Artificial intelligence tools are rapidly becoming integral to daily workflows in medicine, software engineering, and other knowledge-intensive professions. From aiding diagnoses in hospitals to assisting with code writing in tech companies, AI is no longer experimental but a regular workplace tool. However, several studies suggest a potential side effect: professionals' own skills may suffer if they become too dependent on AI.
Growing Concerns Among Healthcare Professionals
According to a recent article in Nature, physicians, nurses, and software engineers increasingly worry about the deterioration of their core skills due to extensive AI use. A survey of American healthcare workers revealed that 70% of nurses and 77% of physicians fear their skills are declining because of AI dependency.
AI in Medicine: Performance Drops Without AI
One early example comes from gastroenterology, where AI is used to detect cancerous lesions during colonoscopies. Research in Poland on experienced endoscopy specialists found a decline in diagnostic performance after they began using AI assistance. These specialists had each performed over 2,000 colonoscopies. Their adenoma detection rate fell from 28.4% before AI use to 22.4% after becoming accustomed to it. As reported in The Lancet Gastroenterology and Hepatology, AI technology may reduce doctors' diagnostic performance when it is absent.
Professor Robert Wachter of the University of California, San Francisco, told Nature that these results raise questions about clinicians' attentiveness when they know a system normally handles part of their work.
Reasons Behind Skill Drift
Scientists involved in the research say the issue is not a decline in knowledge but a change in behavior. With continuous AI presence, specialists may subconsciously exert less effort, concentrate less, or critically evaluate less. Yuichi Mori, a doctor-scientist at the University of Oslo, noted in a Nature feature that specialists may become "less motivated, less focused" and less responsible in cognitive decision-making without AI assistance. This problem is critical for healthcare systems in the UK and elsewhere, where AI diagnostics are being gradually introduced to relieve overworked staff.
Software Engineering: Early Signs from Randomized Trials
The issue extends beyond medicine. In a randomized controlled experiment by Anthropic, 52 software engineers were asked to complete a basic coding assignment. All had access to online information, but only half received additional help from an AI assistant. Researchers found that engineers using AI completed the task faster but showed weaker problem-solving performance when evaluated without AI. Although it is too early for conclusions, this may signal that AI boosts short-term productivity but reduces independent problem-solving in some contexts.
Broader Debate Across Science and Industry
While deskilling has been discussed in scientific and medical labs, the issue is now gaining prominence beyond those environments. Information scientist Kevin Crowston at Syracuse University argues that knowledge itself could be the first form of defense. Experts must determine which skills should remain entirely human and not fully computerized. As AI systems become essential components in workplaces rather than optional tools, deskilling becomes more topical in industries such as law, journalism, and banking.
Implications for UK Workers
UK workers, particularly in medicine and technology, stand to be heavily affected. The National Health Service is trialing AI programs for tasks from diagnosis to management, while British tech companies are quickly incorporating AI coding assistants. The danger lies not in sudden loss of ability but in gradual dependency. Tasks previously performed through skill may become "backup skills," used only when machines malfunction. This is especially troubling in medicine, where AI failure could make independent diagnosis more difficult for some professionals.
Augmentation or Replacement of Skills?
The accumulating evidence does not indicate that AI is inherently bad. In most cases, it increases efficiency, accuracy, and speed. But how people adapt matters. If professionals use AI as a substitute for thought, skill degradation is more likely. When used as a safety net or second opinion, AI can enhance skills. Researchers say the next stage requires creating systems that preserve human skills alongside machine knowledge. Currently, evidence is limited but consistent, raising an important question: as AI becomes more sophisticated, are people quietly losing their skills?
