Meta has updated its performance-review guidance to focus on the actual impact of an employee’s work rather than counting AI usage or labeling employees as AI-native or AI-first.

The company is moving away from tokenmaxxing and relaxing AI usage metrics in performance evaluations, signaling a shift away from heavy AI-use incentives.

Previously, Meta tied AI-driven impact to evaluations and promoted broad AI adoption, which created observable metrics and labels that some staff viewed as targets for using AI.

Staff concerns have grown over Hatch testing, which links work tools to personal accounts and raises fears about data collection and disruption of important tasks.

Despite changes to evaluation metrics, employees continue to report high usage of Hatch, Meta’s advanced AI agent capable of autonomous actions, web browsing, and app control.

Meta has been developing Hatch, an internal OpenClaw-like agent aimed at boosting productivity, with some employees granted early access before a public release.

Hatch reportedly consumes more AI tokens than standard tools, sparking mixed reactions and concerns about training AI on keystrokes, which led to a pause in that initiative.

There are broader workforce implications, including potential layoffs if AI tools scale, though Reuters notes such plans faced implementation challenges due to software issues.

Meta has not provided further comment on Hatch testing; Reuters previously reported that mass layoffs would be scaled back and that no large layoffs are expected this year.

Coverage has noted pushback or constraints around AI initiatives, with reports of Google throttling Meta’s AI tokens and Meta pausing certain data-collection practices.

The story draws on reporting from multiple outlets (Wired, NYT, Financial Times, The Information, Gizmodo, Reuters, Business Insider) to portray Meta’s AI strategy and reception."},{