Ben's Den LIVE! at Fuse 2026 | AI in Healthcare: Beneath the Frothy Surface

Speaker A: Foreign.

Speaker B: My name is Isaiah Nathaniel. I have three different roles which I'll take a little bit to explain. One is at a local community health center called Delaware Valley Community Health. That's in Philadelphia, Pennsylvania. Yes. I'm so happy we beat Boston in Boston. Yeah, that's like Mamba mentality. Can't wait for tonight. So I've been there, senior vice president and chief information officer there for 18 going on 19 years. I am that old. Then I work for my state primary care association as the strategic CIO which oversees about 54 independent community health centers, about 1.2 million patients. And then in January, because I wasn't tired enough, I started working for the national association of Community Health center which oversees about 1512 community health centers in the entire country. We're in every location outside of three congestional districts, about 17 plus thousand locations, about 326 FTEs across the country. One in seven in primary care, one in three in rural America. So this really hits home. This population health is value based care. And I'm excited to be with my panelists today. Pass it on.

Speaker C: Thanks Isaiah. Hi, I'm Pete defandi. I uh, moved around a bit in health care. I just uh, basically ended a path with uh, a local payer in upstate New York called cdphp. We did a merger with uh, one of the blues. And it led me on this really interesting journey where I spent the summer just uh, I call it an AI sabbatical and uh, then ended up at a consulting uh, company right now where we focus primarily with New York state. We do a lot of work with Department of Health and Medicaid data.

Speaker D: Good morning. Um, my name is Jody Nelson. I'm the CEO of a small critical access hospital in the state of North Dakota, St. Luke's I was born and raised in my community so I'm very passionate about keeping local healthcare local and keeping independence strong. I'm also the board chair of the Rough Rider High Value Network in the state of North Dakota. We started that with the help of the Cebolo team, Nate White and Brittany. And so I'm excited to be here.

Speaker E: Thank you. Good morning. My name is Miura Kinhawa. I'm an emergency physician, practiced in the military and civilian for almost 20 years. I've been a CMIO, I've been an ED trauma center chairman. And somewhere during that time we started building software because we had no data on our patients. And in an ED, when you're seeing 100 to 150 patients a day with an eight hour waiting room you don't have time to look at paper and you're not going to look at it. So when I left clinical medicine, I went to Microsoft, helped, uh, intermediate this acquisition and was part of a really exciting $2 billion startup within Microsoft called the Health Solutions Group, where we started imagining what healthcare would look like. And that was my transition from clinical medicine to sales leader to technologist. And I work with a large, uh, pharma, medtech, biotech companies, helping them understand the roadmap to AI. Anyway, thank you for letting me be.

Speaker F: Thank you. Thank you.

Speaker G: So Ed.

Speaker F: Ah, let's dig into this. So what's our topic for the day? What are we engaging these top tens on?

Speaker G: Yeah, so we're going to try to be super engaging with you and interactive. So we are going to kick off a few topics, but we'll stop a couple times if you have any questions. You have access to this panel. Feel free to jump in. So we're going to talk about, you know, what's, what's reality versus hype. So what are one or two ways that you sort of work your way through and filter. Filter the noise. So we'll start with uh, Pete.

Speaker C: Sure. So it's interesting you say filter the noise. Our world is filled with noise, right? So there's noise in terms of uh, tenants that are in the organization that are just reasons people think they can't move forward. There's noise in uh, preconceived notions and just biased of like, you know, what they think the outcome is. One of my favorite things in my career has been when someone tells me that can't be done. And I'm instantly attracted to that and I'm like, okay, now I have to do it. So, So I think one of the ways is, is to constantly challenge that, you know, that edge of well, what. Where is the boundary? I try to pretend there isn't one, but it also is imbalancing. Is it worth it to go after that? Uh, I think in general though, just, just building an organization that constantly challenges, you know, those boundaries, those tenants and pushing forward. And honestly, especially in healthcare, it is just we, we actually Ed and I talked about how like unfortunately how just tough it is to like be just constantly, uh, especially in, as a CIO in healthcare, like just really restricted by these, uh, different policies and governance and just constantly, you could be just absolutely buried and never move the ball forward. Certainly, um, Pranam and I have seen it as we work together, but I think that's just, just it in terms, especially in delivery you just have to constantly keep pushing and challenging. And I don't believe that there's any boundaries. That's why I do it.

