61 – The Future of Healthcare: AI, Innovation, and Impact with Ambar Bhattacharyya

Marcus: [00:00:00] If you enjoy this content, please take a moment to rate and review it. Your feedback will greatly impact our ability to reach more people. Thank you.

Marcus: All right. So, uh, welcome back to our guest series here at Health Further. Vic and I always, uh, trying to educate ourselves, make ourselves smarter by bringing the smartest people at healthcare innovation to the show. And, uh, today really happy to have my friend, uh, colleague who’s, uh, in the Aspen Health Innovators Fellowship with me, Ambar Bhattacharya, who is a managing partner at Maverick Ventures.

Marcus: Ambar, how are you doing, man?

Ambar Bhattacharyya: Doing great. I’m so excited to finally be on their pod. This is awesome.

Marcus: I know we’ve been trying to make the dates work. You’re really busy. We’ve had a bunch of crazy stuff going on, but man, thank you so much for making time for, for us. You, you, you probably don’t even know this, but we have had your work featured on the show.

Marcus: Um, Vic and I, every week when we do our roll ups at the end, we’ve gotten [00:01:00] to the. to the place where now we spend about 10 minutes kind of running down AI stories. And I think before we started that 10 minute AI rundown, one of the things we anchored everybody around was the article you did maybe 10 months ago, uh, on the AI checkup that you and some of your colleagues do on this sub stack.

Marcus: And it was a great sort framework. It had a visual infographic to sort of lay out all the different areas where Gen AI can be effective. So it was a great starting point for us to start our conversation. So thanks for your contribution to the space overall. And uh, yeah, happy to have you on the show, man.

Ambar Bhattacharyya: Oh, this is great. Well, first of all, thanks for. Thanks again for having me to, uh, you know, we put a lot of time and effort into thinking about these about these spaces and we could talk more about that in a second, but I’ve always found frameworks help, you know, and so we’ve been trying to experiment with them, track them, look for them for consistently for consistency.

Ambar Bhattacharyya: Not just, you know, amongst investors like, like, like [00:02:00] you and I Marcus, but also with other key stakeholders, you know, regulators, customers, startups, physicians, nurses, you know, really try to technologists engineers, like really get us, you know, state of the state. And so it’s been really fun to, you know, to publish some of that, right?

Ambar Bhattacharyya: It’s a new skill for, for, for me to, you know, how do you. How do you internalize and then, um, and communicate some of these findings in a space that is moving exponentially fast? Um, and so that’s, you know, that’s kind of where we are for today.

Marcus: So Ambar, before I pass off to Vic, because you and I have had multiple conversations, and I want to make sure he gets his questions in, he doesn’t get to speak to you as often as I do.

Marcus: Uh, I, I was mentioning before we started recording that One of the reasons why I hold you in such high regard as it pertains to the intersection of generative AI and health care is that in our first Aspen seminar, this was the fall of 2022. So this is pre chat GPT or any of that kind of stuff. Everyone was going around the room kind of saying where they think things are going, what they’re really passionate [00:03:00] about.

Marcus: And you were the one, not just in that room that time, but I think against the backdrop of all the things that other people were saying in the seminar. space to make the statement that AI is going to have huge implications for health care specifically. Um, and I, I have to admit, even as a technologist, because chat, GBT hadn’t arrived yet, I didn’t fully understand where you were coming from.

Marcus: So I think before we dig into where we are now and some of the opportunities, I would love to just understand you, you know, you’re based in Silicon Valley. What were you seeing back then? What were you seeing in 2022? Maybe even 2021 or 2020? What were you seeing building up to that point that you had already before the rest of the market got there built your thesis around the intersection of Gen AI and health care?

Ambar Bhattacharyya: Yeah, so maybe I’ll answer with a story. Maybe two stories. You know, about, about why, why I got really excited about this [00:04:00] early. So, um, and they’re a little bit of non sequiturs, right? But maybe they’ll, maybe they’ll, they’ll be insightful. So, um, maybe about 10, 12 years ago, I was, I was investing at, at another fund, Bessemer Venture Partners.

