After two years of rapid experimentation, large corporations are pulling back from artificial intelligence as reliability issues and cost overruns mount.
Fresh Census Bureau data shows that AI adoption among firms with more than 250 employees fell from 14% in June to 12% by late August 2025 — the first sustained decline since tracking began. Smaller businesses continue to adopt AI slowly, but the enterprise retreat underscores growing skepticism around AI’s near-term utility.
The reversal follows an MIT study revealing that 95% of enterprise generative AI pilots failed to show measurable ROI, and a tense summer that saw public disagreements among AI leaders.
At the same time, the pullback is fueling a surge in demand for human expertise:
- Fact-checking became one of the top 10 AI-related skills on Upwork in August.
- Content writing assignments grew 15%.
- Language tutoring skyrocketed 162%.
“Even the best AI models still hallucinate 10% to 12% of the time,” said Kelly Monahan, managing director of the Upwork Research Institute. “We just cannot necessarily overcome that statistical problem yet.”
As companies confront the hidden costs of over-automation, specialists are commanding premium rates. One BBC report detailed how a marketing manager charged $2,000 to rewrite flawed AI copy, highlighting how attempts to cut costs with automation often backfire.
The message for enterprise leaders is clear: AI may remain part of the corporate toolkit, but human oversight has proven more critical — and more valuable — than many had anticipated.