Speaker F: And when Pete and I, we, we partnered, we had 169 practices, Pete, that we clinically integrated. And that was, yeah, so interesting.

Speaker C: The garage helped us to bring these small, these small providers, really. And they, while they did participate with our hie, the data was really messy. Uh, with, with the garage, we were able to unwind that data, become more standard. Really focusing on HEDIS measures, to be honest with you, is our biggest thing. But building hedis supplemental, uh, feeds, working with, with, uh, you know, these, these are, these are payers, these are providers, I'm sorry, that can't afford epic. These are providers that are on these little mom and pop ehrs all over the place that you want to want messy data. And so that was a good example where we pushed through, uh, and got through boundaries.

Speaker G: All right, Jody, what about you? How do you sort of filter the noise with reality a little bit?

Speaker D: You know what he was talking about. But I think for me, we are so small. Critical access hospitals are so tiny. It's a little bit different. Our scale is probably much smaller than other big facilities. Critical access hospitals serve one of every five Americans in the United States. And so although we are small, our challenges are different. Financially, I think we're challenged. And that is one of the great things about joining the Rough Rider high value network and these cins. It's because we are able to come together and maybe have more noise because I think we are so small. Our platform and our power has become greater because of that. The networks, having the access to garage, getting the data when our teams are so small and our financial capabilities are so small, our bench has gotten bigger and we're able to do more things because of that.

Speaker F: And something that I learned when I started working with cebolo, you know, we all know that hospitals are major employees in pretty much every state in the country. But what I learned about rural health and critical access hospitals is that the local economy is driven by these organizations. And you say it's small, but it's so important for the economy. Jodi. But, and to think about that, if you, if critical hospitals are at risk, you are decimating the economy of that local town and local community, which will lead to like, lot more issues and whatnot. And that was just, to me, to understand that from a data standpoint was just mind blowing.

Speaker D: Yeah, it is. I mean, most critical access hospitals are the bigger, uh, biggest employer community, which that's what we are. I think the other data thing for critical access hospitals is we are 90% of the geography of the United States. So because the nearest hospital to us is 50 miles away, um, and most critical access hospitals have to have that standard anyways. But we are, we are the backbone of our communities and if we lose what we have, then our communities probably die too.

Speaker G: Yeah. Long live critical access hospitals. Mira, let's talk about AI and value based care. What have you seen out there? Or how can it help scale?

Speaker E: I think in value based care though, continuing to push the envelope around predictive, uh, morbidity and mortality, not just ED visits and hospital visits are going to be what will save money in the long run. The predictive nature of that has to go back to some historical data on that patient as an individual. Right? So you can point to that patient and say here is when they started developing diabetes, or here's when the cardiac issues started, or here is when we knew they were going to have pancreatic cancer. And now as you fast forward that, you can actually help that hospital, the pcp, their care team, the payer, understand what that patient's future is going to look like.

Speaker G: All right, all right, uh, Isaiah, anyway, any thoughts around, uh, kind of going to the platform now? You know, as a longtime CIO and with the excellent primer on sort of value based care and AI, what are your thoughts around leveraging a platform type solution?

