Kimberly Maurer

I have spent nearly 30 years in nursing. Technology has changed more times than I could count during those years, and nurses have learned how to work with every new system that has come through the door.
Generative AI will be another one.
I am not afraid of AI being used in healthcare. I use AI. I see where it can help. I also know what healthcare looks like outside of a controlled environment, and that is where I believe nurses bring something important to this discussion.
There is a difference between having information and knowing what to do with it.
A patient can have numbers that look perfectly acceptable while the nurse standing in the room knows something is wrong. Maybe it is the way the patient is answering questions. Their breathing looks different. A family member mentions a change that seems insignificant until you ask another question. Something does not fit.
Nursing judgment lives in those spaces.
So when a medical device using generative AI gives a nurse information, a recommendation or an interpretation, I want to know what happens when the nurse looking at the patient disagrees with it.
Is questioning the output easy?
Can the nurse see enough information to understand how the device reached its conclusion?
Is there a clear way to escalate a concern?
Where does that concern go?
Does anyone look at the near miss where the nurse caught the problem before the patient was harmed?
Those answers matter.
I work in care transitions now, and I see another part of this that deserves attention. Healthcare information does not stay in one room. Patients move between hospitals, home health, rehabilitation, specialists, primary care and their own homes. Information follows them, gets repeated, gets interpreted differently and occasionally gets lost.
An AI-generated conclusion that enters that chain can travel with the patient too.
If it is accurate, it may be incredibly useful. If it is wrong, who catches it before the next person treats it as fact?
That is where I want nurses involved in this framework.
Bring bedside nurses into the evaluation process. Include nurses from home health, hospice, emergency care, long-term care, outpatient settings and care transitions. Ask the people using these devices what happens once they leave the testing environment and enter an actual Tuesday afternoon in healthcare.
Training deserves the same attention. Teaching a nurse which buttons to push is not enough. Give clinicians a clear understanding of what the technology knows, where its limitations are and what to do when the output conflicts with their assessment.
Postmarket monitoring needs nursing input too. Near misses matter. Workarounds matter. Repeated concerns that never become adverse events matter. A nurse catching an error before it reaches a patient should be valuable information about the device, not a reason the problem disappears from the record.
Patients deserve consideration here as well. People should understand when generative AI is meaningfully involved in their care and where a clinician’s judgment enters the process. Trust gets complicated very quickly when people do not know who or what helped make a decision about them.
I have spent enough years in healthcare to know that no framework will anticipate every situation that eventually walks through the door.
Patients are human beings. They do not arrive as clean datasets.
The nurse standing in front of that patient carries something a device does not: context from the room, the conversation, the family, the last assessment and the thousand small pieces of information gathered while caring for another person.
Generative AI may become an extraordinary clinical tool. I hope it does.
I also hope we build the guardrails while nurses are in the room, rather than asking them later to compensate for what was missed.
One of the best pieces of advice I have received in my career was simple:
Keep asking questions until there are no more questions to ask.
Please build a framework that leaves room for nurses to do exactly that.