The contemporary refrain “AI is coming for your job” may sound to some like an antitechnocratic exaggeration. Then again, it may also hold some truth. AI may be making workers obsolete in a way more insidious than the “robots taking over the world” trope.
Of course, AI can boost productivity and improve work quality. Still, mounting research suggests that relying on it too heavily may erode workers’ job skills, critical thinking and ability to solve problems independently.
The concern goes beyond AI reliance merely hindering the development of new knowledge and skills; it gradually weakens abilities workers already have. Routine tasks may be completed faster and with better results, but if employees increasingly hand over the thinking behind those tasks to AI, experts warn they may become less capable of working through unfamiliar problems on their own.
Related Article
A Seoul office worker surnamed Kim said they use generative AI to grade their work, revising reports until they receive a score of at least 95. “When a report clears 95 on the first try, I’m sometimes as happy as if my boss had praised me, but I also suddenly worry about whether or not I could write a good report without AI,” Kim said.
A worker at another midsize company surnamed Choi said using AI has become difficult to avoid. “AI points out things I hadn’t thought of, so I can’t not use it,” Choi said. “But as I use it out of habit, I also sometimes feel that my ability to find problems and handle work on my own is shrinking.”
When generative AI first entered the workplace, much of the attention centered on its potential to sharply increase efficiency and productivity. A growing body of research, however, suggests those benefits depend heavily on the task and on how workers use the technology.
In 2023, Harvard Business School researchers conducted an experiment involving 758 consultants at Boston Consulting Group, comparing the performance of those who used GPT-4 with those who did not. The consultants were assigned two types of tasks: one that AI handles relatively well, such as developing new product ideas, and another that AI struggles with and requires greater human judgment, such as solving a client’s business problem.
The results differed sharply. On the product-development tasks, consultants using AI completed 12.2 percent more tasks on average and worked 25.1 percent faster. But on the business problem that fell outside AI’s strengths, those using AI were 19 percent less likely to reach the correct answer.
The findings underscored a central challenge of using AI at work: Productivity gains depend on workers being able to recognize when the technology can help and when human judgment should take precedence.
That ability itself could be at risk as workers become increasingly dependent on AI. Researchers have raised concerns about “de-skilling” — the gradual loss of proficiency as people perform certain tasks less often themselves. As workers lose expertise, they may also become less able to evaluate AI’s output or decide which tasks should be delegated to it, potentially creating a cycle of growing dependence.
Research highlighted by the American Psychological Association (APA) has similarly raised concerns that excessive reliance on AI can weaken critical thinking and job-specific expertise.
“At the moment, most interactions between people and generative AI are transactional. You ask a question; you get an answer; you move on,” Mutlu Cukurova, PhD, a professor of learning and AI at University College London who studies human-AI interaction, told the APA in its July report. “Unfortunately, these types of interactions are more likely to lead to cognitive atrophy than cognitive development.”
Signs of de-skilling have emerged even among experienced professionals. In a study published in the British medical journal The Lancet in August 2025, Polish endoscopists who had each performed more than 2,000 colonoscopies used AI-assisted polyp-detection equipment for three months before returning to procedures without AI. Their adenoma detection rate fell to 22.4 percent from 28.4 percent before they began using the technology.
Experts say preventing such declines requires workers to retain what they call “cognitive initiative” — continuing to think through problems themselves rather than routinely surrendering that process to AI.
“If you depend on AI out of habit, you can’t help panicking when you run into a sudden problem,” said Kim Yi-kyung, a professor of education at Chung-Ang University. “Even when using AI for study or work, you have to keep a habit of thinking actively, with the awareness that you are the one solving the problem, to secure cognitive initiative.”
