For most nurses, when the shift ends, the work doesn’t. Instead of standing at the bedside, they’re sitting at a computer trying to reconstruct the last 12 hours from memory.
A nurse on a medical unit at Atrium Health Union in Monroe, N.C., part of Advocate Health, put it to us plainly: “The burden isn’t the work. It’s the after-work.” However, on a recent shift, that changed. Instead of rebuilding the story later, she charted in real-time, speaking her assessments out loud while an AI-powered, hands-free tool turned her words into documentation she then validated. By the time she handed off to the next nurse, her work was done. She went home on time. Her colleagues stayed behind, still typing.
Hospitals across the country are racing to put ambient AI tools like this one in front of nurses. The pitch is irresistible: less typing, more caring, finally a fix for the documentation burden straining the profession. But if these rollouts follow the path of the technologies that came before them—designed for the system, not the people inside it—they could deepen the very burnout they’re trying to relieve.
We have seen what the right path looks like and the role nurses must play. At Advocate Health, one of the largest nonprofit health systems in the country, ambient AI documentation is reducing charting time and returning nurses to the bedside with their patients. Sixty-five percent of nurses using the tool now activate it six or more times per shift—a level of voluntary use that is rare for any clinical software—more than 80 percent report meaningful time savings. After-hours charting is down. Patients tell us their nurses are more present at the bedside, listening rather than typing.
Those numbers reflect more than the technology itself. They’re the product of nurses helping design it.
Crucially, AI does not replace clinical judgment. Ambient tools capture and organize data; they do not make care decisions. At Advocate Health, nurses retain full control. Every AI-generated draft is reviewed and edited by the nurse who delivered the care—a non-negotiable principle of FAIR-AI, Advocate Health’s standard for responsible AI deployment, now a peer-reviewed model for other health systems.
Nursing is consistently ranked as the most trusted profession in America, and for good reason: no one knows their own work and their patients better than nurses. So, when Advocate Health rolled out ambient AI, nurses weren’t handed a finished product and
trained to use it. They were brought in from the start as co-designers, deciding when to activate the tool, how to narrate care without disrupting a patient encounter and how the data should flow into the electronic health record. Their judgment shaped the workflow, and the workflow shaped the technology.
Most hospitals are getting this backward, and nurses have sounded this alarm. In April, the American Nurses Association convened its inaugural AI in Nursing Practice Think Tank. Consensus findings read like a warning label for executives: erosion of clinical judgment from overreliance on AI outputs, unclear accountability when AI tools influence care decisions and increased cognitive burden from poorly implemented technology. These are the predictable consequences of deploying AI for nurses, rather than with them.
The stakes could not be higher. The U.S. is short hundreds of thousands of nurses, and the Bureau of Labor Statistics projects nearly 195,000 average annual openings for registered nurses every year through 2030. Burnout is one of the leading reasons nurses leave the bedside, and documentation is the leading driver of burnout. AI is one of the few interventions with the scale to shift that trend. If hospitals squander it on rushed rollouts, they will not get a second chance. The crisis is sharpest in rural communities, where a single nurse's departure can leave a unit short-staffed for months, and the pipeline for replacement is thin. For these communities, getting AI right is not just an efficiency question; it determines whether local residents have access to care at all.
Here’s what works: give nurses time to learn the technology in simulation before they use it on patients. Measure whether it returns time to the bedside, not whether usage hits a quarterly target. Design every step around the nurse’s clinical judgment. Treat AI’s output as a draft, not a verdict.
Hospitals that get this right will keep their nurses. Those that treat AI as a quick technical upgrade are likely to be disappointed, along with the nurses they aim to support.
The lesson isn’t new, and the industry keeps relearning it the hard way. If health systems want AI to succeed, they need to invest as much in the people doing the work as in the tools themselves.
Tracy Breece is vice president of nursing innovation, AI and emerging technologies at Advocate Health. Bradley Goettl, D.N.P., is chief nursing officer at American Nurses Enterprise.