Health systems have spent years building online self-scheduling so patients don’t have to pick up the phone. The next party trying to book an appointment may not be a patient at all, but an AI agent working on one’s behalf.
Consumers can already connect their medical records to frontier AI tools. OpenAI launched ChatGPT Health in January, letting users link records from U.S. providers, and Anthropic has said users can upload lab results and medical records to Claude.
Nabile Safdar, MD, chief AI officer of Atlanta-based Emory Healthcare, spoke with Becker’s about what that means for health systems at the publication’s 14th Annual CEO + CFO Roundtable in Chicago.
“I think there’s a lot of potential for us to be looking at access and scheduling. This isn’t something that, frankly, we can ignore forever because patients are trying to use their personal agents, right? Whether it’s by Meta or X or whoever,” Dr. Safdar said.
“We’re starting to see, particularly for those frontier labs that have enabled integrations with their EMR, we’re starting to see traffic,” he said. “This summer, we saw that, and that’s predictable, right?”
Consumer AI agents aren’t confined to healthcare, either.
“I know individuals who are using them quite a bit in their personal lives, managing a busy child’s or a group of children’s combination of events, assignments, due dates, etc., and then using them to call their cable company and try to negotiate a better rate or something,” Dr. Safdar said. “So I fully anticipate that they’re already trying to call our systems or call our people.”
That leaves health systems with a choice about what those agents will be allowed to do once they arrive.
“Those agents that patients are deploying themselves and authorizing to do business on their behalf are going to be trying to interact with our people and with our systems,” he said. “So we’ve got some decisions to make.
“Do we allow them to continue to do that? Some financial institutions have said, ‘No. We’re not going to be allowing individuals’ personal agents, however authorized, to do financial transactions and make decisions.’ So we may say no. We may say, well, we’re going to deploy our own versions that allow pretty highly automated scheduling.”
For health systems on Epic, a version of that second path is already on the menu: The EHR vendor has made AI for self-scheduling available.
Letting outside agents in carries its own exposure.
“I think that could be very powerful in a good way. There are some risks to that as well,” Dr. Safdar said. “It’s a little bit of buyer beware on behalf of the patient.”
There is also the question of which patients any of this would serve.
“I think we have to be honest that there is a certain portion of our patient population that wants to use those yesterday, and then there’s a significant portion that doesn’t even know what I’m talking about,” he said. “So it’s striking the right balance of saying, what’s the right posture, and we don’t want to get distracted by something that’s only going to benefit the people who already have the most access through virtue of their privileged existence. So we’ve got to balance that correctly.”
For Dr. Safdar, the shift changes a long-held assumption about who controls healthcare technology.
“A lot of the power here, or power may not be the right word, but the agency, so to speak, is being decentralized,” he said. “We always assume that the healthcare systems are the ones that have the tech agency, but there’s been a little bit of a leveling of the playing field here. Patients and consumers have a lot of agency now too.”
