Meta’s move-fast-and-break-morale era

TL;DR: Meta’s CTO admitted in an internal memo this week that the company did an “atrocious job” rolling out its AI reorg, per a new Wired report—leading morale to tank. And the planned fixes essentially amount to more social events and better snacks. Meta is the year’s starkest example of a firm trying to transform into a leaner, AI-first workforce, and it’s showing us the biggest hurdle: keeping its remaining employees from revolting.

What happened: While conceding that it botched its explanation of the AI shift, the company has largely stood by the shakeup itself. To recap: In April, the firm announced it would lay off 8,000 people, and then soon after said it would reassign 7,000 more. That reshuffle fed a new division called Applied AI, where staff weren’t recruited so much as conscripted. And the work itself—generating coding puzzles and training data to sharpen Meta’s models—is reportedly so mind-numbing that one worker called Applied AI “literally the gulag.”

The morale-boosting fixes the company has proposed so far: capping managers at about 20 direct reports (down from 50), better “microkitchens” in the office, beefed-up event and travel budgets, and letting Applied AI engineers apply for other Meta roles. (The employee keystroke-and-click tracking, though, is mostly still on.)

“Tell him he's a piece of s--t”: That’s what one Meta employee reportedly demanded be relayed to an unnamed Meta AI executive during a livestreamed all-hands last week—an outburst Wired reports reflected mounting frustration inside Applied AI. It’s just one example of the explosive worker backlash to Meta’s AI pivot—negative posts about it have flooded the anonymous workplace forum Blind, and a team-building hackathon Mark Zuckerberg recently floated went over like a lead balloon.

What it’s all for: The work that Applied AI’s engineers (unhappily) do is meant to improve Meta’s consumer AI products—but the company’s track record in this space has been uneven this year. Muse Spark, Meta’s newest model, shipped in April, yet its developer API has been delayed repeatedly (which has stalled its AI revenue). And some features have arrived less than polished: In June, hackers easily exploited Meta’s AI support chatbot—which rolled out in March—to reset Instagram passwords and take over thousands of accounts.

Bottom line: Meta is now one of the biggest corporate test cases for reengineering a giant firm around AI—and it’s trying to weather the storm of employee discontent while the payoff of its AI bet remains unproven. —WK

Also at Meta…

• Facebook’s new “AI Mode” answers your questions by digging through every public post on the platform, including those from Groups and Reels.