The race for AI supremacy sometimes leads to absurd situations inside the biggest tech corporations. According to journalists’ findings, internal AI usage metrics at Amazon are being artificially pumped up by the staff itself.
At the center of the controversy is MeshClaw – an internal Amazon platform inspired by OpenClaw. The tool allows employees to run local AI agents that automate daily, repetitive tasks and can integrate with third-party apps like Slack or email clients. Amazon officially boasts that thousands of employees use the assistant every day.
However, three employees revealed that some staff members are using the MeshClaw platform to perform completely nonsensical tasks. The goal is singular – to generate as much traffic as possible and join an internal trend dubbed “tokenmaxxing.” Employees are burning through massive amounts of digital tokens to climb to the top of company activity leaderboards and curry favor with the bosses.
Insiders clearly indicate that the work environment forces this artificial padding of stats, describing the phenomenon as:
“perverse incentives”
The sheer pressure to adopt the new technology across teams is reportedly massive:
“there is just so much pressure to use these tools.”
Amazon’s management is pushing back against the allegations. Corporate representatives insist that the number of AI tokens consumed is not factored into individual employee performance reviews. Sources indicate, however, that the reality in the offices drastically deviates from the official guidelines. Additionally, the initially public usage stats for the MeshClaw platform were recently hidden by company administrators.
The widespread and often unnecessary use of AI agents raises further security concerns. Some staff members are pushing back against handing machines such a broad scope of responsibilities, highlighting the specific risk of factual errors and algorithmic hallucinations.