Meta workers inflate metrics with AI tools, showing a person at a computer with data visualizations.

Editorial illustration for Meta Workers Overused AI Tools to Inflate Internal Metrics

Meta Workers Overused AI Tools to Inflate Internal Metrics

By AI Daily Post Edited by Brian Petersen, Editor-in-Chief

September 3, 2026 • 3 min read

Meta told employees this week that their performance reviews will no longer hinge on how often they use the company's AI tools, according to three workers who received the internal message and spoke with WIRED. The shift ends a policy that had pushed staff to rack up chatbot and agent usage as a proxy for productivity, a practice some inside the company came to call tokenmaxxing.

The original program dates back nearly a year, when Meta told employees they'd be evaluated on "AI-driven impact," Business Insider reported at the time. In practice, that meant usage numbers translated into labels like "AI Native," "AI First," or "AI Enabled," designations that later became central to a lawsuit filed in July by roughly two dozen employees who say Meta used those metrics to unfairly target workers on health and family leave during May layoffs. Meta has denied the allegations.

Now, as the company scraps that language in favor of broader "impact" criteria, it's simultaneously rolling out a new agentic tool called Hatch. Employees testing it say their token consumption is climbing anyway, even without the grading pressure attached.

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.

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

Why this matters

Meta says it's dropping the token quotas, but the underlying pressure hasn't gone anywhere. Employees are still watching their token counts climb as they test Hatch, the company's new agentic tool. That tells us the metric didn't create the behavior so much as give it a name.

For anyone building AI products or watching adoption numbers from the outside, this is a useful warning: usage stats generated under any kind of institutional pressure, explicit or implied, are close to worthless as signals of genuine utility. If a company as sophisticated as Meta couldn't stop employees from gaming token counts even after making the incentive official, then vendor-reported "engagement" figures across the industry deserve the same skepticism. Founders pitching investors on AI adoption metrics, and researchers citing internal usage data as evidence of product-market fit, should ask what was actually being measured and who had reason to inflate it.

The real story isn't that Meta had a quota. It's that removing the quota didn't change the behavior it supposedly caused.