Anyone who has opened a document and suspected a chatbot wrote it has company at the top of Microsoft. A senior executive there recently complained publicly about exactly that, and another senior leader wrote about what the company’s own AI rollout taught it. Microsoft declined to comment for the Guardian’s reporting, so the clues come from its own people, with Meta offering the clearest example of a pullback.
The Doc That Said Nothing
Ryan Roslansky, a senior Microsoft executive, described on LinkedIn getting what was clearly an AI-generated document from a co-worker. He read it, had his own AI summarize it, and came away with no new thinking. He called the pattern a “doom loop,” where AI does the reading and writing and leaves behind an “ocean of sameness.” He added that AI slop shows up in everyday work as well as on social networks.
What Microsoft Says It Learned
Microsoft rolled out AI widely internally, and Kathleen Hogan, its chief strategy and transformation officer, shared what the company learned in a blog post. The company first treated AI like a traditional technology rollout. She wrote that “access and usage do not equal transformation,” and that adding agents to a broken process still leaves a broken process. Focusing on training, especially pairing early-career engineers with seasoned employees, boosted productivity.
Not Just Microsoft
Microsoft and McKinsey are among the firms limiting how much employees use AI, the Guardian reported. The flood of new tools and heavy use is weighing on bottom lines. Palantir, Booz Allen Hamilton and Nvidia are restricting use over data security worries. Johnny C Taylor Jr., president and CEO of the Society for Human Resource Management, said leaders feel they are “driving the truck and building it at the same time.”
Meta’s Plan to Shrink Teams
Mark Zuckerberg and his top lieutenants left a January retreat with Project OT, short for Organization Transformation. It pictured an “AI native” company where virtual workers handled much of the daily work, overseen by smaller human teams. Internal documents reviewed by Reuters described two waves, in May and November. Meta acknowledged its most drastic scenarios involved cutting some teams by up to 60%, but said it never planned to lay off 60% of its entire workforce.
The Night Before the Layoffs
On the night of May 19, hours before the first wave, Zuckerberg called off planning for the November cuts. Meta laid off 10% of its employees the next day. Zuckerberg then told remaining staff in a memo he did “not expect other company-wide layoffs this year.” Reuters could not determine what prompted the change, and Meta said the second wave was canceled before it had decided how many people would lose their jobs.
Staff Weren’t Buying It
Employees were convinced the AI push was partly aimed at replacing them, and morale showed it. Meta’s half-year Pulse survey put favorable sentiment at 55%, down from 74%. Many staffers flooded the company’s Workplace network with angry posts and gallows humor, replying to executives with pictures of elephants, their shorthand for the layoffs. Meta later paused a program that tracked U.S. employees’ keystrokes and mouse clicks to train its AI agents.
More Code, Smaller Gains
Meta’s own internal data raised doubts about the agent technology the plan leaned on. In an early June post, Chief Technology Officer Andrew Bosworth wrote that code changes to Meta’s internal platforms were up 220% year over year. Changes that led to new or upgraded features reaching users were up only 36%. Major technical and security incidents rose 40% from the previous year, and firefighting time rose 70%, according to internal posts.
No Productivity Bump After the Cuts
A University of Pittsburgh study published in May found that companies saw no rise in productivity after investing in AI and cutting staff. Professor Mark Ma, who worked on the study, said company share prices tend to dip after AI-related layoffs, which suggests the stock market doesn’t expect them to work out. As of mid-October, Challenger, Gray & Christmas had counted more than 112,000 U.S. job cut announcements in 2026 that mentioned AI.
Betting on People, For Now
A Meta video ad said the company was “betting on people,” and Zuckerberg told employees in June that Meta wasn’t primarily focused on automating work. He kept to the words “company-wide” and “this year” on layoffs, though. In July, he told staff that agent technology hadn’t accelerated as quickly as expected. Microsoft, which sells AI to other companies, and Meta both seem to be sorting out how much of it to use internally.
