
254: UI Health’s Dr. Bhrandon Harris: How a Safety Net Budget Forces Smarter AI Bets, The “Apology Factor” For When AI Can Talk to Patients, and The Hidden Capacity Question Behind Every SDOH Screen
- Written by
- SeamlessMD
- Published on
Subscribe on: RSS | SPOTIFY | APPLE PODCAST | GOOGLE | BREAKER | ANCHOR
On this episode of The Digital Patient, Dr. Joshua Liu, Co-founder & CEO of SeamlessMD, and colleague, Alan Sardana, chat with Bhrandon Harris, MD, Associate Chief Medical Information Officer at UI Health, about "How a Safety Net Budget Forces Smarter AI Bets, The 'Apology Factor' For When AI Can Talk to Patients, and The Hidden Capacity Question Behind Every SDOH Screen, and more..."
Audio:
Guest(s):
- Bhrandon Harris, MD, Associate Chief Medical Information Officer at UI Health
- Joshua Liu, MD (@joshuapliu), Co-founder & CEO at SeamlessMD
Episode 254 - Show Notes:
[00:05:48] Why AI can only do what an organization is ready for — Dr. Harris calls AI "really just a tool," and says any technology has to be assessed for whether it replaces something, enhances an existing process, or creates a new work stream that staff must be reoriented to. Healthcare "is not one thing, it's 10,000 different things," with micro workflows inside each workflow, so a tool that solves one part won't solve the whole, and if people have to "take even a half step to the side" to use it, it won't succeed.
[00:07:24] The real gap between AI vendors and clinics — rather than a disconnect, Dr. Harris says "a lot of vendors know exactly what they're doing. They just don't know who they're doing it for." Every clinic has a waiting room and exam room, but how patients get there and what rooming staff are responsible for differ, and varying digital health literacy among patients and team members all matters when scaling a tool.
[00:08:49] Why "no two people are the same" limits even well-built tools — clinicians want to innovate, but also want to get home to their families and do right by patients. "We're not building machines, we're not making cars, we're not making widgets." Patients with the same diagnosis can have completely different needs and barriers, and not every tool is nimble enough to support that, even when it works exactly as designed.
[00:10:28] How UI Health approached SDOH screening by defining the problem before the tool — before considering any tool, digital or not, Dr. Harris says you have to understand the gap you need to fill rather than "shoehorning something in." The team asked what it would do with the information if screening rates rose and how to absorb the added demand; the care coordination group owned the workflow and worked to expand its own capacity. The key was "spending more time asking the questions than looking for solutions."
[00:13:15] Why AI's speed, and the evidence behind it, demand careful workflow design — AI "just makes things go faster," and that speed can be imperfect, so workflows must accommodate it. Recalling his excitement when Fitbit came out, Dr. Harris says the data quickly proved "too much data to ingest at once," raising the question of what it means. With powerful models now out in front of the evidence, clinicians must ask whether outputs are true and impactful for the specific patient, because "the first thing we want to do is not do harm."
[00:15:25] How wearables can empower patients, if the health system relationship comes with them — people are "really, really interested in their own health" and want to own that process with a partner who helps them understand it. Wearables can show things like heart rate variability, but without continuity of relationship between the patient and the health system, imperfect tools get bought and used without translating into anything. The question, he says, is "How do we use technology to maintain and enhance that relationship that our patients wanna have with us?"
[00:19:55] Why UI Health held off on collecting SDOH data through MyChart — the team wasn't ready to handle SDOH data through the portal and already had existing workflows strong enough to collect what it needed. The decision came down to "what's the best way to serve our patients versus MyChart can do this now." Unlike downloading a beta update on a personal phone, healthcare has to be more measured because "if things go wrong, people can get hurt."
[00:21:23] What the "apology factor" means for trusting AI with patients — Dr. Harris says how much to trust AI with a task, especially patient communication or scheduling, is something he struggles with constantly; at minimum, a tool must not be worse than a human. If a call center agent gives the wrong clinic hours and apologizes, patients forgive it, but "if an AI does it and it's wrong, and there's no one to apologize," it raises the question of what that does to the relationship.
[00:22:36] Why UI Health spent over a year evaluating an early patient-facing AI tool, and what it learned — "the early days of AI in healthcare were not great," and after months of work the team concluded an early text-generation tool for patients "is not ready for primetime." Vendors get better tools every three months, but "people are always gonna be people, and the speed of organizational change is a constant." New tools must move the needle measurably and preserve relationships with clinicians and employees as well as patients.
