Amazon employees are inflating usage data of an internal AI tool called MeshClaw by automating non-essential tasks to maximize AI token consumption, a practice known as "tokenmaxxing" [1, 2, 3]. The company requires over 80% of its developers to use AI tools weekly and tracks AI token usage on internal leaderboards [1, 3].
MeshClaw, developed by a 36-person engineering team, enables users to create AI agents that connect with workplace software to automate workflows. Thousands of Amazon employees reportedly use it daily to streamline tasks [2, 3].
While Amazon states that AI token usage data is not officially part of employee performance evaluations, many employees believe managers track this information informally. "Managers are looking at it. When they track usage it creates perverse incentives and some people are very competitive about it," said one anonymous current employee [1, 2, 3]. Another source added, "There is just so much pressure to use these tools. Some people are just using MeshClaw to maximize their token usage" [1].
Some employees express unease over automating excessive tasks because of worries about AI hallucinations or factual errors introduced by the tools [2]. Similar tokenmaxxing behaviors have been reported at other tech companies, including Meta and Microsoft [3].
In May 2026, Amazon widely deployed MeshClaw internally alongside enforcing AI token consumption tracking and weekly usage targets, setting a clear mandate on developers to engage heavily with AI tools [1, 3]. The company plans to spend $200 billion this year, mainly on AI and data center infrastructure, signaling a deep commitment to AI-driven workflows [1].
Amazon will continue monitoring AI token consumption as it rolls out further AI initiatives and advances its data infrastructure investment throughout 2026 [1].