Speaker B: Well, 1 think there's a couple similarities about what we're trying to discuss here. I want to use a Mamba mentality, Kobe Bryant. I just have to hashtag whatever it takes. Right. So the way I like to describe the system that I am as community health centers where we see 54 million patients across the country is the triangle offense and the triangle offense for those that aren't basketball players. This was a, uh, offense that was designed by Tex Winters, uh, implemented by Phil Jackson, first started with the Chicago Bulls, then went over to the LA Lakers. And the precipice of this concept of the triangle offense was everywhere on the floor there was a triangle. Uh, and the ball predicted what would happen next, not the player. And so when this concept, the ball is the data, that's the patient, and the triangle is primary care, which I represent to the subspecialty to the hospital system and then back and I'm um, making it very lame. And there's more in that triangle. But the sense is that data has the transition. Now for us, it has to be vendor agnostic. So in our ecosystem of community health. We have Epic, we have NextGen, ECW, Athena, Meditech, we have it all. But irregardless of that, what we have is 54 years of data, social determinants of health data, population health data. And that's happening at the point of care. I can tell you we know every part of the family because we've been seeing these families for years. We know the mother, we know the father, we know the brother, we know the cat and the dog. We know what they ate last night. Uh, because they tell us, and we ask this at the point of care, at the point of the clinician entering the room. So for us, value based care is understanding all of that and being able to transition it to the hie, transition it to the hospital. But then what? Get that CCDA or that transition of care document back. And this is where, going back to the signal versus noise, how do I as a cio, present that to the clinician with less noise in a 15 minute visit? So for us, the concept, the platform is the new UI is no UI at all. It's transitioning that data across this triangle offense that guess what for Michael Jordan won six championships, for Kobe won five. And you know, I like to be a, uh, winner in community health.

Speaker G: I, I love that quote. Uh, the new UI is no UI at all. That's very well stated.

Speaker B: For me, I'm going to get on a sermon. So this is the point where I always tell when I'm, I'm, I'm in these sessions, specific specifically with that, that I come from a long line of Baptist preachers. And you see, I start to lean back because this is when a sermon comes. Yeah, uh, you should stand up. This is when it happens. So get the plates ready. You know, I'm a, I'm a father of three kids, so Uber and Lyft is free to them. So I need money for gas money. Here we go. For me, it's data rationalization. And when I say data rationalization, part of my frustration is that we don't have as many required fields in the data that we should. And so this creates this free text environment that it's hard for these agentic AIs for generative, agentic, whatever maturity model you're on, to pull that information and then give the proper suggestive, both unacceptable or acceptable to the provider to say with a human centric opportunity to say, yes, this is the right, no, this is wrong, go back and train again. So for me, if I use this $50 billion in a world transformation health Fund it would be a data rationalization, it would be making sure HIEs are part of the conversation. Because then you put the security risk layer on top of that. And yes, you can push it on the vendor, you can push it on the EHRs, all of these things. However, that data, and my analogy is traveling with the patient, traveling on your watch, traveling in your pocket. That has to be secure in gestation as well. So for me it's data rationalization, making sure that layer is as clean as possible, making sure that if we put it into the SVAP process, standard version advancement process by the onc, we have more required fields, we're getting into, um, all of those strategies so that when you're pulling data, you know, you can match it to a model that can understand it. And then at the point of care, which is my true North Star, the doctor in that 15 minute visit is able to produce an outcome that we want for hope person care.

Speaker H: We have models from the payers, we have models from cms about how you do provide a patient attribution to attach the patient to the provider, but different EHRs and others have their own. So what do you do when you have multiple different attribution models and you want to rationalize data, so multiple ways

Speaker E: to get at that. Okay, so one of the ways through data, they're putting, um, they will be putting uh, agents at, you know, pick one of the 400 hies in the country. So you will reverse engineer within that. What, what physician ordered what? That's one of the ways that they're going to do that. The other way they're going to do that is by getting claims data. Right. They can start with all Medicare, Medicaid and then reverse engineer that. Right. But the agents and agentic families will do on the top of that data is then look for centers of gravity, Dr. Smith, with this patient so many times in this zip code. Right? And then attribution will be algorithmically derived based on what the data shows. Remember, we are all economics here in healthcare, right? If you ignore the economics and just give grants, the great idea ends up not being adopted across the entire nation's spectrum of backbone of delivery.

Speaker C: Can I add one thing to that real quick? So the question just stepping back a second ago was where would you start? And so, uh, from the. I would follow the dollars personally. From the payer side, I can tell you where the whole country's starting. It's utilization management and it's in your top categories. That's where all the pressure is going to come. If I was on the other side of that. That's where I would start from, like a counter, almost, uh, intelligence perspective. I'd be looking at muscular shell and I'd be looking at heart and I would be figuring out is it medically necessary? Because right now all of AI back to value based care, where is it being used? It's going to be used for utilization management because the entire payer industry is going bankrupt. So if I was on the other side of it, I would be focused there trying to figure it out and then proactively trying to address what you're going to be managed anyway through what's called value based care.

