Artificial intelligence may save workers time, but new research suggests those productivity gains could carry an uncomfortable cost. As companies hand more work to AI, employees are increasingly worried that people could become worse at thinking, writing, and making decisions for themselves.
WSJ Intelligence surveyed 2,507 business professionals across the United States, United Kingdom, Italy, South Africa, and Brazil for its Cognition Index Research Report, sponsored by Philip Morris International. Respondents ranged from entry-level associates to executives and board members.
The Philip Morris connection is admittedly unusual for research about artificial intelligence and the workplace. PMI is primarily known for cigarettes and nicotine products, not workplace technology or AI research. The company sponsored the study, while WSJ Intelligence conducted it. WSJ Intelligence is also part of The Wall Street Journal’s commercial advertising organization, rather than its newsroom.
With that context established, some of the findings are truly fascinating.
The report describes a potential problem it calls “cognitive atrophy.” The idea is straightforward: if people repeatedly hand thinking-intensive tasks to AI, some of the abilities required to perform those tasks could weaken through lack of use.
Twenty-nine percent of respondents selected atrophy of human-centric skills as their single biggest concern about human capabilities in an AI-enhanced world. Practical skill loss caused by overreliance on AI followed at 18 percent, while the spread of AI-generated misinformation came in at 17 percent.
Younger and less-senior workers appear particularly worried. Seventy percent of entry-level employees expressed concern about a decline in core writing skills, compared with 61 percent of the most senior executives.
Writing isn’t the only ability workers fear losing. Critical thinking was identified by 46 percent of respondents as a human capability at risk of being undervalued or eroded as AI spreads through workplaces. Creative empathy came next at 44 percent, followed by creativity at 43 percent.
There is an interesting contradiction here. The human abilities that workers fear AI could weaken are also some of the abilities they expect companies to value more in an AI-heavy future.
The survey found virtually every measured human attribute was expected to become more important over the next three years. Creative empathy showed the largest projected increase, followed by adaptability, resilience, and personal accountability.
Workers are also using AI far more often than they trust it.
For research and information synthesis, 83 percent reported weekly AI use, yet only 57 percent expressed high trust in the output without thorough human review. Coding and technical support had a similar divide, with 73 percent reporting weekly use and 55 percent expressing high trust. For data automation, the numbers were 72 percent and 52 percent, respectively.
That skepticism might be healthy, especially considering how few organizations apparently have formal procedures for checking what AI produces. Only one in four respondents said their company has a clear process for verifying AI results.
There is even disagreement within companies about whether those safeguards exist. Thirty percent of C-suite respondents said their organization had a clear fact-checking protocol, compared with just 17 percent of mid-level employees and 19 percent of entry-level associates.
Senior executives also generally trust AI more than workers below them. For coding and technical support, for example, 59 percent of C-suite and board respondents expressed high or complete trust in AI output without thorough human review. Among entry-level associates, that fell to 45 percent.
Workers don’t appear to be rejecting AI, however. Instead, they seem reluctant to surrender final judgment to it.
When an AI recommendation conflicts with human intuition in creative decisions such as branding, campaign concepts, or product design, 62 percent said human intuition should take precedence. Only 12 percent would give the AI recommendation priority, while 25 percent favored equal weight.
Even for strategic and predictive decisions, 50 percent favored human intuition, compared with 20 percent who would prioritize AI and 30 percent who wanted equal weighting.
“Human cognition is a bit like a muscle: use it or lose it,” said Dr. Moira Gilchrist, chief global communications officer at Philip Morris International. “AI can help us work more efficiently, but we need to be careful not to outsource our thinking to technology. The real opportunity is to use AI to think better, not think less. That means continuing to nurture curiosity, critical thinking and active engagement, even as these tools become part of everyday work.”
There are reasons to be cautious about interpreting the findings too broadly. Beyond PMI sponsoring the research, the sample was heavily weighted toward senior leadership. The study was conducted between February 26 and April 13, 2026, and included an equal geographic split across its five countries.
Still, the underlying question is worth considering. AI can write an email, summarize a report, analyze data, generate code, and brainstorm ideas in seconds. Those capabilities can make people more productive, but productivity and ability aren’t necessarily the same thing.
Perhaps the biggest danger isn’t that AI will become better at thinking than humans. It may be that humans become so accustomed to letting AI think for them that some of us eventually become worse at doing it ourselves.
Support independent tech journalism
NERDS.xyz is independently owned and operated. If you enjoy my coverage of Linux, AI, hardware, cybersecurity, and tech culture, consider supporting the site on Ko-fi.
Support NERDS.xyz