Amazon has shut down its internal token-usage leaderboard, which had been tracking how much AI staff consumed. Employees were warned to stop chasing dashboard rank over real results.
The leaderboard had inadvertently encouraged "tokenmaxxing" — gaming AI usage metrics to climb internal rankings rather than using the tools to actually get work done.
The episode is a cautionary tale about how measuring AI adoption can backfire. When usage itself becomes the target, employees optimize for the metric rather than the outcome it was meant to represent.
Amazon's move reflects a broader lesson for organizations rolling out AI: tracking raw consumption can reward the wrong behavior, and real productivity gains come from outcomes, not token counts.