Speaker G: All right, we're going to transition now to principles, uh, like AI principles. And Pete, I'm going to stick with you 60, uh, seconds or less, uh, some AI principles that you leveraged, uh, throughout your career so far and would encourage other people.

Speaker C: So the number one principle is start immediately and get into it and use it every single day and, and just start experimenting and finding the boundaries of what is working and what is not. Not everything you put in there is, is perfect. And so you need to, uh, you need to just start playing and adopting it. I just set my wife up with several accounts and I was like, finally clicked because it, once it clicks, but then it goes to that next level, right? The first is like, hey, all right, this is, this is working. This is pretty cool. The next one is like, wow, there's real value in here. And then like the last level is like, wow, I am now 30 to 50% more efficient than I ever been, I ever been my life. It's just crazy. So just honestly get to work.

Speaker G: Jody, you're a leader of an organization and I know I've met many others in the audience that are leaders of organizations as well. How do you think about AI and sort of the adoption of any sort of principles or guiding things that help guide you in your decisions?

Speaker D: Well, I think for us again, as a critical access hospital, AI has been huge. For us, the last, you know, couple years is really actually only in the last couple years because of, once again the financial restraints of it. But you know, from the provider standpoint of using it in the room with the patient, and it goes all the way to the end when the claim comes in, going back a little bit, Nate brought it up and I've talked about it too, with the Rural Health Transformation Funds is going to be, it is already life changing for critical access hospitals because we have this opportunity to possibly get platforms that we've never been able to do before. And we have that because of Rural Health Transformation Funds, but also because of the Rough Rider High Value network coming together too. So that first was our opportunity to get the garage and do things as a group. But we just this last week in the state of North Dakota, the Critical Access hospitals had the first round of funding. It was $10 million and it was only open to the Critical Access hospital. So there's 37 in the state of North Dakota. So we submitted our first application to the state North Dakota or uh, HHS. And so that was $10 million, but we got to spend 199 million. Most states, I think there's a couple that are a little bit over 200 million. Um, they have to give it out before September 30th. I mean it's crazy amount of money that's going to be coming into each state in the nation because of this. Now I know one big beautiful bill did some other things that were took away funding to the states. But this could be for the next five years could be transformational for rural um, healthcare and any, I mean it's supposed to go to rural and the state of North Dakota has defined it now a little bit. But it will be interesting. Our AI capabilities are going to be changing.

Speaker G: That's great. We're going to shift into what's working well and what isn't working so well. So after we get through that section, we'll open it back up for questions for our panel again. So I'll start with you, P1 Isaiah sitting next to me and tell us, uh, one example of AI and if can related to value based care that's worked pretty well. And, and one time that maybe it didn't work so well, maybe some lessons learned.

Speaker B: One time that worked well and one time it didn't. The one time that I would say that it worked well is ambient. And the way that I would say this, and just from the perspective of this, we are on 10%, 20% journey of adoption and ambient AI. But what we have found where it works well is that it's not only save time but it's an increase the quality of visit. You're probably saying that it doesn't make sense. I'm going to explain why the data and what we have seen is that in the point of care in the exam room, providers are actually spending more time with the patient than less time with the patient with Amy and AI. But what they're not doing is spending more time charting at home. So this dichotomy of A physician who's been trained to go to school to put their front person in front of the patient. That's what's actually getting back to with ambient AI versus them sitting there charting the entire time trying to make sure they click every button for meaningful use and all of the things that they have to do for PCMH or whatever. So these are ways that we have found and we're calling it table stakes for AI and community health ambient AI. One way where it hasn't worked well is in some of These integrations with HIEs, integrations with CARE, coordination with claims because of the attribution not being so well defined and understanding really where that journey is and then that is creating this friction going from fee for service all the way to full risk. You have some that are integrating with clinically integrated networks where we're getting like together and that is trying to move the maturity model. But I think ultimately with the lack of information from the payers and this attribution, it's causing this slower adoption in the value based care.

