Feature · The AI Rollout Backlash

90% of executives say "no productivity gains yet": why companies that rushed into AI are hitting the brakes

"BREAKING: Employers worldwide are pulling back on AI." The post from unusual_whales on X on Oct. 29 passed 24,700 likes and 1.4 million views. Its source is a Guardian analysis published on Oct. 13. Here are eight things the piece and the surrounding conversation show, from Meta, Microsoft and Walmart to a Texas courtroom. Figures from other outlets are as reported by the Guardian.

1. What went viral: a 16-day-old article posted as "breaking"

   On Oct. 13 the Guardian published "Companies rushed in on AI. Now some are pulling back." It says employers are cleaning up a flood of errors and low-quality work that sap time and raise privacy, security and cost problems. When unusual_whales posted it as "BREAKING" on Oct. 29, it drew 24,700 likes. Top replies: "Rehiring the people the chatbot replaced" (439 likes) and "Mandatory AI training in Q1, mandatory AI cleanup in Q4" (79).

2. Meta: backing away from replacing staff with agents

   Citing Reuters, the Guardian says Meta scaled back plans to cut staff and lean on AI agents after employees spent more time fixing technical and security errors. A newsletter summary of the Reuters reporting (AI Weekly) says code changes rose 220% year over year but changes reaching users rose only 36%, while major incidents rose 40%. The Information reported Meta told staff AI usage would no longer count in performance reviews.

3. 라이언 로슬란스키 · 캐슬린 호건 · The Guardian ↗

   Microsoft: an executive laments "AI slop"

   Microsoft's Ryan Roslansky said he summarized "what was clearly an AI-generated doc" with his own AI and came away with no new thinking, and warned of a "doom loop" with an "ocean of sameness." Chief strategy officer Kathleen Hogan wrote that "access and usage do not equal transformation." The Guardian, citing Reuters, says firms from Microsoft to McKinsey limit employee use, and Palantir, Booz Allen and Nvidia restrict it over data-security fears.

4. The numbers: 90% see no productivity gain yet; 36% had an AI incident

   The Guardian cites an Atlanta Fed working paper in which 90% of executives say AI efforts have yet to translate into productivity gains. An EY survey found 36% of senior AI executives at companies with $1bn+ in revenue reported an AI incident causing data loss, financial or brand damage, or disruption. University of Pittsburgh researchers concluded productivity did not rise after firms invested in AI and cut staff. Challenger, Gray & Christmas counts more than 112,000 US job-cut announcements this year that mention AI.

5. Walmart: an AI that assumes workers are "perfect every single time"

   Citing Business Insider, the Guardian says Walmart store and warehouse staff face unreasonable expectations, errors and false alerts from AI tools that assign tasks and schedules. Workers described nonsensical routes, useless safety alerts and unrealistic schedules. One manager said the bots assume workers will be perfect and said "we're not going to hold people to these impossible standards."

6. In court: "What am I looking at?" and a $61.5 million verdict

   Before a 3M trial over a 2020 gas explosion in Houston, plaintiffs' lawyer Will Moye noticed odd formatting and metadata showing 14,500 pages assembled in hours in the company's expert report, billed at $90,000. Chat logs he obtained showed the expert asking ChatGPT "What am I looking at?" about a photo of the gas detector. The jury found 3M failed to inspect the system and awarded $61.5 million to 24 homeowners. The expert's lawyer said ChatGPT did not replace his judgment and helped with formatting and drafting.

7. 조니 테일러 주니어 SHRM CEO · The Guardian ↗

   The prescription: "adding agents to a broken process leaves a broken process"

   Microsoft's own write-up says adding agents to a broken process still leaves it broken, and speeding up one step just lengthens the queue at the next. Pairing early-career engineers with seasoned colleagues raised productivity. An Indeed survey from July found over half of employees say they aren't getting the AI training they need, and just 45% of managers feel equipped to lead "AI natives." SHRM's Johnny C. Taylor Jr. said companies are "driving the truck and building it at the same time."

8. 맷 포콕 · X ↗

   Counter-evidence: a story of AI turning a company cash-flow positive

   A thread from developer educator Matt Pocock drew 1,350 likes the same day. Relaying a Pragmatic Engineer podcast, he says an unnamed CTO weighing layoffs instead had an agent rebuild a licensed component; the result was faster, the licensor asked them not to open-source it and paid, and the company turned cash-flow positive. Replies noted the CTO, company and component are all unnamed. The story is unverified.