Ambar Bhattacharyya: And we were at the, at the back then Bessemer, we were at the forefront of another innovative wave, which was, um. In and around how the Internet was democratizing for the average person, um, and they can actually start earning an income off the Internet. And so this was, you know, when Uber was starting door dash, um, uh, eBay had already started like a, you know, a decade ago, but there was a few companies that had started.

Ambar Bhattacharyya: Uh, there that were in the Bessemer portfolio, uh, Shopify and Pinterest and Twilio in particular. And I was a pretty young, pretty young person, um, at that time. I’m [00:05:00] still pretty young, but I was even younger, younger back then. And one of the ways that I thought about learning about these new tools was trying them.

Ambar Bhattacharyya: And so I said, okay, what I’m going to do is I’m going to try to make a Shopify shop and see what it’s like. I’m going to make a Pinterest board and see if, you know, how, how those bookmarks work. I’m going to go on Twilio and set up my own, you know, text, uh, text type thing. We invested in a company called Zapier and I tried to connect my.

Ambar Bhattacharyya: You know, my, my, my Salesforce updates to getting a text message. It was, it was all like, it was like, how easy is this for the average person? And, uh, and by background, I’m like, I learned how to program at a young age, but I’m not a technologist. Like I’m not a software engineer. And what I found was these tools made it so easy to sell things on the internet, create new designs, integrate the web and text.

Ambar Bhattacharyya: It was so simple. And I was like, Holy moly, like this is going to [00:06:00] be awesome. I don’t know how big these things could be, and they all ended up way bigger than I could have possibly possibly imagined. But I think that first person ability to try something and use it gave me that like wow moment. Uh, so then you fast forward another five years and, um, Yeah, maybe seven years.

Ambar Bhattacharyya: And, and, uh, crypto and blockchain started taking off. And I was like, oh, like this could be it, you know, like, and let me go try all these things. And there was all these things called DAOs, like these distributed autonomous organizations and different things being on the blockchain. So I just went through the same exercises.

Ambar Bhattacharyya: Oh, like, let me try it, right? Like, let me see, like, if this is the same Shopify Pinterest experience that I had, like, you know, seven, nine years ago, and I went through it and holy moly, it was not so hard to use. None of made any sense to me. Uh, we, you know, me and two other, um, two other folks that, [00:07:00] you know, that I collaborate with, we started a DAO, like, ever so briefly.

Ambar Bhattacharyya: It was like the shortest lived DAO of all time. What exactly the purpose was, we tried to integrate all this technology, we made all these NFTs, we did all these things, and none of it was intuitive, like none of it made any sense. And I was like, Maybe it’s me. I’m at least a little bit older, but I think like it just wasn’t there.

Ambar Bhattacharyya: You know, it was like it was like in my mind. It’s a technology looking for a problem. Um, and and I and I just didn’t think it would cross that chasm. At least not that iteration of it. So then you fast forward another four or five years and this stuff started coming out. And, um, again, like just my mentality, like I got to try it.

Ambar Bhattacharyya: Right. And so before chat GPT came out, there were, um, uh, you know, AI existed, right? I mean, there, there was GPT too, right? Um, there was a open AI playground. Um, there were some open source models. Um, there, there were a few different things out there, um, [00:08:00] that were really, uh, you know, cutting edge, but still usable.

Ambar Bhattacharyya: And once I started using those things. I was like, holy moly, like this is actually changing the way that one can actually summarize information search for information. Um, and it was so intuitive. Um, and I started seeing how other folks who are programmers were generating code from it at a really, um, you know, really early time.

Ambar Bhattacharyya: And so then I started just calling as many people as I could about this, um, uh, you know, customers, non customers, just testing these hypotheses. Now that I, I started getting early, you know, early conviction that this might be something. What was fascinating to me was just the myriad of use cases people thought about, not just in health care, you know, but just like outside of health care, you know, customer service search, um, uh, you know, virtual [00:09:00] friends, you know, virtual therapists.

Ambar Bhattacharyya: I mean, you can go on and on and on, um, writing poetry and summarizing Shakespeare, all that stuff. And, you know, It totally just captured my imagination and given the vast majority of what I do is in healthcare, I think I was able to kind of make that connection between like, yeah, this, this technology. It pointed in the right direction with the right guard rails and the right use cases could have a pretty transformative effect.