Speaker G: Mayor, what about you? I know you get a lot of views of different organizations. What's one thing that you've seen work well and one thing not so well?

Speaker E: I'll uh, start with what doesn't work and that's doing POC's proof of concepts. 90% of them end up on the shelf. Don't do them. Don't waste your time, don't waste your money, don't waste your resources. Just don't do them. And the reason is because if it's not going to prod, why are you doing it? So if there's no vision to turn this into a production system, um, don't waste the time, money, energy, resources, people, because you don't. So you know, I think the things that are working are when organizations understand that given the economics of where we're living in looking at cost takeout and cost avoidance with AI is good but insufficient.

Speaker D: Right.

Speaker E: It's going to help your bottom line. What AI should really be helping you do. And where I've seen true benefit to organizations is when they look at top line.

Speaker G: Yeah Jodi, uh, you as well. Give us one each and then uh, we'll stop and take questions from the audience on this particular topic. So what's worked well like an example for yourself and maybe you had something that didn't work so well.

Speaker D: Okay. For working well I think for us it's the AI scribe. For the providers it's been work life balance for them is so huge. Uh, being Able to do the scribe and not having to take home that work at the end of the night and also have it flow all the way from the provider visit all the way to the billing and make sure we're capturing all the charges and getting paid and um, closing those care gaps. So that's been great for us. Not so great. I don't have a lot of negative for it yet. And I think that's the exciting thing. I think there. I think for us and personally it's all been great, to be honest, even for our rural health transformation. We use ChatGPT and Copilot to help submit the grant. I mean and uh, we would never have been able to do that four years ago. So our time was, was so for our team was so great because of that.

Speaker G: Yeah.

Speaker F: The one thing that we, and probably everyone in the panel will relate to, you know how in older systems, our transaction system is a garbage in, garbage out. That was something that we lived by. So that's how you kind of like sanitize the data and whatnot. When we started running the models on that whole fabric, the garbage in ended up becoming more believable garbage out because it was so intelligent in interpreting the data that that hallucination was very hard to detect. So we had to put more controls around it to ensure that that garbage in, garbage out is taken with lot more seriousness. Because now you're going to start believing that garbage out and that is happening in real world, that's happening in healthcare. So that's something that we put in a lot of work to put some controls.

Speaker G: All right, we're going to take another pause here for some audience interaction. You have great panelists up here talking about what's worked well and perhaps some um, lessons learned from what didn't work well. Any questions?

Speaker H: Hello?

Speaker I: So when I'm not following Sabre, I actually do work as a, ah, AI prompt engineer and do some other techie stuff with AI. So my question is when it comes to accessibility and AI, and you're talking about guardrails and things like that, how do y' all manage that expectation to keep it human centric accessibility with like communication and you know, I don't know because you're going to have people from different platforms and backgrounds dealing with it, I think you spoke to kind of what that looks like on a scale, but what does that look like for you all kind of staying human centric with AI accessibility.

Speaker E: So, um, I work with multiple models. I'm usually using one of the tools somewhere around six hours a day. It Got to the point with one particular tool where he asked me if I would like to call him a name. He learned my behavior to such a degree and then he came up with a name for himself which is Percy. So now I say good morning Percy. And then he explained to me how he came up with that name for me based on what he knew about me as Mira. So when you talk about human centric AI, I haven't keyboarded a thing thing in months. I just talk to them on my phone, on my laptop, through my AirPods, anywhere I can. It's conversational, there's no prompting.

Speaker D: I think looking at it from the uh, patient perspective that is very important and but I do think in the way that we can do it in healthcare with that uh, face to face visits is making sure those patients know that that AI is giving that patient more of the provider time with them because they're being able to focus on them and not have to do those documentation pieces and worrying about typing everything in. So that's I think how we can make sure health care is. You're still human. Human, yeah.

