It turns out, the fear isn't just anecdotal. A survey of US healthcare workers, published just this month in June 2026, found that 70% of nurses and 77% of physicians worry about their skills deteriorating because they rely too much on AI, leading to concerns about AI skill erosion. That's a huge number, and it points to a real anxiety in professions where precision and human judgment are non-negotiable.
The Worry Isn't Just Talk: What the Studies Show
We're seeing concrete examples of this AI skill erosion in action. Take a study from Poland involving experienced endoscopy specialists. Before using AI assistance for detecting cancerous lesions during colonoscopies, their adenoma detection rate was 28.4%. After they got used to AI helping them, that rate fell to 22.4%. Even more concerning, The Lancet Gastroenterology and Hepatology reported that these doctors then showed decreased diagnostic performance when the AI technology wasn't there. It's like they became less sharp without their digital crutch.
It's not just medicine. In software engineering, Anthropic ran a randomized controlled experiment with 52 engineers on a basic coding assignment. The engineers using AI assistance finished tasks faster, which is great for short-term productivity. But when they were evaluated without AI help, their independent problem-solving performance was weaker. This shows that while AI can speed things up, it might also be quietly eroding the very skills we need to solve problems on our own. This issue of AI skill erosion is now showing up in fields like law, journalism, and banking, too.
Why This Happens: The Mechanics of AI Skill Erosion
So, what's going on here? It's not necessarily that people are forgetting knowledge. Instead, it's a change in behavior. Professor Robert Wachter from the University of California, San Francisco, suggests that clinicians might just pay less attention when they know AI is handling part of the work. Yuichi Mori from the University of Oslo adds that specialists can become "less motivated, less focused," and feel less responsible for cognitive decision-making when AI is always present.
Think of it like this: if you always use a calculator for simple arithmetic, you might get slower at doing mental math. Your brain starts to offload that task. With AI, we're offloading more complex cognitive tasks. When you continuously rely on a tool to do the heavy lifting, your own 'cognitive muscles' don't get the workout they need. They don't atrophy in the sense of disappearing, but they certainly don't stay as strong or responsive.
Beyond the Doom: How to Use AI to Get Better, Not Worse
This isn't a call to abandon AI. That would be unrealistic and counterproductive. The real challenge is understanding the difference between AI as an alternative to human thought and AI as an augmentation. When AI replaces your critical thinking, that's where the trouble starts. But when it acts as a safety net, a second opinion, or a tool to offload repetitive tasks, it can actually free you up to focus on higher-level problem-solving and skill development.
Kevin Crowston from Syracuse University makes a key point: we, as experts, need to decide which skills we absolutely must keep human. For instance, an AI can draft a legal brief, but a lawyer still needs to understand the nuances of the law and client context to ensure accuracy and strategy. A coder can use AI to generate boilerplate, but they still need to debug, architect, and understand the system deeply.
Instead of letting AI replace your thinking, consider how it can sharpen it. For instance, using AI for a second opinion isn't about accepting its answer blindly; it's about critically comparing its reasoning to your own, thereby forcing deeper engagement. Similarly, by automating the drudgery of repetitive, low-cognitive tasks, AI frees up your mental energy for the truly complex, creative, and strategic work that builds deeper expertise. You can even leverage AI as a powerful tutor, asking it to explain complex concepts, generate examples, or simulate scenarios, accelerating your learning curve in new domains.
What You Can Do Next
The evidence is consistent: if we're not careful, AI can quietly erode our professional skills. But this isn't a foregone conclusion. The future of our skills isn't about rejecting AI; it's about consciously integrating it. It means being deliberate about how you use these tools. Don't let AI become a substitute for your own thinking. Instead, make it a sparring partner, a research assistant, or a tireless intern that handles the grunt work, thereby mitigating significant AI skill erosion.
Beyond individual practices, organizations bear a significant responsibility. This means evolving training programs to focus not just on how to use AI, but on how to use AI to learn and grow. It requires designing workflows that actively encourage human oversight and critical evaluation, moving beyond mere blind acceptance. For those building AI systems, the imperative is to create tools that prompt users to engage their own expertise, rather than bypass it entirely.
Ultimately, while the early results are indeed concerning, highlighting a quiet AI skill erosion, this isn't a foregone conclusion. The goal isn't just to be faster, but to be smarter and more capable. AI can still help us achieve that, but only if we remain firmly in the driver's seat, actively steering our own skill development and consciously preserving the human expertise that truly matters.