Ambar Bhattacharyya: And so that’s kind of what got me jazzed up about it. It was a long answer to your question, but hopefully that that arc, you know, gives you a sense of it.

Marcus: It does. It does. I mean, it says to me that a really two things one, uh, playing. Playing with things early on, uh, is, is key to how you develop your own views of where the world is heading.

Marcus: And then I guess the second thing is that you have a high, um, But you, you, you rate the value of ease of use very highly. Uh, and, and I [00:10:00] think that’s a, that’s a, that’s a, That’s probably something that more VCs should be considering rating very highly, right? Like not, not theoretical utility, but like practical, I played with it.

Marcus: Is it actually easy for me or not? And how do you weigh that into your own decision making process?

Vic: Yeah, I think it goes to the adoption. Problem, the adoption curve, right? If it’s easy to play with, easy to compose into different things, there’s lots of potential use cases that could grow into other things.

Vic: And I think crypto is a good, good example where maybe it’s okay if you’re in another country and you have trouble with your currency. But in the U. S. I don’t mean, yeah, maybe we don’t like how inflation is, but it’s sub 5 percent a year. It has been for a long time. And so, um, the real like pressure to switch over and all of the pain of going through the rotation, and then maybe not worth it.

Vic: And so that, that may be the pretty good, uh, metric to decide is this basic technology going to be adopted at some scale over time or, or not. I think [00:11:00] that’s kind of interesting.

Marcus: Yep. Okay. That was super helpful. So fast forward to today. Um, We’re now seeing lots of companies getting lots of rounds done, uh, around AI and healthcare, uh, sort of high level things that I think we’re seeing, we see a lot of stuff around clinical models, we see a lot of stuff around workflow, uh, and, and workflow efficiency, we see ambient technologies, um, I think in a couple of weeks, we’re going to have Steve Lieber in who, uh, who, uh, Is is, uh, collaborating with Chime and HFMA and a bunch of other organizations around a standard for smart, smart hospitals, um, and working with care, care, AI, et cetera.

Marcus: So those are kind of some, some high level themes. But if you had to lay it out for us and for the listeners in terms of a framework, where would you say we are now? High level categories, high level areas of, uh, of impact where AI is. Meaningfully making progress today in health care. [00:12:00]

Ambar Bhattacharyya: Yeah, I’ll, I’ll state it like this.

Ambar Bhattacharyya: You know, the framework that we put out, um, and updated, uh, both, you know, a few quarters ago and then this quarter, we looked at it on two different dimensions. One is, you know, technical complexity versus technical simplicity of actually building a product. Um, And then we looked at another dimension of, uh, you know, signs of customer adoption versus how visionary it is in the future.

Ambar Bhattacharyya: And, you know, what’s interesting, if you track that same exact grid over a three quarter period, uh, a lot of it. Uh, we have seen some early adopters on the provider and payer side of certain generative a I tools. Um, and plainly speaking, I think to have have really accelerated. Uh, one is in describing space.

Ambar Bhattacharyya: Um, and you mentioned this Marcus. Um, [00:13:00] these are companies that Yeah. Uh, are dealing with the physician burnout issue that is widespread throughout America, frankly, the world. Um, and, you know, it’s the problem of physicians staring into a computer, taking notes afterwards and, uh, all that stuff. And this is a generally is a great use case.

Ambar Bhattacharyya: It’s a great engine for automating a lot of that processes with a lot more accuracy. Um, And we have seen both startup and, um, incumbent adoption of that. Uh, that’s kind of 11 category. And the second category is also in the back office, but as it relates to, um, a revenue cycle, um, and some of the billing parts of this, uh, you know, again, this is another great use case of generative AI, where it can summarize lots of information really fast, and, um, it can go through and make connections that, you know, Uh, might have been overlooked, [00:14:00] uh, might have been done incorrectly, uh, and the field has progressed very rapidly there.

Ambar Bhattacharyya: And I think the ROI for health systems and insurance companies is very immediate once you tie something directly to the revenue cycle or prior off, depending on which side of the of the equation you’re on. Both of those that we’ve seen, you know, real customer dollars flow into it, and I think it’ll continue.

