Meta told employees that performance evaluations would no longer be dependent on how much they used AI tools as it pushes testing of a new agent called Hatch across the company.
The guidance undoes earlier language that had encouraged staff to document "AI-driven impact" and had been linked in practice to labels such as AI Native and an internal era of "tokenmaxxing," when workers repeatedly prompted chatbots to inflate usage metrics.
Contribution, not KPIs
Meta spokesperson Tracy Clayton said the updates were meant to emphasise that the company evaluates employees on their contributions and that labels like AI Native were never used for performance evaluation.
Engineers were told the company would not use AI adoption dashboards or token counts to evaluate impact, shifting review wording back toward assessing outcomes rather than raw tool consumption.
Tokenmaxxing era over?
Workers say the tokenmaxxing era included an internal leaderboard tracking consumption and prompted Meta to later ration employees’ AI usage, practices that many said eroded trust.
Meta has made Hatch available on corporate devices for weeks; the agent can autonomously browse the web and operate other applications as it pursues tasks, and staff are testing it ahead of an expected public release.
Some employees told reporters they hesitate to connect personal email or calendars to Hatch because of privacy worries and the risk the agent could "mess up" important items, concerns sharpened after the company once tracked employees’ keystrokes for model training.
