First-of-Its-Kind AI Voice Tool Redefines GI Procedure Prep at UC San Diego Health

At UC San Diego Health, a new AI-powered voice tool is addressing a long-standing operational challenge: how to reliably prepare patients for GI procedures before they arrive.

Pre-procedure phone calls are a standard part of patient safety protocols, but in practice they are difficult to complete consistently. Patients may not be available during the day when the calls are typically made, contributing to no-shows, late cancellations and underutilized procedure time.

The tool, built with support from a team led by JCHI Co-Director Jeff Pan, is a direct-to-patient conversational AI system. The tool calls patients ahead of colonoscopy or EGD appointments and delivers tailored instructions using data from Epic, including preparation steps, medication guidance and logistics. Clinical teams can review call outcomes and follow up when necessary.

“The team became energized by the creative ideas and collaboration between Operations and IS,” said Shannon Brown, Executive Nursing Leader at UC San Diego Health. ​“The momentum started to build and physicians leaned into the discussion offering their perspective. We had daily engagement meetings which allowed for rapid cycle improvements. We surfaced workflow challenges and the technology shed light on gaps in our patient communication.”

With the new AI GI tool, UC San Diego Health care teams, including nurses, are working alongside the technology, not apart from it. They remain deeply involved in patient care, using AI to connect with more patients than ever before. The tool extends their reach, allowing our teams to engage a broader patient population, while nurses continue to handle follow-up calls that require specific, personalized information, now in a much more efficient way. 

Early results from a pilot phase showed a measurable impact.

“The metrics have been unbelievable,” Brown said. ​“We piloted a small group for several weeks and rapidly observed the no show/​late cancels percentage fall. The technology identified which patients actually required follow up with a nurse, allowing time to be more effectively spent. It also became evident that our ambulatory nurses were better suited to communicate with patients as the procedural nurses did not have skills in coordinating ambulatory care.”

The project was developed through close collaboration between clinicians, operational leaders and data scientists. Jie Cao, Ph.D., who led the data science work, focused on translating clinical workflows into a structured, patient-facing format while standardizing pre-procedure instructions across different GI pathways. The AI tools have been tested extensively, with input from our patient experience team, human factors engineers and designers.

As an AI flow builder, Cao partnered closely with clinical stakeholders to translate GI pre‑procedure preparation guidelines and patient‑specific context into a structured, conversational workflow that patients could easily understand and follow.

Cao said the work behind this revolutionary tool is very much a team effort across Epic analysts, engineers and data scientists.

“I’m energized by the collaboration behind this work,” she said. ​“A single call to a patient touches clinical guidelines, EHR data, conversational design, and operational workflows, getting that right requires genuine partnership across teams, and that’s what makes this kind of innovation both impactful and rewarding.”

For Rohit Loomba, MD, delivering timely care to patients is a top priority for all physicians, and this new tool helps fulfill that. 

“However, the complexity of an academic medical system can be burdensome for patients, so the introduction of AI to ensure patients receive necessary procedures timely and arrive medically prepared is excellent,” said Dr. Loomba, Chief, Division of Gastroenterology and Hepatology and Professor Medicine, UC San Diego. ​“Clarity of communication is critical for patients and timely pre-procedure calls reinforcing bowel preparation and medication instructions is essential to optimal outcomes.”

The tool is now being deployed across multiple UC San Diego Health campuses, with plans to expand further. The team says it reflects a broader effort to apply AI in targeted ways that improve both patient experience and clinical workflow.