Meta Ends Tokenmaxxing Incentives While Rolling Out Hatch AI Agent for Employee Testing

Meta Pushes Its New AI Agent on Employees -- but Eases Off on Tokenmaxxing

Meta is formally ending what amounted to a tokenmaxxing incentive program for employees. In an internal announcement this week, the social media giant told workers that their performance evaluations would no longer be dependent upon how much they used AI tools, three employees who received the message tell WIRED. But as Meta workers simultaneously begin testing a new agentic AI tool known as Hatch, they say their token consumption continues to surge. Almost a year ago, Meta had said workers would be graded on their "AI-driven impact," which in practice, meant evaluating the extent to which they used chatbots and agents to improve their day-to-day work, Business Insider reported at the time. Usage correlated with employees receiving labels such as "AI Native," "AI First," or "AI Enabled," according to a lawsuit filed in July by about two dozen employees arguing that Meta violated US anti-discrimination laws when laying them off in May. (Workers on health and family leaves say they couldn't accumulate usage and were unfairly penalized as a result; Meta has denied the allegations in the ongoing case.) New guidance for performance reviews unveiled this week replaces references to evaluating employees based on criteria like "usage of AI" and their "AI Native" designation, with looser wording, which caveats that "these outcomes can be supported by AI or other means." It restores Meta's focus to grading employees on their impact, with less regard to how it is achieved. Some Meta employees tell WIRED the changes are subtle but welcome, freeing them from feeling pressured to use AI in situations in which it doesn't make sense. Meta spokesperson Tracy Clayton tells WIRED that the updates this week aim to emphasize what has always been the case -- Meta evaluates employees based on their contributions. He adds that labels such as "AI Native" were never used for performance evaluation. Engineers across the company were told this week that the company "will not use AI adoption dashboards or token counts to evaluate impact," the Information reported on Wednesday. Got a Tip?Are you a current or former Meta employee who wants to talk about what's happening? We'd like to hear from you. Using a nonwork phone or computer, contact the reporters securely on Signal at Peard33.24 & Mzeff.88 For months, some Meta employees say their colleagues prompted AI tools repeatedly -- or what they viewed as friviously -- to maximize internal usage measurements, like the amount of tokens, or AI units, they consumed. The tokenmaxxing era reached a high point when one Meta worker created an internal leader board for tracking how much employees used AI, labeling them as, for example, a "Token Legend." After details of the dashboard leaked, the employee took it down in April. A couple of months later, Meta moved to ration employees' AI usage. More recently, Meta has been encouraging -- but not requiring -- employees to use its most advanced AI experiment, an agentic tool that can autonomously carry out actions on a computer. Known as Hatch and similar to the viral sensation OpenClaw, it can browse the web and operate other applications. Employees have been able to test Hatch for the past several weeks on their corporate devices ahead of an expected public release. Some employees have been hesitant to connect Hatch to their personal email, calendar, and other non-work digital accounts because of privacy worries, as well the possibility that the AI could mess up something important, two current Meta employees familiar with the concerns tell WIRED. Meta previously ran a project that tracked employees' keystrokes and other activity on their work devices to collect data for training AI systems. The project, which Meta has since paused, eroded many employees' trust in the company's AI tools, the sources say. However, no longer having to fear being dinged on performance reviews for sitting out AI testing could help boost morale and, ultimately, rebuild trust among Meta's workforce. Already, some employees say they have embraced Hatch for booking personal appointments and organizing their life outside of work. They describe the tool as effective, though some worry about the environmental and economic tradeoffs of using large amounts of tokens and other computing resources to automate tasks. Meta declined to comment on Hatch testing. Though employees may no longer be intending to maximize their token usage, Hatch is still consuming more than standard chatbots and software coding assistants of yesteryear, workers say. And employees believe management still is interested in seeing them prove they can effectively wield AI tools. If Hatch ends up boosting productivity as intended, some employees fear Meta could carry out additional job cuts. Last week, Reuters reported that Meta was gearing up for another round of major layoffs on top of the 8,000 jobs cut in May, but it called off the plans amid an increase in software bugs linked to reliance on AI and slower-than-anticipated progress on agent development. Meta CEO Mark Zuckerberg has said further mass layoffs are not expected this year. Additional reporting by Zoë Schiffer.

