Meta is officially ending its employee “tokenmaxxing” incentive program. The social media company told employees in an internal announcement this week that performance reviews will no longer be based on how extensively they use AI tools, according to three Meta employees who spoke with WIRED. However, Meta is also testing a new AI agent called Hatch, while overall employee AI token usage continues to rise.
Nearly a year ago, Meta announced that employees would be evaluated based on their “AI-driven impact.” In practice, that meant assessing how much employees used chatbots and AI agents to improve their daily work, Business Insider reported at the time. According to the report, higher AI usage was linked to employee labels such as “AI native,” “AI first,” and “AI enabled.” A lawsuit filed in July by approximately 20 employees alleges that Meta violated U.S. anti-discrimination laws when it fired workers in May. Employees who were on medical or family leave claim they could not accumulate comparable AI usage and were unfairly penalized. Meta denies the allegations in the ongoing case.
The latest performance-review guidance replaces references to criteria such as “AI usage” and the “AI native” designation with broader language stating that achievements “may be supported by AI or other means.” The change shifts Meta’s focus toward the impact of an employee’s work rather than the tools used to produce it. Several Meta employees told WIRED that the update is subtle but welcome because it reduces pressure to use AI when it does not provide a meaningful benefit.
Tracy Clayton, a Meta spokesperson, told WIRED that the update reinforces the company’s longstanding policy of evaluating employees based on their contributions. He also said labels such as “AI native” were never used to determine performance ratings. Engineers across Meta were told this week that the company “will not be using AI implementation dashboards or token counts to assess impact,” The Information reported on Wednesday.
For months, some Meta employees said colleagues encouraged them to use AI tools repeatedly—even for tasks they considered unnecessary—to increase internal usage metrics, including token counts and AI units consumed. The practice became known as “tokenmaxxing.” It reached a peak when a Meta employee created an internal leaderboard ranking AI usage and gave employees titles such as “token legends.” After details of the dashboard were leaked, an employee shut it down in April. A few months later, Meta moved to limit employee AI usage as the company faced rising costs.
Today, Meta encourages—but does not require—employees to experiment with advanced AI agents, including a tool called Hatch. Similar to the viral OpenClaw, Hatch can browse the web and interact with other applications autonomously. Meta employees have been testing the AI agent on internal devices in recent weeks ahead of its expected public release.
Some employees remain reluctant to connect Hatch to personal email accounts, calendars, and other non-work services because of privacy concerns and the risk that an AI agent could make an important mistake, according to two current Meta employees familiar with the issue. Meta previously operated a project that tracked keystrokes and other activity on employees’ work devices to gather data for AI training. Although the company has since suspended the project, employees said it damaged trust in Meta’s workplace AI tools.
Source: www.wired.com