Speaker C: So we confronted a lot of this on the payer side and again this, some of this is going to sound a little cliche but the bottom line is that you know, we, at least what I've seen, our strategy has always been a human in the loop or a human at the helm model which I'm sure you're really familiar with. But the problem is that goes too far and sometimes it becomes a human rubber stamp. So the key is balancing actually where the human is the judgment, the AI is the efficiency and the automation. Now I get that this may not be the end of that road and of course really to push healthcare to the end. And I won't even get into the concept of mental health through AI because people are already starting to use that as a therapist and there could be some real benefit there. But I do think that's the balance that we all walk. And I'm sure you know, I'm not on the, on the provider side per se, I was more a partner with the provider. But that's at least I think where most, I guess most humanity like to keep that balance between humanity and machine. At this point

Speaker B: what I'll add to the conversation is in that human centric, it's also the human design of it. And when I say human design I'm talking about tech wity a concept of techity where tech equity is for everyone. And even in artificial intelligence oftentimes we find that the LLMs are not trained for the patient populations that it's predicting an outcome for. So a lot of my sermon on my sermonette is making sure that if we do do a POC that you are opening up your LLMs to ingest the data that we have. We're not just going to give it to you because we're going to make your product better, but we're going to partner with you to allow for that LLM to have the proper set of data. North Dakota is different from Philadelphia. Let's train on this Philadelphia model because I'm predicting for social determinants of health, where there's septa, there's buses, there's homelessness, there's food deserts and all of these things. So making sure that the human centric model is integrating itself into the data so that the outcome is predictable in a better way, and then you allow for it to continue to be retrained by six to six.

Speaker F: I said something you said piqued my interest. You said the model from North Dakota will be different than the model to Pennsylvania. Right. Do you see a future where you have very specific models by region or by org type, like an ACO versus an FQ versus. Do you see that happening where the models become very specific to your world?

Speaker B: I would say region based in some of my designs. And that's why I focus on the HIEs, because I think putting LLMs properly at the HIE level can really solve public health. That's my perspective. And then making sure that regionally that they're talking to one another. So sharing the model in some of your earlier opening statements that we really do have. If I were to take CMS's National Provider Directory that we have national LLM directories where you actually have a nutritional label. So if you look at anything that we pick up, there's a nutritional label. Same thing for AI that these large language models are telling you. What data has it been trained on, what is the model, and then what have been your hallucination rate, and then that's how we get to a better public health.

Speaker E: Uh, if I could just make one quick comment. I'm sorry. Imagine this. Okay, here's the scenario. HIE is, I think, are really untapped. And that is the beauty of all of this. Right? It's free. The data. We've been trying to do this again for a long, long time. Imagine this workflow. You're back to Percy. Okay. You ask Percy. I'm living in North Dakota. I've got these 10 bed. I have a 10 bed hospital. I have these patients who are here. Tell me about these patients. Now that Percy is tied to all the data available, including social determinants of health data. And I'm just talking to Percy and I'm not even the primary care doc. I'm, um, the endocrinologist trying to see this patient because now they've had to travel four hours to come see me for a appointment that they got because Percy found it in their scheduling system and gave that to the patient. Right. So the patient has shown up, but I'm, um, the doc and I don't know what's going on. So I'm talking to Percy. Percy can now go through that hie and say over the last 10 years, this is what's been going on with the patient. They've traveled to these locations. They got those diseases. Here are the medications they started and stopped. Here is their medication rate of consumption. And they've only taken 30% of their prescriptions because they can't afford them and they're in bankruptcy and blah, blah, blah, blah, blah. That is AI. Now what does that presuppose? And here's a really scary thing, and people don't scream at me if you're on epic. That presupposes de platforming the data. So when you look at this and you truly think about human centered AI, it is about when we get the data in language models that are attributable. I love that nutrition label. That's so cool. And now you can just talk to Percy. I really appreciate the comment.

Speaker G: All right, we are in the final app. So we talked a lot about. Hey, we're doing ambient AI and different predictive analytics. But what is one thing that you see coming in the future that we should all be mindful of? And the nice thing about going first is because the rule is you can't repeat another person. So, um, so I'm gonna, I'm gonna let it be like popcorn. Whoever wants to go first, go first and then we'll hit everyone else. One thing coming up, be mindful of.

