Amazon Managers Press Staff To Boost AI Use, Workers Invent Tasks
Amazon employees are reporting new pressure to increase their use of artificial intelligence tools at work, and some say the push is leading workers to invent unnecessary tasks simply to show higher AI activity.
The accounts, reported by Fast Company, describe a workplace environment in which AI usage is being measured and compared, turning what was framed as productivity support into something closer to a performance metric. A separate report by hcamag.com said an internal AI “leaderboard” tied to employee activity was built by HR, encouraging workers to compete on usage. Fortune, citing the broader debate around the practice, highlighted concerns that “tokenmaxxing” could gamify AI use in ways that may not be healthy for employees or productive for the business.
The details in the reports point to a common dynamic: when usage becomes a number that can be tracked, employees adjust their behavior to improve that number. In this case, workers described running extra prompts, generating material they may not need, or routing work through AI even when it adds little value, just to demonstrate they are engaging with the tools.
The development matters because Amazon is one of the country’s largest employers and a major force in corporate adoption of generative AI. How the company measures and incentivizes AI use could influence not only internal productivity and culture, but also how other large employers roll out similar programs.
It also raises questions about what companies should be measuring in the early days of generative AI integration. Counting prompts, tokens, or usage frequency can be easy to track, but those measures don’t necessarily align with better outcomes. If employees feel they are being judged on AI activity itself, rather than the quality and impact of their work, it can create inefficiencies and add stress, while muddying any true assessment of whether the tools are helping.
The reports also suggest potential risks for managers trying to evaluate performance fairly. Metrics that can be “gamed” may disadvantage employees whose roles don’t naturally involve heavy AI usage, or who are cautious about when and how to use such tools. They may also encourage behaviors that increase internal noise, including unnecessary documents and extra steps in routine workflows.
What happens next will depend on how Amazon responds to the reported effects of its tracking and competition mechanisms, including whether the company changes how it encourages AI adoption, how it measures engagement, and how managers interpret those signals. The broader corporate world will be watching for what sticks: usage-based scorekeeping or outcome-based evaluation.
For now, the reports underscore a familiar lesson of modern workplace management: when a metric becomes the goal, employees will find ways to hit the number—even if the work itself doesn’t need to be done.
