Please welcome back to the stage, Mike Sicilia.
Welcome back from the break. I'm especially looking forward to this next session because we're talking about what it takes to keep pace and to move AI from promise to real impact. And at Oracle Health, as you've heard in the last couple of days, we see a real opportunity to connect clinical and financial intelligence in entirely new ways. And today, I'm excited to share quite an important next step with all of you.
We're bringing AI agents into the full revenue cycle, from clinical documentation through reimbursement and into enterprise finance. Seema also already mentioned the financial impact of our clinical AI agent that we've seen in early adoption. And building on that note-taking capability, we're extending AI across coverage determination, prior authorization, documentation integrity, coding, and even appeals. So really connecting that clinical and financial context across the entirety of revenue cycle management. And these agents are just a few of the tools available to help you turn what we think is AI ambition into scaled impact. Today, we've got three extraordinary healthcare leaders with us today, and very many of them are doing just that. It's really their innovative work that pushes us to create new solutions like the agents that I just described. So please help me in welcoming our distinguished panelists.
Please welcome to the stage His Excellency Dr. Majid Al-Fayad, Dr. Brendan Carr, and Dr. Pratiba Varky.
Okay. Well, again, thank you all. Thank you all for your leadership, and thanks for being with us here today. So we've heard a lot about and we've seen no shortage of AI-based pilots. The challenge, I think, really for the industry and industry in general has been: how do you turn them into capabilities that are going to be used and scaled every day across a health system? So I guess as we go down the line, you know, a question for all of you is: what is it that separates the AI experimentation phase from something that you're willing and confident that you can clinically deploy at scale? So Dr. Majid, we'll start with you. What is it that you think is the distinguisher?
Yeah. Thank you for having me. It's a great honor to be here. In healthcare, we're bombarded by this. As a physician, I'm talking, we're bombarded by this technology, and we have to be ready. And I say that because we get ideas, and we are inviting our healthcare workers to come up with problems that AI can solve. And we're getting all these problems to solve. We have the technology that can answer that, and that creates a lot of demand. But we have to have gates, and we have to make sure that what we engage in can actually translate into value. So it has to be aligned with our strategy. It has to be visible. It has to be useful and create value. And then monitoring later on, we have to demonstrate effect. And I think that creates the difference between ideas and actual product that can help the patients. This August, for example, our team got about 38 ideas that were filtered through the gates of AI products that they want to work on, between something that we acquire and something that we build. However, when we put them through these gates and make sure that they will produce value, so it's not only a good product, but it's also a useful product. It's feasible. It's technically doable, and it can stand the test of our governance, which we have set up as well separately. Then we go for it, and I think we have been successful in many areas that we sort of introduced to our healthcare workers and our professionals, both in the hospital as well as in the research center, but also in the non-medical areas where it helps us with operations. And the examples I could sit here and tell you about examples that we did this in radiology, we did this in breast cancer, and we did this in... but I think the examples are plenty, and hospitals are racing to make best use of these AI products in order to capitalize and not only create efficiency and value, but also answer to the demand of the society and the new generation that are seeing how easy it is to deal with the government or banks or any other facility or aspect of life. Yet they come to the hospital and they're faced with these long lines, expensive care, and inefficiency. So we not only have to answer for our operations and our boards and our healthcare workers, but also we have to answer to society.
Yeah, I think you made a great point around governance, you know, governance and regulatory compliance and these things. And you know, when you have a technology that's moving so quickly, how do you make sure that, you know, we're keeping the regulators and the oversight agencies involved here? You know, very good point. So yeah, we'll just move down the line. Dr. Carr, experiment, scale, deployed in the middle of New York City. What's the tipping point?