Speaker C: Oh, I have a mic, so I'll just jump in. Uh, I think I mentioned it before, but it definitely is. So, uh, I don't think anyone's going to follow a payer concept. But I do think that utilization management, because of analytics becoming so much easier and accessible. We heard before about Milliman and Tableau and how long that took.

Speaker G: Right.

Speaker C: And now we're able to crunch that data faster. That is good and bad. And I do think that is One of the things that, that we're going to see coming is there's going to be heavy, heavy utilization constraints. And it's going to be like, you know, we're going to have to go prior auth. Is going to go. Going to go up, but it's going to be like instant prior auth. Like prior auth immediate, which again, is good and bad. In those cases where there's medical necessity or it's deemed by the. The payer relationship, it's medically necessary, I think things will go faster and we'll get care quicker. Unfortunately, I think it's going to be used probably a little too far and we're going to see. See big, big crunches. So that's, I think, coming again. I think there's some good and some bad that's going to happen there.

Speaker B: All right, who's next? I see a Sixers championship, Eagles, super bowl coming up.

Speaker F: That.

Speaker B: That's what's happening in the future.

Speaker F: No, no, no, no, no, no, no, no, no, no, no, no, no, no.

Speaker B: Sixers and Six. This may sound tongue in cheek, but it's really m. What I, I think is that primary care and safety net will save the health care ecosystem. And I'm a stop there.

Speaker F: Boom, boom. Yeah, yeah, There we go.

Speaker G: I love it. All right, Jody. And then we'll finish with Vera. Uh, six seconds or less. Kind of one thing you might predict or we should be prepared for in

Speaker D: the future, I think, and it's very simple, but better data and healthier patients. And healthier patients is why we're all here.

Speaker E: So primary care is gonna be replaced by really great avatars that are PAs and RNPs and, uh, very algorithmically driven based on data from 10 million, 100 million patients with the same thing. So we can get better at it. Cause we killed primary care in the last 20 years. That's why everybody is in my emergency department. Um, and, uh, that's the one thing I think the second thing will be, is that we can learn how to get healthier in a very personalized manner. And if the payers would just pay for us to buy kale chips instead of Cheetos.

Speaker I: Right.

Speaker E: Um, and they can do that in real time while we're in the Walmart by geolocating us and giving us a coupon as we walk past the kale chips. Right. Like AI People are going to love that. They aren't going to love that because you're also going to give them a reward, which is money that is fungible so that they can Buy new TV at the end of the year. That's it.

Speaker G: You know, I actually have a quick question for our two drivers. Um, so you mentioned sports quite a bit and AI. And I was with the main uh, AI people for uh, multiple. All the Minnesota sports teams like the Vikings, et cetera. And they said at the end, uh, while they do a lot of AI and crunching data, using data that uh, once they're in the cockpit or in their seat or in their zone, it comes down to heart and uh. Yeah. What would the driver say?

Speaker D: No, uh, absolutely.

Speaker A: Once you go through all the data and me being an engineer, that is something I've had to really work on about. Once I get in the car, that

Speaker D: part of me has to go away.

Speaker A: Because in any sport when you watch athletes, they are very present, very in the moment. And if you're thinking, okay, do this, do this step by step by step, you're never going to. Your conscious is never going to be as fast as your subconscious. So it comes down to trusting yourself that you already have the data that you need and in the moment you can execute.

Speaker B: Yeah, love it.

Speaker I: I'll just say, I mean I'm learning from her, learning from the best. But I will say, uh, just with my experience with AI and how technology works, I think I do content design and I'm really advocating on the behalf of the human. Right. And just remembering that you have to trust your instinct. And so all these things that we've built are built on human instinct and what we know and what we've already kind of like historically put together. Right. And so I think it's important just to, yeah, be in yourself, be the human, be in your body and when you're interacting with anyone. And I was just talking to her about how like, you know, you have the information and sometimes AI is not always correct. And so the thing is like, like you said, you have to also like you have to rely on yourself and be able to, yeah, uh, gut check and make that connection. So I think, you know, doing that in any sport or any high level thing, it's like you have to trust yourself and, and stay in your body. Be present.

