Scaling AI across a health system demands governance structures, cultural alignment and a willingness to redesign workflows from the ground up.
During a Becker’s Healthcare webinar sponsored by Nabla, two leaders from UC San Diego Health shared how they have transformed AI from a collection of point solutions into a true operating model and what it takes to make that shift stick.
Below are four key takeaways from the conversation.
1. AI decisions are health system decisions
UC San Diego Health’s approach to AI governance is built on the foundational principle that AI should be governed the same way the health system governs everything else.
“Getting rid of AI exceptionalism and making sure that we govern AI the same way that we govern other things is a key early step,” said Karandeep Singh, MD, chief health AI officer at UC San Diego Health. That philosophy shaped the design of UCSD’s AI Think Shop, a structured intake process that evaluates every proposed AI initiative not as a technical question, but as a health system decision.
Marlene Millen, MD, chief medical information Officer at UC San Diego Health, described the practical effect: embedding AI review into existing governance meetings, procurement workflows and medical records oversight rather than standing up a separate silo. “The best thing we did was think about it as part of our work and not a separate thing,” she said.
2. Ambient documentation as infrastructure
Ambient scribes became UCSD Health’s highest-profile AI rollout but what made it notable wasn’t just adoption. It was how quickly the tool graduated from use case to foundational infrastructure.
“AI scribes went very quickly from being cars to being the road,” Dr. Singh explained. “Scribes are the way that we are going to be able to improve our documentation based on what was actually discussed.” He described scribes as the platform on which future clinical AI will run, from accurate billing capture to real-time clinical decision support.
What drove adoption was organic demand unlike anything the team had seen before. Instead of the usual change management struggle, clinicians were asking to join wait lists. “I had to be careful to not put my name out there because everyone would start emailing — I want to be part of the pilot,” Dr. Millen said. That demand helped leadership make the case for enterprise-level investment, framing ambient AI as the cost of doing business rather than a departmental line item.
3. Workflow redesign
One consistent theme across the session was the danger of automating a broken process. Both panelists emphasized the importance of mapping workflows in detail — not as leaders describe them, but as frontline staff actually execute them.
“Go and watch someone go through the whole process,” Dr. Millen said. “You need to understand not just the best person and the best workflow, but go through the spectrum.” This includes understaffed days, surge scenarios and edge cases that expose where automation will fail.
Dr. Singh described the value AI adds to this layer as the ability to handle the messy, free-text, variable reality of clinical operations without first forcing everything into structured data. Now organizations can move straight from the current process to automation.
4. Why trust matters
As UCSD Health’s AI portfolio has grown, ensuring that both clinicians and patients understand how AI is being used and can trust it is paramount.
On the clinician side, Dr. Singh noted the growing reality of shadow AI — staff finding and using unapproved tools when health system options don’t meet their needs. His goal is to close that gap.
“I want to create an environment where we’ve got the menu of tools to choose from,” he said, ensuring that compliant options have feature parity with what clinicians would seek out on their own.
On the patient side, Dr. Millen emphasized transparency as a trust-building strategy. UCSD Health has standardized language clinicians can add to notes disclosing when AI was used — a practice she sees as fundamental to maintaining the patient-provider relationship.
“If we start losing trust, then people will start fighting back,” she said. “The proof is the feedback we get and the happiness of our providers that finally something is helping them and not adding stress to their day.”
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.