Ambar Bhattacharyya: I think, um, on that back office admin side, I think things have been, um, frankly, more, a more rapid pace of adoption than I would have expected. Um, the EMRs have actually played relatively friendly so far in this, um, you know, the epics and the Cerners, the Athenas of the world have all, Um, and others have have all done a pretty, uh, pretty good job of allowing startups toe exist and flourish in that.

Ambar Bhattacharyya: Of course, that can change any given moment. So I think that’s kind of one thing that we have seen. I’ll I’ll say in the next breath. [00:15:00] Um, you know, something that we didn’t really talk about in either post, but, um, I think we’re beginning to see it is the, um, it’s the ability of a I to start. Um, uh, you know, a lot of people use the word copilot, right?

Ambar Bhattacharyya: I think it’s like a very safe word. Um, there’s probably less less safe words you could use to, but like, we’ll start with the safe word. Um, uh, a lot more copilots on the clinical side. That’s something that I, I didn’t necessarily expect to happen, um, from an adoption perspective as early, um, as it is. Um, because, you know, the, the, the copilot is a very well accepted term.

Ambar Bhattacharyya: Uh, the, the, the next, the next word that you use after copilot is autopilot. And that is that, that, that is a less, uh, a less well accepted work , at least in the healthcare world. So somehow for pilots, it’s okay. Um, but for, um. For health [00:16:00] care. It’s not quite there yet. Um, but I think we’re beginning to see some green shoots.

Ambar Bhattacharyya: Of things that are certainly co pilots and potentially over the course of the time could be autopilots. Um, and that is, uh, and that’s a whole different, fascinating, um, you know, can of worms.

Marcus: Yeah. Tarun Kapoor, uh, is, has been a guest on our show and I guess he’s kind of a recurring guest over at Virtua Health.

Marcus: We, we brought him on to explain to us. Uh, the rationale at Virtua for launching Wobot. Um, and he’s actually going to come back in August and sort of talk a little bit about like how it’s gone and what they learned through that process. But he was very clear about two things that I think relate to what you just said.

Marcus: So the first was, He was, he was clear with us that Wobot is not, uh, gen, gen AI. Uh, it, it is machine learning, but it’s all sort of pre programmed effectively. So it’s, it’s choose your own adventure, but it’s pre programmed adventures. Right. So, um, the, the risk of hallucinations was not really there. Um, the, the second thing though, is that [00:17:00] doctors are not going to be able to keep up with the rate of change of information in the medical space and.

Marcus: And there’s going to be demand for the latest and greatest technologies and therapies and modalities, et cetera, from the patient population. And the intersection of those two things is going to force, I think what you said, which is a more rapid adoption of co pilots than people anticipate. But then, I would say, even going further than that, once you layer on clinician shortages, um, And in other workforce related challenges, I think we are going to have to start stepping in and this is really where we’re robot is going, right?

Marcus: It’s we have a larger behavioral health, um, you know, population of need than we have clinicians. And so can we? triage the less severe, um, you know, issues with programs, right, as opposed to with people. And I think if [00:18:00] we just play this all out 5 10 years, all the trends sort of point to exponential, you know, change of information in the medical space, less availability of providers, um, certainly there’s going to be a huge mismatch when we get to elder care.

Marcus: Um, and you’re going to need some of these technologies that, oh, by the way, are improving in their, uh, quality to deliver the right answer every week, as far as we can tell, you know, every week we’re, we’re reporting the news on what’s going on one week. It’ll be clause number one, then it’s GPT 4. 5 is number one.

Marcus: Now it’s, you know, uh, uh,

Vic: uh, met

Marcus: Gemini is number one. It’s literally every week. You know, I don’t think we would know that if we weren’t recording the show every week, but it’s every week. There’s a new winner right in the LLM space against all those different, um, metrics. So it does seem like those, those things that you called out there are, are right.

Marcus: Can you talk a little bit about [00:19:00] what you’re seeing in hallucinations, right? Like. Like, is, are the LLM, uh, uh, makers really getting after that problem? Because it seems like even just from a uneducated perspective, you know, you talk about the word autopilot being a word that people fear. It feels to me like Hallucination is another big, big worry that people have about, you know, letting Gen AI sort of do the work in place of people and then getting something wrong, it having a catastrophic outcome.

