For years, AI has been waiting for instructions. Now, it’s writing its own to-do list.
At least, that’s the idea behind AI agents. This is a more autonomous type of AI that can use reasoning and planning to solve multistep problems, compared to generative AI that only creates outputs when specifically prompted, though definitions vary, Tech Brew reported in February.
Healthcare is buzzing with the possibilities. Electronic health record giant Epic, for instance, announced at the March HIMSS25 Global Health Conference and Exhibition that it plans to create AI agents to help patients prepare for medical appointments, Fierce Healthcare reported. These agents can independently take a number of steps, including identifying and scheduling tests—something that non-agentic AI, like a chatbot or note-taking scribe, might need to be prompted to do.
“Generative AI, it’s very task driven, be it you’re saying ‘Write me a letter’ or ‘Summarize this document,’” Aashima Gupta, global director of healthcare strategy and solutions for Google Cloud, told Healthcare Brew at HIMSS25. “Now we are in the era of agentic AI. Agentic AI is not just simple task-driven search and synthesization. It is actually an agent, your personal collaborator.”
But what does that actually mean?
How healthcare can use agents
Think of it this way, Mika Newton, CEO of data platform xCures, said: An AI-driven medical record aggregation and information retrieval platform could replicate an employee through agentic AI rather than just helping that employee complete their tasks.
His company, for instance, uses agentic AI to extract info from medical records and fill in healthcare forms.
“I try to think, ‘What is the persona or person that we are trying to train this AI to simulate?’ So, in our case, it’s the reader and re-typer of the medical record,” he said.
Don Woodlock, head of global healthcare solutions for data technology provider InterSystems, told Healthcare Brew at HIMSS25 that surgery is another good example of how agentic AI can help healthcare providers.
Navigate the healthcare industry
Healthcare Brew covers pharmaceutical developments, health startups, the latest tech, and how it impacts hospitals and providers to keep administrators and providers informed.
“If you were to come in for surgery, there’s all the presurgical testing, and the visits, and booking the room, and booking the resources,” he said. “You can just tell an [AI] agent: ‘Hey, get this patient ready for surgery. Here’s the kind of surgery. You go handle the rest.’”
Agentic AI is a potential solution for the current and growing healthcare workforce shortages, Newton said. By 2037, the National Center for Health Workforce Analysis predicts a shortage of 187,130 physicians, 207,980 registered nurses, and 302,440 licensed practical nurses.
It could also be a way to reduce clinician burnout. A survey of 12,400 physicians from the American Medical Association found that 48.2% reported at least one symptom of burnout in 2023.
New era, new problems
Of course, nothing is perfect. “Safety and validity” are probably people’s “biggest concern” about agentic AI, Newton said.
Of concern is how to keep these autonomous AI agents accountable and make sure they aren’t operating on false information or producing hallucinations, which is when the AI creates a fabricated output.
There are guardrails to address these issues, Gupta said. Developers have reduced agentic AI’s hallucination risk by building AI that cites sources for its answers, allowing clinicians to double-check a program’s work, for instance.
Another concern is that agentic AI will remove humans from the loop, the World Economic Forum wrote in a December 2024 post. The group said organizations should consider human oversight of every AI decision and establish “clear ethical guidelines” as to how it should operate.
In line with that thinking, Gupta agrees that humans should play a big role in agentic AI’s future.
“Healthcare is a complex thing. We need human judgment, empathy,” she said.