Meta has ended the internal program that tied employee performance reviews to how much they used AI tools, according to three employees who received the announcement this week. The shift comes as Meta simultaneously pushes staff to test Hatch, its most advanced agentic AI project, and after May layoffs cut 8,000 jobs at the company. New review guidance drops references to 'usage of AI' and 'AI Native' designations, replacing them with wording that notes outcomes 'can be supported by AI or other means.'
The previous regime, which Meta rolled out almost a year ago, told workers they would be graded on 'AI-driven impact' — in practice, how much they leaned on chatbots and agents in day-to-day work. Usage correlated with internal labels including 'AI Native,' 'AI First,' and 'AI Enabled.' About two dozen former employees sued Meta in July, arguing the company violated US antidiscrimination law when it laid them off in May, in part because workers on health and family leave could not accumulate the usage that drove those labels. Meta has denied the allegations, and the case is ongoing.
Meta spokesperson Tracy Clayton framed the update as a clarification rather than a reversal, saying the company has always evaluated employees on their contributions and that labels such as 'AI Native' were never used for performance evaluation. That characterization sits awkwardly next to the plaintiffs' filing, which cites those exact labels as central to how staff were sorted ahead of the May cuts.
“Meta evaluates employees based on their contributions.”— Tracy Clayton, Meta spokesperson
Key facts
- 01Meta told employees this week that performance reviews will no longer depend on AI tool usage, replacing 'AI Native' criteria with looser wording.
- 02About two dozen employees sued Meta in July, arguing usage-based 'AI Native,' 'AI First,' and 'AI Enabled' labels contributed to discriminatory May layoffs of 8,000 workers.
- 03Meta is internally testing Hatch, an agentic tool that browses the web and operates apps, positioned against OpenClaw ahead of a public release.
- 04Meta called off a second round of major layoffs after AI-linked software bugs rose and agent development slowed.
- 05CEO Mark Zuckerberg has said further mass layoffs are not expected this year.
Engineers were also told this week that adoption dashboards and token counts would not be used to grade their work, a message reported inside the company by The Information on Wednesday. Employees describe the change as subtle but meaningful: for months, colleagues had prompted AI tools repeatedly, sometimes frivolously, to maximize measured token consumption. The high point of that era came when one worker built an internal leaderboard tagging heavy users as 'Token Legend'; after details leaked, the employee took the dashboard down in April, and Meta moved to ration AI usage a couple of months later.
The pullback on usage-based grading does not mean Meta is easing off AI internally. The company is now encouraging — though not requiring — employees to test Hatch, an agent that can autonomously browse the web and operate other applications on a computer. Staff have had access on corporate devices for the past several weeks ahead of a planned public release. Meta declined to comment on the testing.
Hatch is positioned squarely against OpenClaw, the viral agent that has set the reference point for consumer-facing computer-use tools. Some Meta employees say they have started using Hatch for personal tasks like booking appointments and organizing schedules and describe it as effective. Others have declined to connect it to their personal email, calendar, and non-work accounts, citing privacy concerns and the risk that the agent could break something important.
That hesitation has roots. Meta previously ran a project that tracked employees' keystrokes and other activity on their work devices to gather training data for AI systems. The company has since paused that effort, but employees say it damaged internal trust in Meta's AI tooling — trust that removing the token-count pressure may help rebuild.
Even without a scoreboard, Hatch is heavier on compute than the chatbots and coding assistants it is replacing. Employees say token consumption keeps rising and that management still wants to see staff demonstrate they can put AI tools to work. Some worry about the environmental and cost trade-offs of running agents on routine tasks, and about what higher measured productivity implies for headcount.
Reuters reported last week that Meta had prepared another round of major layoffs on top of the 8,000 May cuts, then called off the plans after software bugs linked to AI reliance increased and agent development moved more slowly than expected. Mark Zuckerberg has said further mass layoffs are not expected this year, a commitment employees are watching closely as Hatch expands.
The retreat from token-based grading is a quiet admission that the metric was measuring the wrong thing. Counting AI calls rewards ceremony over output, and, per the July lawsuit, it created a review artifact that penalized workers who were physically absent from the tooling. Meta's shift back to grading on impact is the more defensible position, but it also puts more weight on Hatch actually delivering — because if the agent boosts real productivity, the pressure to translate that into headcount reductions will return regardless of what Zuckerberg has said about this year.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