Marcus: And even if we improve all of the metrics statistically, the story of that one person. who was impaired from a hallucination, I think will just not be accepted very well in society. So what can we say about the progress the LLM makers are having with hallucinations?

Ambar Bhattacharyya: I’ll say a few things. So one is, I think we’ve both used this word, uh, but I’ll, I’ll use it [00:20:00] again, uh, in, in describing how fast the field is, is advancing.

Ambar Bhattacharyya: And it is, It is, it is advancing at an exponential rate. And I think there’s been many studies talking about how hard it is for. The human brain to comprehend what exponential means myself included, put myself to that. If you have

Vic: a human brain, you’re in there. Yeah.

Ambar Bhattacharyya: Yeah. The, the, um, your comment on the leaderboards is spot on.

Ambar Bhattacharyya: And, uh, your, your comment on, you know, what, you know, things changing on a weekly basis, um, it’s not just changing anyways. It’s actually the other way to put it. It’s improving on a weekly basis, right? These things are getting more and more accurate now. What a few as it relates to hallucinations. I think there’s a lot of there’s a lot of different angles in which, uh, to think about that.

Ambar Bhattacharyya: So one is and perhaps the most. Yeah. Uh, [00:21:00] well accepted one today, uh, is going back to copilot, right? Another word for that is a human in the loop, right? And you, whether you’re, uh, you’re dropping generative AI to read a, um, uh, an image like a radiology image, and it can actually do image recognition, and it’s trained on a whole bunch of, um, uh, x rays or MRIs or CTs, whatever it is, And it’ll, uh, recommend something there’s some radiologists that will have to say agree or disagree, right?

Ambar Bhattacharyya: That is, that is one definition of it. Um, and that, that reinforcement learning is, um, is what is really, uh, hopefully going to eliminate hallucinations in the long term because you find some model model gets better, et cetera, et cetera. Now, like a sub, a sub thread of that is, you know, the, the next wave of reinforcement learning.

Ambar Bhattacharyya: Uh, is is a I, you know, reinforcement learning. So instead of human in the loop, it’s a I in the loop now. Now we’re getting really meta here. And, um, but the [00:22:00] but the difference with that is that it could be, uh, you know, the A. I itself is training on the reinforcement learning, which is really fascinating to think about.

Ambar Bhattacharyya: Um, and when you think about going back to the exponential learning curve, Reinforcement learning does take a long time, right? Because you need a lot of radiologists to click on a lot of images to say, hey, thumbs up, thumbs down. And if it’s up and down, what should it be? Just, you just need a lot. And, uh, AI, you know, reinforcement learning, and you, you can, you can automate that.

Ambar Bhattacharyya: And now still an open question in the academic literature as to what the Delta is between those two. But, um, if over the course of time, AI enforcement learning, uh, reinforcement learning. Uh, if it can get to the stage of human reinforcement learning, that’ll be a very interesting moment for us, um, as, not as a society, but also in, in healthcare because then you, [00:23:00] you are the medical literature, the, the academic literature could argue you’re better to actually have an AI in the loop than a human in the loop.

Ambar Bhattacharyya: As the guardrail, which would be really fast. We’re not there yet by any and all means we’re not there yet, but it’s just, just one, one sub thread to go down. But that’s like on like the human loop. Second thing is like guardrails, right? You know, and there’s a lot of different guardrails that could exist and they’re starting in different ways.

Ambar Bhattacharyya: Um, there’s, uh, government sponsored guardrails, you know, like the chi of the world, et cetera, um, that are industry consortiums, which. Uh, I think the industry is very rightly trying to regulate itself. Um, you know, as what does rapid change? Um, there’s a lot of, uh, uh, startups that are trying to do this, you know, larger, large conglomerates, the ethics and Googles are trying to do this too, um, because they realize what’s at stake and Marcus, what you said exactly, right?

Ambar Bhattacharyya: Like, If a hallucination is a, uh, that leads to a bad medical event is a really bad situation and that story will be told and it should be told, um, in a, you [00:24:00] know, in, in a real, in a real, um, important way and you don’t want that to, um, you don’t want that to happen, right? Just kind of pull stuff and then there’s like a third, like, cut of it, which is, again, I like analogies in different industries, which is looking at self driving cars and self driving cars.

