As patient portal use continues to grow, Nashville, Tenn.-based Vanderbilt University Medical Center is exploring whether AI can help patients get answers faster while reducing administrative burden on clinicians.
The health system launched an AI-powered assistant inside its patient portal on April 1 at Primary Care One Hundred Oaks, marking the first patient-facing AI tool deployed at scale by Vanderbilt University Medical Center.
Integrated into My Health at Vanderbilt, the assistant activates when a patient selects “medical question.” The tool then asks follow-up questions tailored to the concern, including about symptoms, medication issues and postoperative updates, and drafts a more detailed message for the patient to review before sending. Patients can opt out, edit the draft or submit their original message unchanged.
CIO Neal Patel, MD, of Vanderbilt University Medical Center said the impetus for the tool was a well-documented friction point in portal messaging: the repeated back-and-forth exchanges between patients and providers that often stretch a simple question into a days-long thread.
“More often than not, as patients are sending messages, there’s lots of back and forth that results in clarification of the message or asking subsequent questions,” Dr. Patel told Becker’s. “Our thought was it often takes too much time back and forth to get the patient the answer they need.”
The burden falls on both sides, Dr. Patel noted. Physicians absorb a growing volume of portal messages during time outside of direct patient care. The goal, he said, is to have the AI collect enough information upfront so that when a message reaches a provider, it arrives with enough context to be resolved in a single response.
The assistant uses a large language model from OpenAI through a Microsoft Azure cloud environment already in use at Vanderbilt University Medical Center for clinical applications. The health system said patient information is not retained by the AI-powered virtual assistant. The tool incorporates Vanderbilt’s triage protocols and nursing triage guidelines through retrieval-augmented generation and is designed to avoid providing medical advice.
Dr. Patel did not share specific adoption figures but said uptake across participating clinics has met expectations, though the rollout remains in a beta phase and the health system is moving carefully. The two metrics Vanderbilt is tracking most closely are time to resolution and repeat adoption.
“What we’d love to see is that the question resolution time from initial message to final message on a given topic is reduced markedly,” he said. “If we find that the patients are choosing this mechanism, repeat customers per se, then that tells us that it is palatable and something that is potentially even a preference for patients.”
The health system is watching closely for unintended consequences. Dr. Patel said the stakes for patient-facing AI are higher than in administrative settings, where errors carry less clinical risk. That calculus is driving a measured pace that he said stands in contrast to how quickly the underlying technology has matured.
“We are really moving slowly, but at the same time, what’s been neat, given that even though we’re being very methodical, we actually have been able to create the concept and then begin to deploy it and begin to evaluate it much faster than a lot of things in the past,” Dr. Patel said. “The tools just become so much better.”
Looking ahead, Dr. Patel framed AI as an augmenter rather than a replacement, a tool for reducing the administrative and bureaucratic load on clinicians so they can focus on clinical work and for giving patients more self-service access to their own care information. He also flagged a broader challenge the industry has yet to resolve: patients are increasingly turning to large language models for clinical guidance, and health systems need to offer a trustworthy alternative.
“Our goal is to provide as effective tools as we can, so that the patients engage more directly with us, but on the terms that they want to dictate,” Dr. Patel said.
At the Becker's 11th Annual IT + Revenue Cycle Conference: The Future of AI & Digital Health, taking place September 14–17 in Chicago, healthcare executives and digital leaders from across the country will come together to explore how AI, interoperability, cybersecurity, and revenue cycle innovation are transforming care delivery, strengthening financial performance, and driving the next era of digital health. Apply for complimentary registration now.
