Meta stops judging employees on how much they use AI tools

In short: Meta says employee performance reviews will no longer be tied to how much they use the company’s AI tools, even as workers test a new AI assistant called Hatch.

What happened

Meta told employees in an internal message this week that performance evaluations will not depend on “usage of AI,” according to WIRED, which spoke with employees who received the note. Meta is replacing earlier language that pushed workers to show “AI-driven impact” with wording that says results can be achieved with AI “or other means.”

This change follows months of complaints from some employees about “tokenmaxxing.” Tokens are small pieces of text that AI systems process, and more tokens usually means higher computing costs (like a car’s fuel meter, but for AI). Some workers said colleagues repeatedly prompted internal chat tools to drive up token counts, including through an internal leaderboard that ranked usage.

At the same time, Meta is encouraging employees to try Hatch, a more hands-on kind of AI tool. WIRED describes Hatch as an “agentic” tool, meaning it can take actions on a computer for you (like an assistant who can click around, open apps, and fill things out). Employees have been testing it on work devices ahead of an expected public release, but some are hesitant to connect it to personal accounts like email and calendars due to privacy and mistake concerns.

Why it matters

This shows a shift in how Meta wants people to use AI at work. The company is stepping back from measuring effort by raw AI usage, but it still wants employees to practice using more powerful tools that may change how office work gets done. For regular people, it is a reminder that these tools can be useful, but they also raise practical questions about privacy, errors, and cost.

Source: Wired