After racing to roll out generative AI, a growing number of employers are tightening controls as quality problems, security risks and higher costs surface. Meta has slowed plans to replace staff functions with AI agents and is rebuilding management layers. At Microsoft’s LinkedIn, CEO Ryan Roslansky warned of “AI slop,” where automated writing and reading create an “ocean of sameness” with little new thinking. Walmart workers report scheduling glitches, false alerts and unrealistic tasking from store and warehouse bots.
The pullback reflects a broader reality: 90% of executives say AI has yet to deliver productivity gains, according to an Atlanta Fed working paper, while an EY survey found 36% of large firms suffered AI-related incidents, from data loss to operational disruption. Companies including Palantir, Booz Allen Hamilton and Nvidia have curbed internal use over data-security concerns. Microsoft says training and process redesign—not just tool access—are required to unlock value.
A courtroom episode underscored the risks: a 3M expert’s reliance on ChatGPT drew juror skepticism in a $61.5 million verdict. Research from the University of Pittsburgh ties AI-related layoffs to weak performance, with share prices often dipping after such announcements. The new corporate playbook emphasizes governance, role-specific training and measured deployment over breakneck adoption.
Related articles:
— NIST AI Risk Management Framework
— Will AI Fix Work? (Microsoft Work Trend Index)
