Epic has developed AI agents that can execute clinical protocols and is exploring how they could help health systems manage routine patient visits, a company executive told Becker’s.
Epic founder Judy Faulkner raised the possibility of an “Epic doctor” while discussing the company’s AI strategy during an Oct. 13 fireside chat for Harvard’s Clinical Informatics Lecture Series, according to a health system executive who attended. Asked what that would entail, Garrett Adams, Epic’s vice president of research and development, set the name aside and described AI agents that could handle portions of routine visits under human-defined protocols, with physicians retaining oversight.
Epic has not set a timeline for making the capabilities available, and Mr. Adams did not identify any organizations using them in patient care. The company is in a discovery phase with health systems, he said.
Mr. Adams said the work is driven by two access challenges. The first is delays in getting appointments, compounded by current and projected workforce shortages across healthcare. The second is patients who may lose access to routine care because they lack insurance or are affected by changes in government funding.
Mr. Adams said physicians in some specialties report that as much as 40% of their visits involve lower-acuity, routine follow-up care. Using AI agents to help manage portions of those visits could free physicians to see more higher-acuity patients sooner.
He described a potential workflow in which health systems establish clinical protocols and AI agents gather relevant patient information, including laboratory results, imaging or cardiology studies, and evaluate it against those protocols.
The technology could then present its findings to a physician, along with information about the data reviewed and the protocol followed, allowing the clinician to make a determination with less time spent gathering and analyzing information.
In some cases, Mr. Adams said, the approach could spare patients an appointment altogether, such as saving a rural patient an hourlong drive to the hospital when the matter could be handled remotely.
Mr. Adams said Epic has already proven out some of the underlying technology.
“We have agents that can execute protocols,” he said.
However, he emphasized that proving the technology is different from finding a way to use it safely and effectively in clinical settings. Epic is now exploring how to operationalize those capabilities while maintaining quality and trust.
Mr. Adams said the concept builds on Epic’s existing AI work, including Agent Factory, which orchestrates agentic workflows; Art, which focuses on clinical intelligence; and Emmie, which supports clinical intelligence as well as patient communication and outreach.
He described the capability as a combination of pieces of those tools with a new protocol component and new mechanisms for presenting results and following up.
Epic has been in conversations with many health systems to understand which clinical workflows they would prioritize, Mr. Adams said. Interests vary by organization, he said, noting that customers from the United Arab Emirates he met with a few months ago had different priorities than “our friends from UW Health right down the street.”
Those conversations include questions about when an AI agent should escalate a case to a clinician if a protocol isn’t met, how much control individual health systems should have over the process and when patients should be able to choose between trying the technology and seeing a human clinician.
Rather than building for a single use case, Epic is developing what Mr. Adams called a “generically capable engine” that could be deployed across many workflows following a similar pattern: a protocol, data gathering and execution of the protocol.
Mr. Adams pointed to medication refills and certain routine urgent care encounters, which often involve lab work followed by a straightforward protocol, as areas of potential interest.
He emphasized that Epic is not envisioning AI independently conducting patient visits without human oversight.
“You could think [of] one end of the spectrum where AI is just going rogue and doing visits. That’s not what we’re talking about here,” he said.
Instead, the approach would rely on protocols established by humans and policies defined by health systems, with clinicians retaining oversight.
Mr. Adams said Epic intends to move gradually, emphasizing safety, transparency and trust as it explores potential clinical applications.
The post Epic explores AI agents for routine visits appeared first on Becker's Hospital Review | Healthcare News & Analysis.