Meta Promotes Its Latest AI Assistant to Employees While Reducing Focus on Tokenmaxxing

Meta is officially discontinuing what had become a tokenmaxxing incentive program for its staff. In an internal memo issued this week, the social media giant informed employees that their performance assessments would no longer rely on the extent of their AI tool usage, according to three employees who shared the information with WIRED. However, as Meta employees start testing a new agentic AI tool called Hatch, they report an escalation in their token usage.

Nearly a year ago, Meta announced that employees would be evaluated based on their “AI-driven impact,” which effectively meant assessing how much they utilized chatbots and agents to enhance their daily tasks, as reported by Business Insider at the time. According to a lawsuit filed in July by around two dozen employees, which alleges that Meta breached US antidiscrimination laws during layoffs in May, such usage led to labels like “AI Native,” “AI First,” or “AI Enabled.” (Workers on health and family leaves claimed they couldn’t achieve usage and were unjustly penalized; Meta has denied these allegations amid the ongoing case.)

The new performance review guidelines introduced this week eliminate references to evaluating employees based on factors like “usage of AI” and their “AI Native” status, replacing them with broader language that notes “these outcomes can be supported by AI or other means.” This refocuses Meta’s evaluations on assessing employee impact, irrespective of how it’s achieved. Some employees expressed to WIRED that these changes, though subtle, are appreciated as they relieve the pressure to use AI in situations where it may not be appropriate.

Meta spokesperson Tracy Clayton informed WIRED that this week’s updates aim to reinforce what has always been true—employees are judged based on their contributions. He noted that labels like “AI Native” were never intended for performance assessments. The Information reported Wednesday that engineers throughout the company were informed that the company “will not utilize AI adoption dashboards or token counts to gauge impact.”

For months, numerous Meta employees alleged that their peers prompted AI tools excessively—or what they considered unnecessarily—to inflate internal usage metrics, such as the volume of tokens or AI units consumed. The tokenmaxxing trend peaked when one Meta employee established an internal leaderboard to track AI usage, designating top users as “Token Legends.” After details about the dashboard were leaked, the employee removed it in April. A few months later, Meta decided to place limits on employees’ AI usage.

Recently, Meta has been promoting—but not mandating—employees to utilize its latest AI experiment, an agentic tool capable of autonomously executing tasks on a computer. Named Hatch and reminiscent of the popular OpenClaw, it can browse the web and interact with other software applications. Employees have been testing Hatch on their corporate devices for several weeks as they anticipate a public release.

Some employees have been cautious about linking Hatch to their personal email, calendar, and other non-work-related accounts due to privacy concerns and the risk of the AI potentially causing issues, according to two current Meta employees familiar with these worries who spoke with WIRED. Previously, Meta had conducted a project that monitored employees’ keystrokes and activities on their work devices to gather data for training AI systems. This initiative, which Meta has since halted, diminished many employees’ trust in the company’s AI tools, the sources revealed.