Is AI Deskilling Healthcare Workers? What Nurses and Clinicians Should Know

A new survey is putting a name to a worry many nurses and doctors already feel in their gut. As clinicians lean on artificial intelligence for everything from charting to literature reviews, a growing share fear they are quietly losing the very skills that make them good at their jobs. According to the Future Ready Healthcare survey from Wolters Kluwer Health, covered by Healthcare Dive, 74 percent of clinicians now worry about deskilling, the slow erosion of critical thinking and decision-making that can come from leaning too hard on automated tools. The same survey found AI use exploding inside hospitals and clinics, which makes the question urgent rather than hypothetical. At Metaintro, we track how AI is reshaping real careers, and this story matters because healthcare is one of the largest employers in the country and one of the fastest to adopt these tools.

What the Survey Actually Found?

The Wolters Kluwer survey, fielded by Ipsos in March 2026, polled more than 350 healthcare professionals and over 250 patients across the United States. The headline is the speed of adoption. Roughly 73 percent of doctors and 70 percent of nurses said they now use AI at work at least once a week. A year earlier, those figures were just 38 percent for doctors and 46 percent for nurses. Daily-or-more use jumped too. About 38 percent of doctors now reach for AI multiple times a day, up from 10 percent, while 32 percent of nurses do the same, up from 16 percent.

The starkest evidence comes from outside the survey. A multicentre observational study in The Lancet Gastroenterology & Hepatology tracked 19 experienced endoscopists at four colonoscopy centres in Poland before and after they began using AI-assisted colonoscopy. Each had already performed more than 2,000 procedures, yet their unassisted adenoma detection rate, the share of exams in which they spotted precancerous growths without the software, fell from 28.4 percent to 22.4 percent after AI was routinely introduced. That is a 6 percentage point drop in cancer-relevant detection among seasoned doctors working without the tool, and it happened in a matter of months.

Researchers called it the first real-world documentation of automation-induced deskilling tied to patient outcomes in medicine. The mechanism is intuitive. When a tool reliably highlights what to look for, the human stops practicing the search, and the underlying skill atrophies. Independent experts gathered by the Science Media Centre urged caution on the size of the effect, but coverage in STAT and TIME framed the finding as a warning shot for every specialty rushing to adopt AI, not just gastroenterology.

The same dynamic threatens nurses. Researchers have long warned about automation bias, the tendency to accept an algorithm's output without questioning it, and alarm fatigue, where a flood of AI-generated alerts dulls vigilance rather than sharpening it. A National Nurses United survey found that most nurses say algorithmic patient-acuity scores do not match their own bedside assessment, a sign of how easily software can override hard-won judgment. A nurse who learns to trust a sepsis-prediction model may slowly stop running the mental checklist she once did by reflex. The skill is still in there, but it gets rusty when it is rarely used.

What Does Deskilling Actually Look Like?

The pattern is not unique to medicine, which is part of what makes the colonoscopy data so credible, a point echoed in commentary from the American College of Gastroenterology. Aviation studied the same effect decades ago and found that pilots who relied heavily on autopilot saw their manual flying skills degrade, which is why airlines now require periodic hand-flying to keep proficiency current. Healthcare is only beginning to think in those terms. The lesson from other high-stakes industries is that automation does not remove the human, it changes what the human needs to practice, and skipping that practice is where the danger lives. A clinician who never deliberately works without the tool may not notice the gap until the tool fails or produces something wrong.

Radiology offers another preview of where this can go. In a controlled study published in Radiology, 27 radiologists read mammograms with AI assistance, and when the system offered an incorrect suggestion, accuracy dropped sharply across every experience level, with the least experienced readers falling the furthest. That is automation bias in action, and it shows the deskilling risk is not limited to colonoscopy or to junior staff. Experts warn that introducing AI support too early in training is especially dangerous, because residents may never properly learn to read subtle, atypical findings if the software always flags them first. The skill cannot atrophy if it was never built, which is the version of the problem that worries educators most.

Why Are Younger Clinicians Most Worried?

One of the most striking findings is who is most anxious. Deskilling was the top concern for younger providers, with 33 percent ranking it among their top two worries, compared with just 11 percent of the rest of the provider pool. That flips the usual assumption that older workers fear new technology most.

The logic makes sense once you think about it. A veteran clinician built decades of pattern recognition before AI existed, so the tools augment an already deep base of judgment. A newer nurse or resident is still forming that base, and if AI does the reasoning during those formative years, the foundation may never fully set. Early-career professionals seem to sense that they could become dependent on tools before they ever develop independent expertise.

That concern points to a training problem more than a technology problem. The American Academy of Nursing approved a formal AI position statement in February 2026 that makes human nursing judgment the priority and calls for human-in-the-loop oversight in every AI governance policy. The message from the profession is that AI should support clinical reasoning, not quietly replace the chance to build it.

Are Hospitals Ready to Manage This?

