Meta drops AI token-count performance metric as workers test new Hatch agent

Meta has formally ended the practice of tying performance evaluations to AI token consumption, removing references to “AI Native” designations in its latest review guidance. The change comes after a lawsuit claimed the previous system discriminated against employees on leave who could not accrue usage. Engineers were also told the company will no longer rely on AI adoption dashboards for impact assessments.

Meanwhile, staff are testing Hatch, an agentic AI capable of performing tasks across devices, but hesitations remain over linking personal accounts and potential errors. Earlier internal efforts to track token usage, including a leaderboard dubbed “Token Legend,” have been discontinued. The company says the new policy refocuses assessment on overall contributions rather than the tools used. Meta declined further comment on the Hatch trial.