Meta Pushes Its New AI Agent on Employees—but Eases Off on Tokenmaxxing
The company is reducing pressure on workers to use artificial intelligence tools while encouraging them to experiment with Hatch, its most advanced AI project yet.
META EASES OFF ON AI USAGE METRICS
Meta announced this week that it is ending the tokenmaxxing incentive program that linked employee performance reviews to the amount of AI tools they used. In an internal memo shared with staff, the company said future evaluations will no longer depend on how many AI tokens an employee consumes. The change arrives as workers begin testing a new agentic AI system called Hatch, which is already driving a rise in token consumption.
Shift in Performance Evaluations
The updated review guidelines replace language that previously measured staff on “usage of AI” and on designations such as “AI Native.” The new wording notes that outcomes can be supported by AI or by other means, shifting the focus back to overall impact rather than the method used to achieve it. Several employees told WIRED that the adjustment, while subtle, relieves the pressure to employ AI in situations where it adds little value. Meta spokesperson Tracy Clayton told WIRED the revisions are meant to reaffirm that contributions, not AI adoption metrics, drive performance ratings. He also said that labels like “AI Native” were never part of formal evaluations. According to a report from The Information, engineers were informed that the firm “will not use AI adoption dashboards or token counts to evaluate impact.”
Labels and Performance Evaluations
The tokenmaxxing era began roughly a year ago when Meta started grading workers on “AI‑driven impact,” effectively rewarding those who logged high usage of chatbots and agents. Employees who amassed large token counts earned tags such as “AI First” or “AI Enabled,” a practice cited in a July lawsuit filed by about two dozen former staff alleging violations of U.S. antidiscrimination law. The lawsuit claims that workers on health or family leave could not accumulate usage and were penalized as a result, a charge Meta denies. At the height of the program, an internal leaderboard tracked token consumption, awarding titles like “Token Legend.” After the dashboard leaked, the creator removed it in April, and Meta later moved to limit AI usage.
Introduction of Hatch
More recently, Meta has been urging—though not obligating—employees to experiment with Hatch, an advanced agentic AI capable of browsing the web and operating other applications autonomously. Staff have been testing the tool on corporate devices for several weeks ahead of a planned public launch. Some workers remain cautious about linking Hatch to personal email, calendars, or other non‑work accounts, citing privacy concerns and the risk of the AI making critical errors. Meta previously ran a project that recorded keystrokes and other activity on employee devices to train AI models; that effort has since been paused, further eroding trust in the company’s AI initiatives.
Employee Concerns and Morale
Eliminating the threat of a lower performance rating for opting out of AI testing could improve morale and help rebuild confidence among the workforce. A number of employees have already adopted Hatch for personal tasks such as scheduling appointments and organizing life outside of work, describing it as effective. Nonetheless, they voice worries about the environmental and economic costs of consuming large volumes of tokens and computing power to automate routine duties. Even without the incentive to maximize token counts, Hatch still uses more resources than traditional chatbots or coding assistants.
Future Productivity and Job Cuts
If Hatch delivers the promised productivity gains, some staff fear it could trigger additional layoffs. Reuters reported last week that Meta was preparing another round of major cuts beyond the 8,000 jobs eliminated in May, but the plan was halted amid a surge in software bugs tied to AI reliance and slower progress on agent development. CEO Mark Zuckerberg has said further mass layoffs are not expected this year.
Source: wired.com · 2026-09-03