promote awareness and encourage wider use of AI across engineering teams. However, employees reportedly began using AI agents to perform tasks that were not always necessary, largely to boost their position on the rankings.
As a result, the volume of AI tokens consumed increased significantly. Tokens are units used to measure the amount of data processed by AI models, and higher usage directly translates into greater infrastructure and computing costs. According to the report, the behaviour became widespread enough for Amazon to eventually shut down the dashboard.
The incident occurred amid a broader push by Amazon to integrate AI into everyday software development. The company has reportedly set adoption targets that encourage developers to regularly use AI tools, while employees also have access to internal platforms designed to automate coding and engineering tasks.
The episode has drawn attention to a growing issue facing employers: distinguishing between AI adoption and AI effectiveness. While usage metrics can indicate whether employees are engaging with new technologies, they do not necessarily reveal whether those tools are improving productivity, reducing costs or delivering better outcomes.
In response, Amazon is reportedly placing greater emphasis on performance-based measures that assess whether AI is helping engineers produce useful work rather than simply generating more activity. The shift reflects a broader trend as organisations move beyond measuring AI usage and begin focusing on the business impact of their investments.
The development comes as Amazon continues to expand its artificial intelligence capabilities and invest heavily in data centres and AI infrastructure. As enterprises increasingly integrate AI into workflows, the company’s experience serves as a reminder that adoption targets alone may not guarantee efficiency gains. Instead, organisations are learning that the real challenge lies in ensuring that AI use is purposeful, measurable and aligned with business objectives.
