Guardian: companies that rushed into workplace AI are pulling back as staff clean up errors; 90% of executives see no productivity gain yet
Confirmed
The takeaway
It cites an Atlanta Fed working paper in which 90% of executives report no productivity gains so far, and a court case in which an expert’s ChatGPT-assembled report preceded a $61.5M verdict against 3M.
Status
- Claim
Confirmed
- Our reporting
- High confidence
- Importance
- 2 of 5
- Last verified
- 13 October 2026
Your AI and this story
- GPT-6 Astra166 days after its cutoff
- Claude Opus 5.5105 days after its cutoff
- Gemini 4 Argon196 days after its cutoff
- Grok 4.7135 days after its cutoff
None of these four assistants can know about it. The closest, Claude Opus 5.5, stops 105 days before it.
Key facts
- Meta ‘has scaled back plans to axe staff and use more AI agents’ after finding employees ‘spending more and more time cleaning up technical and security errors’ (Guardian, citing Reuters); it is expanding managerial ranks again, paused tracking employees’ computer activity to train AI (Reuters) and told staff AI-tool usage will no longer count in performance reviews (The Information)
- Usage limits: ‘Firms from Microsoft to McKinsey are limiting usage for employees’ as consumption weighs on costs; Palantir, Booz Allen Hamilton and Nvidia restrict employee use over data-security concerns with AI companies (Reuters)
- Atlanta Federal Reserve working paper: 90% of executives say AI efforts have yet to translate into productivity gains at their companies
- Ernst & Young survey (September 2026): 36% of senior AI executives at companies with at least $1B revenue reported an AI incident causing data loss, financial damage, brand damage or operational disruption
- Indeed survey (July 2026): more than half of employees say they ‘aren’t getting the AI training they need’; 45% of managers feel ‘equipped to effectively lead AI natives’
- Microsoft’s Kathleen Hogan (blog post): ‘We learned that access and usage do not equal transformation’; ‘adding agents to a broken process still leaves a broken process’. Executive vice-president Ryan Roslansky warned of a ‘doom loop’ in which AI writes and AI summarizes, leaving an ‘ocean of sameness’
- SHRM chief Johnny C. Taylor Jr.: ‘we’re driving the truck and building it at the same time, and that’s a problem’; ‘If we could all rewind ... we would have introduced [this] in a more thoughtful, slow way’
- Walmart store and warehouse staff report AI tools that set unrealistic schedules, plot nonsensical routes and issue useless safety alerts (Business Insider)
Show 2 more
- 3M case: attorney Will Moye found that 3M’s expert witness Josh Autenrieth (Knighthawk Engineering), who billed $90,000, had assembled a 14,500-page report in hours with ChatGPT, asking it ‘What am I looking at?’ about a photo of the gas detector at issue; the jury ordered 3M to pay $61.5M to 24 homeowners over a 2020 Houston plant explosion. 3M says he ‘used AI as a tool’ after reaching his own conclusions
- Layoffs: more than 112,000 US job-cut announcements in 2026 so far have mentioned AI (Challenger, Gray & Christmas). A University of Pittsburgh study (May 2026) found productivity did not rise after companies invested in AI and cut staff; co-author Mark Ma: share prices tend to dip after AI-related layoff announcements
What happened
The Guardian’s “Reworked” series published an analysis of companies slowing down on workplace AI. Its examples are mostly taken from other outlets’ reporting (Reuters, Business Insider, The Information) and from surveys: Meta reversing parts of its agent-for-staff plan, Microsoft and McKinsey capping usage, security-minded firms restricting it, and workers at Walmart managed by error-prone scheduling tools. Wharton’s Jessica Reif explains the mechanism: AI produces “plausible-looking output” whose errors are not caught at once and so travel through a workflow.
The NYT’s DealBook described the same pattern three days earlier as a “verification tax” (surveys by Workday, Google and MIT, per Techmeme’s summary); that article was not readable.
Why it matters
The piece is a counterweight to the week’s layoff and “AI first” announcements (Kroger, Epic): the same months show companies discovering that checking machine output costs human time, and that the promised productivity is not yet in their numbers.
Sources
2 sources from 2 sites. Numbers match the chips in the text.
2 sources: 2 press
Press
Changes
- Filed from the Guardian article, read in full