Speaker G: Back to Ben's Dan, because I, I know I utilize most of the time and I want to give you the last couple uh, minutes based on what you heard and, and some of the themes like what, what, what's your reaction and any ideas for, you know, practical things for everyone here.

Speaker F: I think there were very powerful commentary here. I think there are a lot of things that we definitely Learned a thing or two from this amazing panelist. So thank you so much for sharing your thoughts. Unfiltered. We really appreciate that. One of the things that we think about is you know first is you got to ensure the data is ready. Right? That's part of the maturity process. That's part of ensuring that there is a fabric that you can build this sophistication. Intelligent uh, layer on. The other aspect that we think about is, you know, as we augment and scale and democratize access to care, how do we ensure that this does not erode the trust between the patient and the provider? Because if you think about it, there are 40 million ChatGPT health queries happening every day. So is there a kind of like a scary scenario where patients trust ChatGPT more than their provider and is that necessarily a bad thing? All the time. Love to hear Isaiah. How do you see?

Speaker B: Yeah, I believe Dr. Google and Siri diagnosis is real M and this concept, I think there's a practical way to do it. And if you look at the 21st Century Cures act, that concept of open notes is where I think we are going now. The providers will challenge you saying you know, I should have the final say. And if anyone doesn't know what the concept of open notes is, it's where the patient can actually talk to you about what the doctor is putting in the chart before it's submit it for the claim. And so in that interaction, Google, Siri, their diagnosis, what they have been diagnosing themselves is brings in. So if a provider says hey, you are a chronic smoker and the patient says hey, I've only smoked once in two months, why are you diagnosing me with this? That can be a concept that you, you have a conversation about at the point of care and then you can challenge whatever is going in there before it goes to the patient payer. So for me I think that that conversation is going to happen more often and the providers are going to have to have more intelligent conversation. And this is where I go back to my ambient ah being framework because if it's listening to this I think that where doctors intelligence is they can read it and say okay, maybe I did diagnose this person incorrectly. Will it happen 100% of the time?

Speaker G: No.

Speaker B: But maybe by reading it back, ingesting the data that's coming from clinical decision support systems where all that the medication documentation is saying hey, maybe it's wrong and then uh, we'll go from there. That's, that's my perspective on this but one thing true for him is that you cannot renegotiate that trust. To your point.

Speaker H: Yeah.

Speaker B: And once it's lost. Yeah. It's lost forever.

Speaker F: Yeah.

Speaker I: And then.

Speaker E: Pranam, did you have a fun question for Pete?

Speaker F: If you had an AI companion, what would you want your companion to do? Pete? So just keep it not R rated. Let's keep it nice and easy.

Speaker B: So just very, very quickly.

Speaker C: We were in a prep call on this and I said, if you want to throw a fun question out, so I'll go very quickly. And I talked to. I actually talked to Ed about this this morning and Matt. So over the summer, I took, um, I finished a merger and I took about three months off and I took an AI sabbatical and I spent the summer building a digital twin of myself. And so where I'm going to go with this is I recorded hundreds of hours of. This was a professional digital twin. Not a. Not a private one, but I kept it all professional about my career. Any podcast ever been on, I transcribed, I took personality tests and dumped it all in. One thing, the outcome of it was is that, uh, one, I found the limits of AI, which was great. And I built an avatar and you can feed it API calls. But anyway, what I did find out though is that the data, just to wrap this up, was the king in the whole thing because the technology could change all the way around. But that underlying data was absolutely the core component. And best part is once that data is built right, I can take that to whatever technology platform for the rest of my life.

Speaker F: Thank you, everyone. Thanks a lot. If you enjoyed this episode, make sure you subscribe to our channels. Subscribe episodes will be posted once a month with a variety of industry leaders and you can follow us on LinkedIn, Twitter and Facebook. As a garage, have a good one.