Technology news and discussion on TalkFuse.
Meta Pushes Its New AI Agent on Employees—but Eases Off on Tokenmaxxing
Meta has quietly softened its internal push for staff to maximise use of AI tools, updating performance-review guidance this week to drop explicit references to "usage of AI" and labels like "AI Native" in favour of judging employees on overall impact, "supported by AI or other means". The shift follows months of controversy, including a lawsuit from former employees alleging that AI-usage-linked labels contributed to unfair layoffs, and an internal "leaderboard" that ranked staff by AI token consumption, which was taken down after leaking in April. Meta says the changes simply restate its existing approach and denies the legal claims, but employees told WIRED the revised wording removes pressure to use AI when it isn't useful.
Alongside this, Meta is encouraging—though not requiring—staff to trial its new agentic AI tool, Hatch, which can autonomously browse the web and operate other applications, comparable to the "OpenClaw" tool. Some employees remain wary of linking Hatch to personal accounts such as email and calendars, citing privacy concerns and past unease over a now-paused Meta project that tracked keystrokes to train AI systems. Even so, some staff say they have found Hatch useful for personal tasks like booking appointments, while others flag concerns about the environmental and computing costs of heavy token use, which reportedly remains high despite the end of formal "tokenmaxxing" pressure.
In short
- Meta drops "AI usage" as a formal performance-review metric.
- Change follows a lawsuit and a leaked internal AI-usage "leaderboard".
- Meta now promotes its Hatch AI agent, but usage stays optional.
Steelman: both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
Encouraging staff to try Hatch and assess work by its overall impact could help employees discover useful ways to automate routine tasks without making raw AI use a goal in itself. Supporters may argue that giving staff room to experiment helps Meta understand where capable AI agents improve productivity, while allowing people to avoid tools that do not suit a task.
The case against
Employees may reasonably question whether an agent that can access email, calendars and other applications is worth the privacy and security risks, especially given earlier concerns about workplace data collection. They may also argue that even without formal usage targets, sustained promotion of AI can create pressure to adopt it, consume costly computing resources and divert attention from tools that offer clearer benefits.
Explained simply
Meta is a technology company that makes Facebook, Instagram and WhatsApp, and it is also developing artificial intelligence tools. One way it has encouraged staff to use AI was by tracking how much they used these tools, including through a ranking based on “tokens”, the units AI systems process. That approach drew criticism after former employees alleged AI-related labels were involved in unfair layoffs; Meta denies the claims.
Hatch is a newer kind of AI tool, called an agent, that can carry out tasks by browsing the web and using other applications. Meta is encouraging staff to try it, but some are cautious about giving it access to personal accounts and information. The debate matters because companies are deciding how to measure AI use at work while weighing its usefulness, privacy risks and computing costs.
AI Art Business Companies Culture Research Science Technology