Upon the technology’s arrival, Kensaku Kawamoto, MD, PhD, University of Utah Health’s inaugural chief health AI transformation officer, considered himself something of an AI skeptic.
The longtime clinical informatics leader spent a lot of his career building digital health tools and testing AI models that often looked promising in demos but struggled in real-world healthcare operations.
Now, he is leading the health system’s enterprise AI strategy in a newly created role. Dr. Kawamoto was appointed in April and sees opportunity in the technology as it has matured.
“What really changed was the emergence of transformer-based AI models and, more recently, reasoning models. Early versions of large language models made for impressive demos, but in real-world healthcare applications they weren’t quite ready for prime time,” Dr. Kawamoto said. “I was honestly somewhat skeptical of AI for a while because we tested many of these approaches and found they often didn’t perform reliably enough for operational use.”
According to the AI chief, reasoning models improved significantly last fall and hallucination rates dropped, allowing the organization to see evidence that the systems could work effectively in healthcare settings.
“That was the turning point,” he said.
In his role, Dr. Kawamoto said his aim is to help the health system become an AI-forward organization that can leverage the technology to advance University of Utah Health’s mission of clinical care, research and education.
“It’s obviously a broad scope, but in practice, the real change happens at the organizational-unit level throughout the health system, where the people and processes for change already exist,” he said.
On a day-to-day basis, that often means meeting with colleagues across the enterprise to identify recurring needs. University of Utah Health also has an in-house development capability which includes AI engineers, informaticists and other experts who help build solutions.
“Much of my day involves partnering with teams to develop internal capabilities when vendor-based solutions don’t fully meet our needs,” Dr. Kawamoto said.
Some of the projects for which the health system has prototyped AI tools include pathology workflows, ambient clinical note summarization and clinical trials recruitment.
One pathology-focused tool, built within a day and integrated into the organization’s Epic EHR, helps physicians review patient records during cancer evaluations and has significantly reduced chart review time.
“We found that these approaches reduced review time by roughly tenfold,” Dr. Kawamoto said. “Even more importantly, the AI systems consistently surfaced clinically relevant findings that physicians had sometimes overlooked.”
The health system has also built infrastructure that allows end users to create many of these AI tools themselves through simple configuration.
“A pathologist can now create a tool like this in just minutes,” Dr. Kawamoto said. “The longer part of the process is no longer technology development — it’s validation. We still rigorously test tools across large patient samples to ensure performance and safety before broader rollout. But the shift in timelines is incredible. Earlier this year, we were building these tools in hours or days. Now we’re approaching the point where some can be created in seconds or minutes.”
When it comes to where University of Utah Health is headed in its AI journey, Dr. Kawamoto said the organization is closely evaluating financial return on investment and identifying efficiencies that allow it to continue investing in AI capabilities sustainably.
The health system is also exploring how AI can support researchers.
“Many of these technologies are highly adaptable,” he said. “For example, the same AI workbench infrastructure we’re building for clinical applications is also being used for areas like clinical trial recruitment in partnership with our cancer institute.”
Ultimately, Dr. Kawamoto said the goal is not AI for AI’s sake.
“The goal is advancing the mission of the institution. Success means helping leaders, researchers, clinicians and educators accomplish what they’re already trying to achieve — and using these technologies safely and responsibly to help move that work forward.”
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.
