As artificial intelligence becomes increasingly integrated into professional environments, there is a concern that essential skills may deteriorate.
This concern is particularly significant among medical professionals, computer scientists, and other specialists. A recent survey of healthcare workers in the United States revealed that 70% of nurses and 77% of physicians fear that their skills might decline due to excessive reliance on AI technologies.
These apprehensions appear to be well-founded. Evidence indicates that AI-induced ‘deskilling’ is already occurring in fields like medicine and computer science. Researchers are actively seeking ways to maintain human expertise in the era of AI.
On supporting science journalism
If you’re enjoying this article, consider supporting our award-winning journalism by subscribing. By purchasing a subscription you are helping to ensure the future of impactful stories about the discoveries and ideas shaping our world today.
“Awareness of this issue hopefully encourages self-reflection on which skills individuals want to retain and which they’re comfortable delegating” to AI tools, says Kevin Crowston, an information scientist at Syracuse University in New York.
Spoiled by AI?
A study involving Polish physicians specializing in endoscopy — a procedure using flexible probes to view the interior of the human body — highlights how AI tools can quickly diminish human capabilities. These physicians, each having conducted over 2,000 colonoscopies, were provided with an AI system that examines colonoscopy images in real time to identify precancerous intestinal lesions known as adenomas. The AI tool was accessible to them on certain days but not others.
Their performance noticeably declined on days the AI system was unavailable. Prior to the tool’s introduction, the physicians detected at least one adenoma in 28.4% of colonoscopies. After the tool’s introduction, this rate dropped to 22.4% for procedures conducted without AI assistance.
These findings, published in The Lancet Gastroenterology and Hepatology last October, suggest that even highly skilled professionals might experience a decline in task performance as they grow more reliant on AI tools. Robert Wachter, a physician at the University of California, San Francisco, notes that extended use of these tools can lead clinicians to become “less motivated, less focused, and less responsible when making cognitive decisions without AI assistance.”
Co-author Yuichi Mori, a physician-researcher at the University of Oslo, emphasizes the need for further studies to verify this phenomenon. He warns AI users about the potential risk of skill degradation, adding, “There is no established solution against deskilling right now. It should be a very hot research topic in the next decade.”
No lesson learnt
To explore skill loss in computer science, researchers at the AI firm Anthropic in San Francisco, California, conducted a randomized controlled trial with 52 software engineers tasked with a basic coding assignment. All participants had access to online instructions, but half were also encouraged to use an AI assistant.
Subsequently, participants completed a quiz on what they had learned. Those using the AI assistant scored significantly lower, with an average of 50% compared to 67% for those without AI assistance. The AI group struggled particularly with questions requiring code error diagnosis, indicating a lack of understanding of the code they generated. The study was made available on the preprint server arXiv before undergoing peer review.
These results are concerning, especially for students and emerging professionals in the field, says Crowston, who studies how generative AI affects software developers’ learning and skill retention. “There is now a strange disconnect between performance and learning,” he states. “Individuals can perform at a high level by essentially borrowing skills from the AI, but they are not developing those skills themselves.”
Outsourcing cognition
Technological advancements have historically rendered certain skills obsolete, observes Tapani Rinta-Kahila, an information-systems researcher at the Hanken School of Economics in Helsinki. For instance, GPS navigation systems have diminished people’s navigation abilities. However, generative AI tools represent the first technology to automate a range of cognitive functions related to thinking and interpretation, which were long considered uniquely human.
Rinta-Kahila’s research supports these concerns. In 2018, he published a study on accountants who had used an automated, non-AI accounting system for over a decade. When the tool was removed, the accountants forgot how to perform several routine tasks. He predicts that AI systems will continue to impact work by taking over tasks once done by early-career professionals. “Future generations of programmers may not grasp the fundamentals of coding well, lacking hands-on experience,” he asserts. “This is also true for many other knowledge-intensive professions, such as accounting and law.”
To combat AI-driven skill erosion, individuals must be conscious of how much they delegate to generative AI tools. Understanding how these models function and their limitations is crucial. It’s essential to scrutinize AI outputs rather than accept them blindly. “People need to balance the dynamics of relying on generative AI while remaining vigilantly aware,” he advises.
This article is reproduced with permission and was first published on June 18, 2026.
