Meta Stops Grading Workers on AI Use — Now Tests Its New AI Assistant Hatch

Meta is changing how it judges its employees. For almost a year, workers were told their performance reviews would partly depend on how much they used AI tools. Inside Meta, that led to something employees jokingly call "tokenmaxxing" — tokens are the little units of computing that AI tools consume. People felt pressured to use chatbots and agents even when they didn't need to, just to look good. Some were even labeled "AI Native" or "AI First." A group of former employees sued Meta this summer, saying workers on leave couldn't pile up AI usage and were unfairly laid off.

This week, Meta told workers that performance reviews will no longer be tied to AI usage. The new guidance says employees will be graded on their actual contributions, not on how they achieve them. A Meta spokesperson said the company has always evaluated people based on impact. The internal employee labels like "AI Native" were never officially used for reviews. For many employees, this comes as a relief. They no longer have to use AI in situations where it doesn't make sense just to protect their performance rating.

At the same time, Meta is quietly testing a new AI tool called Hatch. Hatch is what's known as an "agent" — an AI that doesn't just chat but can actually take actions on a computer, like searching the web, clicking on apps, or booking appointments. Some employees are testing it on their work devices before it is likely released to the public. A few say they've used it to schedule personal appointments and organize their lives. But others are cautious about connecting Hatch to personal email and calendars because of privacy worries, and because the AI might accidentally do something important wrong.

What does this mean for the rest of us? Meta's reversal is a real-world example of a problem every company will face: how to get people to use AI without turning it into a box-checking exercise. When employees feel forced to use AI, trust drops. When they're allowed to use it because it's genuinely helpful, they actually want to. Meta's shift away from counting AI usage and toward measuring real work could help rebuild that trust — and it offers a lesson for other employers watching closely.