I'll pick up on a couple of themes here. One is listening to the people and so solving for the pain points. And you know, the partners that we work with, the companies that we work with know this as well. It is the onslaught of firms that want to, you know, want a laboratory in which to experiment with things makes it really hard. So I guess I would say for us, listening to the two biggest stakeholders: our employees, you know, the care providers, and also the patients. If it's a pain point for either of them, it gets bumped up on the priority list for us. In parenthesis, you know, being honest, also if there's a big financial return, we need to prioritize that too. So, you know, those three things come through a funnel. And I, you know, I sort of... and I said early days to... I said to someone who was pitching me over and over, 'Look, I need you to come through the front door.' And they said, 'Oh, we hate front doors. It's so much better for us, and we can just come in the window on the side, grab one of your doctors, let us do a pilot within their practice, and then somebody signs an MOU or a contract, and you know, now we're in the organization.' And then it gets... that's true, that it makes good sense, but it does get in the way of us actually intentionally deciding what to scale. So solving for the key pain points, being intentional, use the word governance, being intentional about where you're going and tethering it to your strategy. And then, you know, another big piece I would say is communicating it, because if you're explaining to people what you're bringing in, what the timeline is for pilot, how you're going to... we think of ourselves as a learning health system, if you're clear about when you're going to either kill it or scale it, then people can adjust. You can't pilot everything all the time. You can't, you know, pilot and never scale. And you have to communicate. I guess those are some, you know, top-of-mind thoughts for me.
Yeah, you mentioned the word listening a lot, and I think it's worth saying that, especially for us as a technology provider. You know, we've been in lots of businesses providing mission-critical operational technical things to, you know, banks, utilities, retailers, and so forth. And we go through a lot of these big technical... and I'm not saying that they don't listen to their constituents, right? But what you know, as I think about this, but certainly, you know, in healthcare, one of the things that I've observed here for the last five years or so is just really the importance of getting the provider buy-in, the importance of getting the people that are actually delivering the care to not feel like it's being pushed on them, so to speak. I feel like they have a seat at the table. You don't want to have that 'hey, we're here from HQ, we're here to help' mentality and say here's the best system in the world. So I think that, you know, we've kind of learned that lesson, to be frank, as we go through there. So it's a great point. You know, Dr. Varky, I mean, obviously the Mayo Clinic has been no stranger to cutting-edge innovative technology for decades. This AI thing, relatively new, you know, it certainly has scale. And you know, what's been the tipping point, if any, yet from experiment to scale for you?
Yeah, I think I'll touch on something you raised, which I think AI at scale is really not about technology success. It's really about trust. It's about redesigning the underlying processes and getting the stakeholders to buy in and actually carry out the change. So, you know, the questions that we ask are: is this really an important issue that needs to be solved that addresses either a patient concern or, you know, an issue that's affecting a care team member? Second is: has it actually been validated in the clinical environment where it's going to be implemented versus just in a test environment? And has it answered the typical questions of, you know, quality, safety, affordability, you know, financial sustainability, efficiency, etc.? And then is it embedded within the workflow, or does somebody have to go finding the AI algorithm to execute it? And then finally, the trust issue. It's really about: do we know how the algorithm or the agent came up with the recommendation? Is it traceable to the source of information? Is it auditable? And are there governance processes so that these systems can be monitored on an ongoing basis and stopped and pulled if we have unpredictable outcomes or if we have concerns about safety?
And have you seen any tangible impact on the provider-patient interaction yet from AI? Have you seen any difference in how doctors, nurses, other caregivers focus really, you know, less on the tech, if you will, and more on the patient interaction?
Yeah, it's been an incredible time. We have over 300 AI algorithms embedded in the practice currently. I think we are just at the beginning. We're learning a lot, and I'd say there's been some tremendous opportunities that we've observed where it returns time back to our talented, hardworking staff, and more importantly, helps us singularly focus on what matters most to us, which is the needs of the patient come first, which is our primary value at Mayo Clinic. You know, early observations suggest four themes that I'll share in maybe a few quick examples. One is early detection and diagnosis earlier than perhaps clinical symptoms show up or traditional imaging might show. For example, AI-enabled EKG analytics that's predicting CHF for primary care physicians and alerting them about more imaging that perhaps needs to be done if the patient is asymptomatic, or it is in early lung cancer screening for entire populations versus waiting for the patient to come to us. So that's automated processes. The second is really in triage. Should this patient be admitted to a smaller critical access hospital or a hub hospital because of the acuity? Perhaps it's a staffing issue and various things. We are able to run AI algorithms on the entire group of elderly patients that are at risk for readmissions, for example, and overnight pull out the 10 people that are at risk and then send nurses home or have a telehealth connection so that we're able to reduce ED and hospital readmissions. So it's amazing the kind of changes that have occurred. Workforce transformation: we had issues with CRNA shortages during COVID, post-COVID, and you know, there were large numbers of patients waiting for colonoscopy. And marrying AI algorithms to evidence-based algorithms, we were able to triage down, you know, complex anesthesia to conscious sedation that nurses could administer, and 30% increase in the number of colonoscopies, you know, 150 colonoscopies, for example, in just one town, and that we were able to disseminate and scale. So, quite a lot of excitement in this space. Of course, ambient listening that many institutions have implemented, but 50% reduction in chart reviews for complex specialties, you know, 260 such summaries that have been embedded into the EMR. So, I think we're just at the beginning stages. Lots more to be learned, lots more to be done.
