Updated
Updated · WIRED · Sep 3
Meta Ends AI Usage Reviews as Hatch Testing Drives Higher Token Consumption
Updated
Updated · WIRED · Sep 3

Meta Ends AI Usage Reviews as Hatch Testing Drives Higher Token Consumption

1 articles · Updated · WIRED · Sep 3

Summary

  • Meta told employees this week that performance reviews will no longer depend on AI adoption dashboards, token counts or labels such as “AI Native,” reversing a year-old push to measure “AI-driven impact.”
  • The change shifts evaluations back to employee impact regardless of method, easing pressure some workers said led colleagues to spam chatbots and chase internal token leaderboards before Meta later rationed AI use.
  • Hatch — a new agentic AI tool now being tested internally ahead of an expected public release — is still pushing token consumption higher than older chatbots, even without the formal review incentive.
  • Some employees welcome the policy reset but remain wary of Hatch because of privacy risks, trust issues after Meta’s paused device-tracking project, and fears that successful automation could revive layoff pressure after 8,000 May cuts.

Insights

Why did Meta's obsession with AI tokens backfire, and could their powerful new Hatch agent secretly pave the way for massive job cuts?
How does the shift from token-chasing to the million-token Muse Spark 1.1 reveal a hidden crisis in measuring true AI business value?