A new Wolters Kluwer Health survey, published in early June 2026, puts a number on something a lot of people have felt and few have measured. Doctors and nurses are reaching for AI constantly now. And the risk they name most often has little to do with a bad answer from the tool. The thing they fear losing is their own judgment.
Ipsos surveyed 355 clinicians and 254 patients across the country in mid-March for the report. Two findings sit side by side, and reading them together is the whole story.
First, daily use exploded. The share of doctors using AI multiple times a day jumped to 38%, up from 10% a year earlier. For nurses it doubled, to 32% from 16%. Roughly three-quarters of doctors and 70% of nurses now use AI at least once a week at work.
Second, 74% of clinicians said losing critical thinking and decision-making skills will be one of the greatest AI risks ahead. The report has a word for it: deskilling. The worry that leaning on a tool every day quietly erodes the skill the tool was supposed to support.
So the people adopting AI fastest are also the ones naming the most pointed fear about it. Read together, those two findings describe one thing: what happens when use climbs faster than proficiency.
of doctors now use AI multiple times a day, up from 10% a year earlier. For nurses the figure doubled, to 32% from 16%.
Source: Wolters Kluwer Health, 2026Adoption is a number. Proficiency is a practice.
It is easy to read a chart like this and assume heavy use means high skill. The survey suggests the opposite can be true.
Confidence is high: 73% of clinicians said they were somewhat or very confident they could spot an incorrect AI response. That leaves about a quarter who are not sure they could catch a wrong answer. In medicine, a wrong answer that slips through is not a typo.
The governance picture is thinner still. Only 27% of doctors and nurses said they knew how their workplace was addressing AI governance. Among those who did, 63% said they understood how privacy rules like HIPAA apply to AI use. So a small slice of clinicians has clear guidance, and even inside that slice, roughly one in three is unsure about the privacy rules.
Put plainly: the tools arrived first. The training, the verification habits, and the written rules are still catching up.
Deskilling is what heavy use without proficiency looks like
Here is the helpful read on the deskilling worry. Treat it as a sign that the skill underneath the tool needs deliberate attention, rather than as an argument for using AI less.
Think about a clinician who asks AI to summarize the latest literature, drafts a note from it, and moves on. Do that once and it is a time-saver. Do it daily, without ever checking the source, questioning the summary, or knowing when the model tends to invent a citation, and the muscle that used to do that checking starts to soften. The tool is doing the thinking. The person is approving it.
The fix keeps the tool running and keeps the judgment switched on at the same time. That is a learnable practice, and it has a name in The 7 Levels of AI Proficiency.
The tool is doing the thinking. The person is approving it.
Where this maps in The 7 Levels of AI Proficiency
The 7 Levels of AI Proficiency is a measurable model for how a person grows from first awareness of AI to orchestrating it across a whole operation. Inside it, critical thinking is a defined stage with a name, rather than a vague virtue.
In The 7 Levels of AI Proficiency, Level 3 is The Lieutenant, the Critical Thinker. A Lieutenant interrogates the answer instead of accepting it at face value. They notice when a confident reply is built on a thin source. They know the tool's failure patterns and check the places it tends to slip. The emotional skill underneath this level is self-management: the discipline to slow down and verify when the easy path is to accept and move on.
The deskilling fear in the survey is, in plain terms, the fear of skipping Level 3. When use races from awareness straight to daily reliance without building the Critical Thinker stage in between, the judgment that catches a bad answer never gets strong enough to do its job.
That is the quiet lesson in the data. The clinicians are not wrong to worry. They are pointing at a real stage that heavy adoption can leap over. The good news is that the stage can be built on purpose, and the survey shows clinicians already sense where the work is.
What this means beyond healthcare
If you lead a team in any field, the healthcare numbers are a preview. Adoption in your shop is probably climbing too. The questions the survey raises travel:
Do the people using AI every day know its failure patterns well enough to catch a wrong answer? Is there a written, understood policy for how AI gets used with sensitive information? And is anyone building the verification habit on purpose, or is everyone assuming heavy use will teach it by itself?
Heavy use does not teach judgment by itself. Judgment is built. That is the difference between a team that uses AI and a team that is proficient with it.
What to do this week
Three reps you can run this week, whatever field you are in.
-
Name one AI task you do without checking.
Pick the one place you accept an AI output and move on. Awareness of the habit is the first step out of it.
-
Run one verification rep.
On your next AI-assisted task, trace one claim back to its actual source before you use it. One rep, done on purpose, is how the Critical Thinker muscle rebuilds.
-
Find your AI rule, or notice there isn't one.
Ask whether your team has a written policy for using AI with sensitive or regulated information. If you are in the 27% who can find it, read it. If you are in the rest, that is the project worth starting.
Related reading: Level 3: Lieutenant.
Sources
- Wolters Kluwer Health, 2026 Future Ready Healthcare Survey Report (primary report page)
- Wolters Kluwer, Future Ready Healthcare Survey (report family page)
- Wolters Kluwer, AI survey insights: newer providers concerned about deskilling
- Healthcare Dive, Healthcare AI adoption accelerates as providers worry about deskilling (June 2, 2026)
- Fierce Healthcare, Wolters Kluwer Health survey examines AI use and concerns among clinicians, patients in 2026
Frequently Asked Questions
What did the Wolters Kluwer survey find about AI use among clinicians?
The 2026 Future Ready Healthcare Survey, fielded by Ipsos among 355 clinicians and 254 patients, found daily AI use rising sharply. Doctors using AI multiple times a day reached 38%, up from 10% a year earlier, and nurses reached 32%, up from 16%. About three-quarters of doctors and 70% of nurses use AI at least weekly at work.
What is deskilling and why are clinicians worried about it?
Deskilling is the erosion of a person's own skill from relying too heavily on a tool. In the survey, 74% of clinicians named losing critical thinking and decision-making ability as one of the greatest AI risks. The concern is that daily reliance without active verification softens the judgment needed to catch a wrong AI answer.
Does the survey say clinicians should use AI less?
No. The data points to a need for proficiency and governance to keep pace with adoption, not to a need for less AI. Only 27% of clinicians knew how their workplace handled AI governance, which suggests the rules and skills are still catching up to the usage.
How does this connect to The 7 Levels of AI Proficiency?
In The 7 Levels of AI Proficiency, Level 3 is The Lieutenant, the Critical Thinker, who questions and verifies AI output rather than accepting it. The deskilling fear is essentially the risk of jumping from early awareness to heavy daily use without building that Critical Thinker stage in between.
What is the single best first step for a leader?
Check whether your team has a written, understood policy for using AI with sensitive information, and build one verification habit on purpose. The survey shows most clinicians cannot find their organization's AI governance approach, and that is a fixable starting point.
Find your AI Proficiency level
The free 7 Levels assessment places you across seven stages of AI capability. Under ten minutes. Research-backed scoring.