[00:24:56] The reason patients are moving faster than health systems — citing literature that 90–95% of what clinicians do doesn't make the biggest impact on overall health and life expectancy, Dr. Harris notes most of what makes someone healthy or unhealthy happens outside the clinical setting. Patients have long brought direct-to-consumer lab results and DNA risk tests to him asking what they mean, putting them ahead of the health system's curve in ways other industries are more insulated from.
[00:27:25] What UI Health measured in its ambient AI pilot beyond whether clinicians liked it — "feeling better about something doesn't actually mean that it's better," so the team defined workflow metrics such as time patients spent in the waiting room and time spent with clinicians. The pilot spanned over a dozen specialties to reflect a representative group of clinicians, rather than proving value in one clinic, and tested both current scalability and expected improvement. The team also assessed what the technology would mean for its balance sheet, because "technology generally is not free."
[00:30:23] How ambient AI is moving from shiny object to "just what we do" — with accelerated AI development cycles, waiting two years to implement something is "a very long time," but systems also can't be "blinded by some really cool shiny object." Ambient "feels magical," and while the novelty is wearing off as many vendors have gotten good at it, Dr. Harris predicts that five years from now "we're not gonna be having a conversation about ambient," much as no one talks about the mouse as a new tool, and that audio interfaces may change how clinicians interface with EHRs.
[00:32:46] Why voice won't replace the physical interface for clinicians — "our medical decision-making is not verbalized most of the time," and much of what clinicians process and rule out is never part of any discussion, which makes it difficult to communicate to a system. On alternatives like haptic technology, Dr. Harris says the hard part is choosing a winner, comparing it to VHS and Betamax or HD DVD and Blu-ray: solutions will come, but adoption and operationalization are harder.
[00:36:11] How a safety net budget forces teamwork — at a safety net academic system, "I just can't do anything on my own, which I think is a good thing." Limited resources push groups across care delivery, administration, and research to talk, avoid unnecessary overlap, and pool resources, expertise, and goals, especially with the financial headwinds he expects in the coming months.
[00:38:44] Why constraints make UI Health more innovative — the team can't do everything that sounds cool, or even everything that might benefit it, so it focuses on what's high value for the mission and patients. The constraint is often people, not dollars: enough trainers, analysts, and operational leaders, and whether attention can be diverted from current problems. Constraints push the team to solve problems in ways it might not have chosen at first, making it more nimble.
[00:39:45] How to say no to good ideas the right way — Dr. Harris says he didn't expect to say no as much as he does in this role, and part of his job is saying no to really good ideas well. His approach: offer to spend time understanding what the person is trying to solve and provide an alternative, asking "This is a great idea. How can we support you even though we don't have a dozen people to throw at it and $2 million to buy this thing that you wanna use?"
[00:41:42] How the CMIO role has shifted from "evangelizer in chief" to strategist — the role varies by organization and reporting line, whether to a chief medical officer, CEO, CTO, or CIO. It used to be the doctor who liked computers and helped turn tools on; now many in the CMIO role help institutional leadership strategize, including structuring the organization to take advantage of current or emerging technology and developing a roadmap that fits the institution's goals.
[00:43:58] The four mindsets a CMIO needs, and why the list is growing — informatics leaders must be a clinician, speak technology, communicate well with other clinicians, and understand hospital administration. Those slices are "only gonna be growing," Dr. Harris says, as AI raises new questions beyond EHR implementation: how to train the workforce on AI, what the data and people infrastructure look like, and whether the right resources are at hand.
Fast 5 Lightning Round:
1. What is your favorite book or book you've gifted the most?
How to Talk So Kids Will Listen & Listen So Kids Will Talk by Adele Faber and Elaine Mazlish
2. If you could instantly master any skill, what would it be?
"Going to sleep exactly when I want to"
3. Would you rather have super strength, super speed, or the ability to read people's minds?
"Super speed"
4. What is something in healthcare you believe that others might find insane?
"Fax machines"
5. What is the last movie or TV show you saw, and what did you think of it?
F1. "Equal parts completely unrealistic, but still really enjoyable."
The Digital Patient has been recognized as Feedspot's #1 Patient Engagement Podcast of 2025. Thank you to our listeners for making this happen!