Key Takeaways
- Meta has formally discontinued its internal “tokenmaxxing” incentive program, removing AI‑tool usage from performance‑evaluation metrics.
- The change follows employee complaints and a pending lawsuit alleging that AI‑usage labels unfairly penalized workers on health or family leave.
- New guidance tells managers to assess impact broadly, noting that outcomes “can be supported by AI or other means.”
- While some employees welcome the reduced pressure to use AI indiscriminately, many are simultaneously testing a powerful agentic AI called Hatch, driving token consumption higher than before.
- Privacy concerns linger, especially after a paused keystroke‑tracking project that eroded trust in Meta’s AI initiatives.
Meta Ends Token‑Based AI Incentives
In an internal announcement this week, Meta told workers that their performance evaluations would no longer depend on how much they used AI tools. “We will not use AI adoption dashboards or token counts to evaluate impact,” the company stated, according to three employees who spoke with WIRED. The move formally ends what had become known internally as a “tokenmaxxing” program—a system that rewarded employees for racking up high volumes of AI‑generated tokens, essentially measuring how often they prompted chatbots or agents. The announcement marks a clear shift away from tying career advancement to raw AI usage, a practice that had drawn criticism for encouraging superficial or frivolous interactions with the technology simply to boost metrics.
Origins of the AI‑Driven Impact Metric and Employee Lawsuit
Almost a year ago, Meta introduced the idea of grading workers on their “AI‑driven impact,” which in practice meant evaluating the extent to which they incorporated chatbots and AI agents into daily tasks. Employees who logged heavy usage received labels such as “AI Native,” “AI First,” or “AI Enabled.” Those designations later became central to a lawsuit filed in July by roughly two dozen employees who claimed Meta violated U.S. antidiscrimination laws when laying off workers in May. The plaintiffs argued that employees on health or family leave could not accumulate token usage and were therefore unfairly penalized. Meta has denied the allegations, but the lawsuit kept the issue in the spotlight and contributed to the pressure that led to this week’s policy reversal.
Revised Performance‑Review Guidance
The new guidance unveiled this week replaces explicit references to evaluating employees based on “usage of AI” and their “AI Native” designation with looser wording that acknowledges alternative paths to success. “These outcomes can be supported by AI or other means,” the updated policy reads, emphasizing that managers should focus on the substance of an employee’s contributions rather than the specific tools they employ. A Meta spokesperson, Tracy Clayton, told WIRED that the updates aim to reiterate what has always been the case: Meta evaluates employees based on their overall impact, not on how much they interact with AI. Engineers across the company were reminded that the firm will no longer rely on AI adoption dashboards or token counts when judging performance.
Employee Reaction to the Policy Shift
Some Meta workers describe the changes as subtle but welcome, saying they free them from feeling compelled to use AI in situations where it does not make sense. One engineer noted, “I no longer feel like I have to chase a token legend badge just to look good on my review.” Others, however, remain skeptical, worrying that the shift might be superficial if underlying incentives—such as promotion criteria tied to project outcomes that still heavily favor AI‑heavy solutions—remain unchanged. The sentiment is mixed: while the removal of overt token metrics alleviates immediate pressure, employees are watchful to see whether the broader culture truly values impact over tool usage.
Rise of the Agentic AI Tool Hatch and Surging Token Consumption
Concurrently with the policy rollout, Meta has been encouraging—though not requiring—employees to test its latest agentic AI experiment, known as Hatch. Described as similar to the viral OpenClaw prototype, Hatch can autonomously browse the web, operate other applications, and carry out multi‑step tasks on a computer. Employees have been able to run Hatch on their corporate devices for several weeks ahead of an anticipated public release. Despite the new emphasis on impact over usage, many workers report that their token consumption has continued to surge as they experiment with Hatch’s capabilities, prompting some to joke that the “tokenmaxxing” era has simply moved underground. One employee told WIRED, “We’re still burning through tokens like crazy, just now we’re doing it to see what Hatch can actually do, not to chase a leaderboard.”
Privacy Concerns and the Legacy of Keystroke Tracking
The enthusiasm for Hatch is tempered by lingering privacy worries. Several employees have hesitated to connect the agentic tool to their personal email, calendar, or other non‑work digital accounts, fearing that the AI might inadvertently alter or expose sensitive information. Two current Meta staff members familiar with the concerns explained that prior initiatives—such as a project that tracked employees’ keystrokes and other activity on work devices to harvest data for AI training—had already eroded trust in the company’s AI systems. That project has since been paused, but its memory fuels apprehension about granting any AI broad access to personal data. As one worker put it, “After the keystroke‑tracking episode, I’m not comfortable letting an AI roam my inbox, even if it’s supposed to be helpful.” The combination of renewed AI experimentation and historic privacy breaches creates a complex environment where employees balance curiosity about cutting‑edge tools with a demand for stronger safeguards.
https://www.wired.com/story/meta-pushes-its-new-ai-agent-on-employees-but-eases-off-on-tokenmaxxing/