AI May Be Boosting Productivity — But It's Quietly Deskilling Workers, a Professor Warns

Companies are racing to deploy AI tools they expect will dramatically raise productivity. But Anastasia Berg, an assistant professor of philosophy at the University of California, Irvine, warns that heavy reliance on AI may be eroding workers' core abilities rather than genuinely building them.

What the professor observed

Speaking on the Philosopher podcast, Berg said recent research and reports from colleagues in multiple industries point to a worrying pattern: employees who lean on AI tools too much are losing essential skills. She stressed that acquiring a skill is only the first step — maintaining and deepening it requires practice, feedback and the friction that AI often removes.

We have a tremendous amount of empirical data on this question of skill attrition or skill atrophy, Berg said, adding that skill maintenance is frequently overlooked when people focus only on how skills are initially learned.

Evidence and concerns

Berg did not point to a single definitive study in the interview, but there is a growing body of research published through outlets such as Oxford University Press and journals hosted by Springer and MDPI that supports a mixed picture: AI can speed learning and increase engagement, but it may also reduce depth, critical thinking, creativity and long-term skill development.

One large analysis of ChatGPT conversations by researchers at OpenAI, Duke University and Harvard University examined 1.58 million chats and found that by June 2025, about 73% of messages from adult users were non-work-related. While that study did not break down every non-work category, the high share of personal usage underscores the broader trend of cognitive offloading — relying on AI for routine decisions, emotional support and everyday advice.

Junior workers are most at risk

Berg singled out junior employees as particularly vulnerable. She and computer science educators report that some students and early-career developers lean on AI to write and debug code before they have built the foundational knowledge needed to understand, verify or correct what the tools produce. In her words, experienced coders can productively use AI, but newcomers who never learn core techniques become dependent and less capable independently.

That dependency, she argues, removes the productive challenges that develop reasoning, troubleshooting and decision-making. Over time, those capacities can atrophy when they are not exercised.

Broader implications

Berg warns the implications extend beyond the workplace: when adults routinely turn to chatbots for emotional support, daily planning and social interaction, independent judgment and everyday problem-solving may weaken. The result could be a workforce and a public that appear more efficient on paper but lack resilience without digital assistance.

Her takeaway: AI does more than automate tasks — it can automate the cognitive processes by which people learn. If organizations adopt AI across every workflow without deliberate strategies to preserve learning and verification, they risk producing employees who rely on tools but cannot perform without them.

Organizations that want the benefits of AI should pair tool adoption with training, mentorship and assessment practices that preserve deep skills and human oversight. That balance can help ensure AI augments rather than quietly dismantles the competencies people need to think independently and solve hard problems.