Meta Drops AI-Usage Metrics From Performance Reviews While Pushing New Agent Hatch
Meta has told employees their performance reviews will no longer hinge on how much they use AI tools, even as the company encourages staff to test its new autonomous agent, Hatch, which continues to drive up token consumption.
Meta this week told employees in an internal announcement that performance evaluations will no longer be tied to how heavily they use the company's AI tools, according to three employees who received the message, Wired reports.
About a year ago, Meta had said workers would be graded on their "AI-driven impact," which in practice meant assessing how much they relied on chatbots and agents in daily work. Heavy users received labels such as "AI Native," "AI First," or "AI Enabled." The practice became central to a lawsuit filed in July by roughly two dozen former employees who allege Meta violated US antidiscrimination laws during May layoffs, arguing that workers on health or family leave couldn't accumulate usage and were unfairly penalized as a result. Meta denies the allegations.
New performance-review guidance issued this week drops references to "usage of AI" or "AI Native" status, instead noting that desired outcomes "can be supported by AI or other means." Some employees welcome the shift, saying it removes pressure to use AI in situations where it isn't useful. A Meta spokesperson said the company has always evaluated employees based on their contributions rather than AI labels.
End of the tokenmaxxing era
For months, some employees reportedly prompted AI tools repeatedly—sometimes seen as frivolous—to boost internal usage metrics such as token consumption. The trend peaked when one employee built an internal leaderboard ranking colleagues with titles like "Token Legend"; it was taken down in April after details leaked. Meta subsequently began rationing employees' AI usage.
A new agent, Hatch
Meta has more recently been encouraging, though not requiring, employees to test its newest AI experiment: an autonomous agent called Hatch, similar to the widely discussed tool OpenClaw, capable of browsing the web and operating other applications on a computer. Some employees remain hesitant to link Hatch to personal email or calendars due to privacy concerns and fear of mistakes, distrust reinforced by an earlier, since-paused Meta project that tracked employees' keystrokes to gather AI training data.
Even without pressure to maximize usage, Hatch still consumes more computing resources than standard chatbots, employees say. Some worry that if Hatch significantly boosts productivity, it could lead to further layoffs. Last week, Reuters reported Meta had prepared another major round of job cuts following May's 8,000 layoffs but shelved the plan amid rising software bugs linked to AI reliance and slower-than-expected progress on agent development.