Meta Is Reportedly Sending Mixed Messages to Employees on Tokenmaxxing

Imagine you work a weird job where you fire guns at targets all day. At one point you were told your value to the company would be tied to how many bullets you fired, so you pulled the trigger like crazy. What if, when that caused a bullet supply issue, you eventually received a memo saying you were no longer expected to fire as many bullets as possible. But what if, around that same time, you were given the minigun from Terminator 2: Judgement Day? According to Wired, something a little like that is happening with AI at Meta. A March Kevin Roose column in the New York Times clued most of the world into the idea that certain AI-happy workplaces, including Meta, and also OpenAI, were attempting to maximize AI usage among staff -- thus the term "tokenmaxxing." Roose alluded to the popular (at the time) token-chugging agentic AI platform OpenClaw as a potential driver of the trend. Roose said Meta was "rewarding workers who make heavy use of A.I. tools and chastening those who don't." The report also mentioned leaderboards, which turned AI profligacy into a competition. A few month later, in June, a report from the Financial Times said Google was throttling Meta's AI tokens, forcing the company to rethink tokkenmaxxing to some extent. Now tokenmaxxing, it seems, is officially on its way out. The Information reported Wednesday that Meta engineers had received word that "AI adoption dashboards" and "token counts" would no longer be part of evaluations. Wired, informed by anonymous Meta insiders, says newly published performance review guidance for Meta employees no longer makes "usage of AI" a criterion for assessment. When describing desired outcomes, the new guidance apparently says they "can be supported by AI or other means." Some of Wired's sources in Meta say they're relieved to not feel under pressure to make constant use of AI. But as all this has been happening Meta has also been trying to make a competitive AI tool of its own. Back in 2024, Mark Zuckerberg said, "This is going to be the year when a highly intelligent and personalized AI assistant reaches more than 1 billion people, and I expect Meta AI to be that leading AI assistant." It's not clear that that ever happened with Meta's ChatGPT-style app. Personally I can say that I used the Meta AI web app exactly once, for a story. Amanda Siberling of TechCrunch apparently also used Meta AI for a story, although she had a worse experience because Instagram notified her friends that she was using it, and they made fun of her. It seems Meta may be hoping its luck will be better with a productivity-focused tool, rather than one oriented toward the same market as ChatGPT. In May, the Information reported that Meta was developing something called Hatch, which would be Meta's own OpenClaw-like agentic AI platform -- a consumer-friendly version of the exact type of token-guzzling agentic platform that had helped give rise to the tokenmaxxing trend in the first place. According to Wired's story, alongside the apparent sunsetting of tokenmaxxing incentives, Meta "has been encouraging -- but not requiring -- employees to use its most advanced AI experiment." That experiment is Hatch, which still hasn't been released, but employees have had access "for the past several weeks," Wired claims. The tool apparently consumes more AI tokens than standard chatbots and AI coding tools. Some apparently like using Hatch, but some of Wired's sources point to a news event from April, in which it emerged that Meta planned to use employee keystrokes to train AI, as a reason to be wary of the new tool. (Meta later paused that program). And it's simply a fact that heavy use if AI agents is still part of Meta's workplace culture. For instance, a memo earlier this week from AI chief Alexandr Wang explained the company's switch from Google Chat to Slack was to further enable the use of AI agents. Slack, he wrote is "the strongest platform available today for agents." Even though "not all of us are building agents today," Wang wrote, "everyone at the company will benefit from a robust and useful agent ecosystem." If Hatch is a success internally, according to Wired, the fear is that layoffs will soon follow. According to a Reuters report from last week, Meta planned to cut up to 60% of its workforce in certain areas. According to that report, the plan never moved forward due to bugs in the AI software that would have taken over for dismissed employees. Gizmodo reached out to Meta for information about the Reuters report, and about internal testing of Hatch. We did not receive a reply.