Ambar Bhattacharyya: There was this huge question nine years ago. You know, eight years ago, like what happens the first time a self driving car gets into an accident? Who’s liable? Is it Waymo or cruise or, you know, Zooks or whatever? Uh, or is it the, uh, you know, is it, you know, the algorithm? Is it nobody? Is it the other driver?

Ambar Bhattacharyya: You know, there’s all these like open and open ended questions. And, um, and like now, like you fast forward a bunch of time. And there’s been a ton of accidents with the self driving cars. Right. And unfortunately there’s been casualties with self driving cars, right. And the people have been [00:25:00] like hurt, killed all that stuff, terrible things, right.

Ambar Bhattacharyya: Like things that you, uh, things that you, you really never want to hear about with self driving cars. And granted, I live in a city full of self driving cars, so I see them every day. But I think what’s interesting to look at is how that industry has talked about those events. And again, everyone talks their book, but if you look at what the data that they have collected, the self driving car industry, um, and they’ve compared it to human drivers.

Ambar Bhattacharyya: Saying like, Oh, here’s our rate of accidents per million miles. Here’s humans, you know, you know, rate of accidents, every million miles. And again, everyone collects different data. As I understand, there’s like different, different, different cuts of everything. But I do think the self driving car industry has presented a compelling enough case.

Ambar Bhattacharyya: of accident occurrence, accident avoidance, such that in multiple states now, self driving cars are expanding, they’re growing, not just in popularity, but from a regulatory [00:26:00] perspective. And so the question is what in healthcare, what’s the equivalent, right? Because there’s some, there’s some benchmarking here, not just to no hallucinations, right?

Ambar Bhattacharyya: Like, you know, no never events would be the healthcare word. Or is there a perfect bar like the health care system as it exists today? And when will the data sets be large enough such that you can compare both of those? And what is an, what is a societal, you know, acceptance, you know, between those two bars?

Ambar Bhattacharyya: Um, and that’s for us as citizens, regulators to decide, industry members to contribute to, patients for us to think about. Uh, because it all goes back to what you said earlier, markets. Access, it remains a huge issue in health care. NPS, even when people can access healthcare is very low. And to really the pace of medical literature that is being published and how to face a change is, is, is overwhelming.

Ambar Bhattacharyya: Um, and so how do you compare all of those real [00:27:00] tailwinds that are demographical and they’re not going to get solved with some of those risks that we, that we have to deal with? That’s for all of us to come together and try to solve.

Vic: John, I think I’m excited about that. I’m more interested in the next six months.

Vic: Like your assessment of something I’ve been thinking about a lot, which is the two things you called out, which I completely agree that revenue cycle management or, um, the other side of the payer side, sort of a duty claim pre authorization. And then also, uh, scribing had been the two big winners early on and there’s other, there’s other spaces.

Vic: But. It strikes me that those services, those, those jobs to be done in the health care system already were seen as, uh, being outsourced to maybe to another department in the health system where the payer, um, or to, there are scribing like actual people in other countries or in this country. And so that was an easy place for [00:28:00] participants to say, well, I’m already sending this.

Vic: Down the hall, maybe I should try sending a few claims to this other group and see how they do. And, um, it strikes me that we already have, uh, lower licensure extenders in healthcare, with lots of, Physicians, maybe not every specialty, but lots of them have that concept. And we have care coordination and chronic care management, uh, methodologies where if you have diabetes, there’s a set of things that maybe some of the details we don’t all agree on, but the broad things that a patient should do when they’re first diagnosed with diabetes, pretty well established.

Vic: And, um, is there an opportunity to take sort of That example or others where like it already is an existing thing, but we, we keep the doc in the loop, just like we have the doc in charge with a nurse extender. Um, and that’s a place [00:29:00] where the humans that are making the decision say, well, I, I can sort of conceive of that.

Vic: And the hallucinations are not that big of a deal because I, I want the AI to explain why this activity for the diabetic patient is important. Maybe teach them about it, talk to them and, and help them understand. But the, the methodology or the care management protocols have been widely established. Is, is that a reasonable place to think we might see more of the adoption?

