Two survey results are sitting side by side right now, and most healthcare executives are reading them wrong.
In a National Nurses United survey of more than 2,000 nurses, six in ten said they do not trust their employer to prioritize patient safety when using AI. Read that carefully and notice where the distrust lands: on the employer.
Meanwhile, Healthcare IT News reports that among nurses who use AI tools regularly, 67% believe AI will help the workforce. Only 16% of surveyed nurses said it will hurt.
Hold those two findings together and the picture changes. The people closest to the technology are not the ones most afraid of it. Familiarity is producing optimism, not dread. What frontline clinicians distrust is what leadership will do with the tool, and that distrust is being earned one rollout at a time.
What earned it
This week, The Markup reported on Kaiser Permanente nurses who say technology is making their jobs and patient care worse. The details deserve attention from every health system leader, because none of them are about diagnostic accuracy.
Nurses described call-length tracking that flags them for evaluation when a patient call runs past fifteen minutes. Algorithmic systems predicting their daily productivity. An AI tool, piloted in 2024, that rates their empathy and tone of voice.
Sit with that last one. A nurse spends twelve hours advocating for patients, and a piece of software grades her warmth.
In the same NNU survey, two-thirds of nurses said their own clinical assessment has at some point disagreed with a computer-generated prediction about a patient. That number should not alarm anyone; disagreement between an experienced clinician and a statistical model is exactly where judgment lives. What matters is what happens next. California lawmakers are now weighing legislation to protect doctors and nurses from retaliation when they override automated care recommendations. Think about what it means that a law like that feels necessary. Somewhere between the pilot program and the press release, clinicians concluded that disagreeing with the algorithm could cost them.
Every rollout is a message
I write persuasion for a living, so let me name what is actually happening here. Every AI deployment doubles as a piece of communication, and your staff reads it more closely than any memo you will ever send.
When the first AI tool a nurse encounters measures her call time, the message received is: we bought technology to watch you. When the tool scores her empathy, the message is: we trust software to judge the thing you have given your career to. No town hall can talk over that. The deployment already said it.
Now flip it. Let her first encounter be a tool that takes documentation off her plate, or flags a deteriorating patient she would want flagged, and the deployment says something else entirely: we bought technology to carry what was crushing you. Same budget line. Opposite sentence.
This is why the same technology produces strikes in one system and adoption in another. Health systems are not choosing between AI and no AI anymore; roughly 80% of hospitals already use it somewhere. They are choosing what their deployments say, and most leaders have never once read their rollout as a message.
Three questions before your next deployment
If you lead in healthcare, put your next AI initiative through these three questions before anyone signs.
Who named the problem? If the answer is a vendor or a spreadsheet, stop. The strongest signal in all the current reporting is that adoption follows co-design. Clinicians who help choose the problem worth solving become the tool's defenders. Clinicians who receive it from above become its critics, and their patients hear the criticism.
Can your people override it without fear? Two-thirds of nurses have already disagreed with an algorithm at the bedside. That will keep happening, and you want it to. The question is whether your system treats an override as clinical judgment or as non-compliance. Decide this in writing, before deployment, and say it out loud. If a state legislature has to guarantee your nurses that right, you have already lost the trust the tool needed.
What does it measure? Measurement is a values statement. A tool that tracks minutes-per-call tells your staff that speed is the point of care. A tool that tracks time returned to the bedside tells them the patient is. Your people will find out what the dashboard displays within a week of go-live, and they will conclude, accurately, that it displays what you value.
The part no vendor will tell you
The technology is going to keep getting better. The trust arrives separately, built or spent in how you deploy, what you measure, and what happens to the first nurse who overrides the machine and turns out to be right.
Six in ten nurses distrusting their employers with AI is a leadership grade, posted publicly, and every health system is currently earning its own. The systems that thrive in the next five years will be the ones whose clinicians say "this place uses AI to back me up." That sentence cannot be purchased. It can only be deserved.
Deserve it on purpose.
If you lead people through change and this landed, follow my portfolio here on Authory. I write on leadership, persuasion, and the messages organizations send without meaning to.
Reyes