Amazon employees are reportedly using an internal AI tool to automate non-essential tasks, partly to demonstrate active AI usage to managers, according to a report by Financial Times. The e-commerce giant has set targets and begun tracking how much its staff uses the technology, creating significant pressure among workers.
MeshClaw: The Internal AI Tool
Amazon has rolled out an internal tool called MeshClaw, which allows employees to build AI agents that connect to workplace apps and perform tasks automatically. The tool can send emails, manage Slack messages, and handle code work. More than three dozen Amazon staff built it, and the company says thousands of employees use it daily to reduce repetitive work.
However, some employees have found another use: running MeshClaw to generate unnecessary AI activity purely to increase their "token" count — a measure of how much data the AI processes. Internal leaderboards track this usage across teams, creating what one employee called "perverse incentives."
Safety Concerns Raised by Employees
Beyond the pressure to perform, some Amazon employees have raised safety concerns. MeshClaw requires permission to act on a user's behalf — sending messages, making deployments, and taking actions across apps. For some staff, that is a step too far.
"The default security posture terrifies me," one employee told the FT. "I'm not about to let it go off and just do its own thing."
Amazon, in a statement, said it remains "committed to the safe, secure and responsible development and deployment of generative AI." The company described MeshClaw as a tool that "empowers teams" to experiment with AI in their everyday work.
The Trust Gap Between Amazon and Its Staff
Amazon told its workers that token statistics will not be used in performance evaluations. Yet multiple employees told the FT they believe managers are watching the numbers anyway.
"Managers are looking at it," one current employee told the publication. "When they track usage it creates perverse incentives and some people are very competitive about it."
Another employee said: "There is just so much pressure to use these tools. Some people are just using MeshClaw to maximise their token usage."
Amazon has since restricted access to team-wide AI usage stats. Now only individual employees and their direct managers can see the data. Managers are also officially discouraged from using token use as a performance measure — but the concern among staff remains.
Not Just Amazon
Meta employees have reportedly done something similar, a practice some are calling "tokenmaxxing" — deliberately inflating AI usage stats to rank higher on internal leaderboards.
The trend reflects a wider push across Silicon Valley to show that massive AI investments are paying off. Amazon alone is expected to spend $200 billion on capital expenditure this year, with the bulk going toward AI and data centre infrastructure. Companies need their staff to actually use AI tools — and they are tracking whether they do.