Yeah. Well, it's great to hear some early success stories. You know, I think the world needs to continue to hear those, and certainly there's no better place, you know, except for healthcare, to hear positive impact from the technology. You know, with that said, barriers remain. There are lots of opinions, lots of opinions. I say, you know, Dr. Carr, what have you found to be the biggest barrier so far in, you know, thinking about your clinical and administrative operations and, you know, getting people to embrace or think about rolling out AI? And what role do you think technology providers have, or can technology providers help overcome barriers?
So I mean, I'll piggyback a little bit. The frameworks are really important for me, and I love how your brain is organizing those things. I mean, I think in back-office administrative work, I think operational efficiency, and then clinical, the barriers are a little bit different in each of the three. Let's pick, you know, the most fun one and the one where we've made the least progress and where the barriers are the biggest: it's in the clinical space, right? So similarly, we have a bunch of algorithms that we rely on, and all you need is a couple wins, and those couple wins do start to build the trust and the culture that you're talking about. If you solve for something and demonstrate out here, a couple concrete ones are predictive analytics on whose wound is not going to heal or is going to break down if you send them home based on their nutritional status before and during their hospital stay. Or the fact that every, you know, twice a day our critical care and rapid response team gets served up people that get flagged for perhaps decompensating in the next, you know, 6 to 12 hours. You know, and then now you just have proactive critical care folks walking around, taking a look at the folks, talking to the nurses, talking to the primary teams, potentially, you know, transferring folks. It's sort of the light-touch stuff that helps, that gives you the wins that then gets you beyond the piece that you're talking about here, the barriers. And so if you get the trust right, if you're solving for the right things, if you get a couple dedicated demonstrations that really work, and then you're clear about articulating scaling them, that's really where the tech partners come in. Because the other barrier is that my Instagram feed and probably yours is full of all kinds of narrative around AI. And when it's this overwhelming thing, and all, you know, and the key frontier labs are talking about how it's going to solve all the problems or end the world, one or the other, you know, it makes it really complicated for people to accept it as a part of their clinical workflow. When you make it much simpler and you don't brand it as AI, but rather talk about the fact that we're using the information that is available inside of our electronic medical record to identify patients who we're worried might decompensate, and here's a list of, you know, some folks that you should go see before you head home tonight. Boy, that's, you know, that's a big win. So it's a weird thing to say in response to your question, but the more invisible our tech partners are, the better off we are.
Yeah, I think we certainly share that vision. The more the technology disappears into the background, the better. We had this experience when we first rolled out last year the clinical AI agent, which I'm going to trivialize and call it an ambient scribe. There's a lot more to it, but there was a lot of resistance, you know, in the first systems that we piloted with, a lot of resistance among the providers until they actually started to use it. We said, 'Just give it a chance, see what you think.' The uptake right now, it's certainly the fastest uptake that we've had of any clinical technology that we've ever rolled out. And the reason is you don't have to get trained to use it. It disappears in the background. And I'm not a provider, but what I hear from providers, it really does allow them to practice at the top of their license because the technology is not in the way. It's in the background. So that's, you know, it's a really... and it's an interesting blend. We spend so much time in the technology industry worrying about and thinking about user interfaces and patterns and how do you get people... and actually the best thing that we can do is just to get rid of it all, right? And that's really been something for us is we've focused on. So I appreciate your validation on that. We'll try to disappear as much as we can into the background because I think that's probably the ultimate deal.
Lean in until we get it right, and then, you know...
Yeah. We'll surface when you need us, but we'll try to be in the background. But you know, I think, you know, back to barriers, back to benefits now. So obviously, hospital capacity is always at a premium, right? You know, whether it's available beds, available staff, all these things. Dr. Majid, are you thinking about the opportunities to rethink care models and technology such as remote monitoring, personalized medicine? How might it help, you know, a large organization like yourself, a scaled organization, countrywide provider, think about potentially reserving hospital capacity for those who need it most because you've got potential telemetry that you don't have in traditional technology?