Vic: How do you think about the next part of it?

Ambar Bhattacharyya: Yeah, I think he said two really important things. You know, one is, uh, a large part of healthcare, uh, outside of the clinic is around engagement. Like, how do you engage that patient? Uh, in their time and need, whether it’s chronic care management, recovering from a knee surgery, preventative medicine, you know, medicine, pre diabetic, you know, becoming diabetic, you know, the key to any sort of successful, you know, risk [00:30:00] based model, value based model.

Ambar Bhattacharyya: Uh, is is is engagement, right? Like, kind of full stop. One thing that you’re alluding to are these physician extenders, nurse extenders that are been tasked with that role today, right? Whether it’s a, you know, uh, on the provider side or a pair side, um, that’s their job. And I think, you know, Industry wide, uh, engagement rates are still pretty low, right?

Ambar Bhattacharyya: I think just, you know, like, there’s there’s some spikes, some lows, but generally speaking, like, pretty low. I think one opportunity here is, you know, how, how do you, how do you take that the power of AI to have a conversation real time? Both voice or chat, um, and use that, you know, in a way that can actually drive up more engagement, more personalization, things of that sort, uh, in the last few years, I think most of us have thought about generative AI just in chat form, right?

Ambar Bhattacharyya: We are chatting with, you know, chat GPT on our phone or computer and it spits out some answers. Uh, the [00:31:00] voice side is taken off. It’s, um, uh, This is like a embarrassing example, but, you know, we have a young kid at home and, uh, very often I have to like, you know, burp the kid and walk around until you hear a burp or whatever.

Ambar Bhattacharyya: Right. You know, and, and it’s, it’s, you know, uh, it’s more or less boring. And there’s not a lot of people to talk to you while you’re, while you’re burping. And so. Um, and this will, this will speak a little bit, don’t judge me too much, but, um, you know, one of the things I used to do while burping this kid was, um, there was a, uh, a company called inflection, which Microsoft kind of, you know, bought the team of, but it exists today

Marcus: and

Ambar Bhattacharyya: it is voice voice bot called pie.

Ambar Bhattacharyya: And while I was burping this kid, uh, I would just talk to pie. You know, like I just keep talking to it. I keep asking pie questions. It would it would answer back Sometimes I would ask me questions. I would have to answer back. It was [00:32:00] like the strangest thing ever like I mean truly I’m just talking to like You know, like a I buy, um, but like I bring that up because, you know, this is, you know, you know, um, you know, 20 fall 2023 like I’m describing this memory and it’s come so far from that, right?

Ambar Bhattacharyya: You know, like now there’s multiple voices. There’s no latency, all this stuff. And you can just fast forward that again, another nine months, another 12 months. And it’s going to be just so dramatic, different languages, the ability to like speak at different grade levels and translate real time. Uh, you know, how do you communicate complicated medical things in simple ways?

Ambar Bhattacharyya: Medication adherence, all these things that engagement is at the core of, I think are now like solvable, um, or at least there’s a different way of solving them. I don’t know, a different way of solving them. And it’s super cost efficient because you don’t need to hire 40 case managers to make 400 cold calls in a day to see if like 18 people pick up, right?[00:33:00]

Ambar Bhattacharyya: That, that’s the state of the art today, by the way, right? This is just, you know, how do you do that? So I’m a firm believer that we’ll see some really interesting examples there.

Vic: Yeah. So just to build on that, I mean, I agree. I agree completely. I’ve been using both Gemini and Chatty Patee’s voice thing on the way driving.

Vic: Like I might be driving to Atlanta or driving home. And instead of just listening to a podcast or listen to the music, you can have a conversation and, and actually, uh, it goes in all different directions. Like if, if Marcus and I were riffing on something, Uh, and then I have the whole record of it because it’s, it’s stored there and I think to your point about the extenders, the state of the art is them dialing all day long and trying their best to engage, but engage within a very limited time, right, engage with on bar and then get onto the next call and instead you could have a voice A.