Mostly, no. The survey exposed a wide governance gap. Only 27 percent of physicians and nurses said they understood their organization's approach to AI governance, a figure that crept up only modestly from 21 percent the prior year. Among those who were informed, just 63 percent understood how HIPAA applies to AI, only 35 percent knew the guidelines for verifying AI accuracy, and a mere 22 percent reported clear employer policies defining who is responsible when clinician and AI disagree.

The survey also makes clear that clinicians are not blindly trusting these tools. About 74 percent flagged AI hallucinations, confident-sounding answers that are simply wrong, as a major worry, and 77 percent said they already double-check AI output against original sources or trusted medical databases. Roughly 53 percent want any clinical AI to show the detailed reasoning behind its answers rather than a bare recommendation. The Fierce Healthcare coverage frames this as a healthy skepticism that hospitals are failing to support, because the verification habits clinicians describe take time and training that most organizations have not built into the workflow. The result is a workforce doing the safety work informally, on its own, with no policy backing it up. That mismatch is precisely the kind of gap that burns people out and exposes them to blame when something goes wrong.

Dr. Peter Bonis, Chief Medical Officer at Wolters Kluwer Health, tied the surge in usage to real demand rather than hype, saying it reflects "a combination of increased exposure, increased familiarity. But really importantly, it's addressing an unmet need." Greg Samios, CEO of Wolters Kluwer Health, was blunter about the downside, noting that rising real-world use "comes with a significant trust gap."

The accountability question is the one that should worry workers most. If an AI tool contributes to an error and no policy spells out who owns the decision, the clinician at the bedside is often left holding the liability. More than 90 percent of clinicians in the survey said human experts should validate AI-generated healthcare content, a near-universal demand that current hospital policies are nowhere close to meeting. The survey also found that about 72 percent of clinicians worry about advertiser-driven bias creeping into AI-generated information, a reminder that not every model answering a clinical question is neutral or built for medicine.

This gap between fast adoption and slow governance is where the career risk and the career opportunity both sit. A workforce racing ahead of its own rulebook is a workforce that badly needs people who can write the rulebook. Clinicians who understand both the medicine and the technology are rare right now, and that scarcity is exactly what makes the skill valuable. The hospitals that figure this out first will lean on the staff who already speak both languages.

What Does This Mean for Your Career?

It also means building a new, hireable skill on top of clinical training, and that skill is AI fluency. Knowing how to prompt a tool, recognize a hallucination, and verify an output is fast becoming a core competency, and the survey's governance gaps suggest most employers desperately need staff who can do it. Workers who can bridge bedside judgment and AI oversight will be the ones promoted into the safety, informatics, and quality roles every health system is now scrambling to fill. Nursing informatics, clinical documentation improvement, and AI safety committees are all growing functions, and they reward exactly the hybrid profile this survey describes. If you are mapping a healthcare path, our guides to high-demand nursing careers and pharmacy careers can help you find the corners of the field where this expertise pays off.

The good news for anyone early in training is that the foundation still matters more than the software. Models change every few months, but the ability to take a history, read a patient, and reason through an ambiguous case is durable. Build that core first, then layer the tools on top, and you end up with the profile employers will fight over rather than one they can automate around.

There is also a job-security angle. AI is not eliminating the need for licensed clinicians, because someone with accountability has to validate what the machine produces. The roles most exposed are the ones that are purely routine and rule-based, while jobs that pair human judgment with technology are getting more valuable, not less. Demand for skilled clinicians remains strong even as automation spreads, which is one reason healthcare keeps appearing on lists of the most AI-resistant careers.

How Can Workers Protect Their Skills Right Now?

Start small and practical. Pick one or two clinical tasks you do with AI and occasionally do them the old way first, then compare. That deliberate friction keeps your reasoning muscles working. Push your employer for the training the survey shows is missing, because formal education on spotting hallucinations and verifying outputs protects both you and your patients.

Document your involvement in AI rollouts too. As hospitals scramble to write the governance policies most still lack, the clinicians who help shape those rules become indispensable and visible to leadership. That visibility matters in a field where layoffs, like the recent cuts at major pharmacy chains, remind workers that no role is fully insulated. Pairing irreplaceable clinical judgment with fluency in the tools is the surest way to stay both safe and valuable as AI keeps spreading through healthcare.

People Also Asked

Q: What does deskilling mean in healthcare?

A: Deskilling is the gradual loss of a clinician's critical thinking and hands-on judgment caused by over-relying on AI tools. A Lancet study found doctors detected fewer precancerous growths on their own after months of routine AI assistance, the first real-world evidence the effect is real.

Q: Will AI replace nurses and doctors?

A: No, not in the near term. AI is automating routine tasks like charting and literature summaries, but licensed clinicians are still needed to validate outputs and own accountability. The roles most at risk are purely rule-based ones, while jobs blending human judgment with technology are becoming more valuable.

Q: How can clinicians avoid losing their skills to AI?

A: Periodically perform clinical tasks manually before checking the AI, push employers for training on spotting hallucinations, and build AI fluency as a hireable skill. Clinicians who pair durable bedside judgment with the ability to oversee AI are the ones getting promoted into safety and informatics roles.

Future-proof your career with Metaintro, AI-powered tools that help you adapt and thrive. Sign up to stay ahead.