Vic: I. Talking to me about college football, talking to Marcus about jiu jitsu, talking about whatever the [00:34:00] person likes they can interact with. And that’s a lot of engagement. It’s not even necessarily healthcare details. It’s more of their kids and where they are thinking of going on a trip this summer. And all that personal stuff is how humans kind of bond and engage with each other.

Vic: It’s

Ambar Bhattacharyya: totally right. You know, you guys are in Tennessee, like, you know, one of my first conversations with Pi was about Paramore. You know, you know, band band. Yeah, I see. And I don’t know much about it. I became moderately obsessed for a minute about it. And I just kept asking questions. And I kept asking, Oh, what are your favorite songs?

Ambar Bhattacharyya: Why do you like them? And you can just flip that in terms of if that is an AI case manager, next time that he or she or it, whatever it calls me, be like, Oh, have you heard that? Like, like the most recent Paramore song, right? Like, that would just be a way to just You know, do it or a text message being like, Oh, I think you’d like to listen to this or whatever totally unrelated to healthcare, but it builds that trust, builds that connectivity.

Ambar Bhattacharyya: It sounds really strange to use those words when you’re talking about technology, but I think that’s part of it. [00:35:00] Um, so I I’m, I’m totally with you, but going back to what you said, the second important thing you said that I’d want to lose track of here, which is you talked about, are there non clinician ways of doing this?

Ambar Bhattacharyya: And. You mentioned a few of these things, you know, nurse extenders in the diabetes care management area. I think you hit on something that I really strongly agree with, which is there’s a whole categorization of health care workers, which are allied health professionals, and those folks are critical to the health care ecosystem.

Ambar Bhattacharyya: Um, they do so much hard work, um, and there’s not enough of them. Um, and it’s a really hard job. And again, whether it’s copilots or autopilots. Um, a lot of those, a lot of those, um, allied health professionals, um, uh, do have a lot of rules, right? A lot of, you know, um, a lot of, um, set boundaries, uh, of which to go on, which you can train a model on, uh, and so, you know, will we see the first, um, [00:36:00] uh, generation of generative AI co pilots and or autopilots In that area.

Ambar Bhattacharyya: Right. Which is not quite replacing doctors, not quite replacing nurses, but it’s replacing those extenders to allied health professionals and not repla. Maybe replacing is the wrong verb too. Um, uh, but augmenting is probably a safer word to use. Um, you know, I, that, that, that’s where I’m looking. Like, I, I think that, I think you’ll, there’s a high likelihood of that.

Vic: Yeah. I, I like that and, and. that they are using a pretty defined, uh, Marcus called, called it sort of a choose your own adventure. It’s pretty well defined, a set of scripts that the AI can use. I think you can have a balance with maybe on the medical, uh, advice, medical questioning. It could be pretty prescriptive, but then it could be riffing, kind of talking about your band or college football or what happened this weekend.

Vic: And if it hallucinates about a college football thing, half my friends have hallucinations about their [00:37:00] college football team. So I don’t, I mean, there are places where I think, you know, AI could be much more socially accepted in that. That chit chat, which is how you build trust with patients, and then that, that leads to engagement over time, I think.

Vic: And then, I, I like your positioning around, uh, augmenting, I’d even go further to say empowering. Like, the workforce and healthcare, right, we don’t have enough clinicians. We don’t have enough physicians, not enough nurses, not enough extenders, not enough allied professionals. We don’t need to replace them, but we just don’t have enough humans.

Vic: So, we, we need more care broadly. And I think we can help the clinicians be more empowered to do the work, like why they get into the field. What was their calling? Let’s let them do that.

Ambar Bhattacharyya: Yeah, I’ll, I’ll, I’ll respond to that for a second, but I have a very ill, uh, or ill form daughter, right? Ill form, but it’s not even a half baked.

Ambar Bhattacharyya: It’s like, you know, one 10th bake. Um, so definitely not ready for podcast, you know, conversation, but whatever, perfect [00:38:00] setting then. Um, I think on that thread. Like, what I’ve been really interested in is, you know, like, what business models could exist? That would not and take it would not antagonize the existing medical system, you know, like would not antagonize a nurse or a nurse union and would not antagonize a physician or a group of solutions or a medical association, but in fact would be embraced by them.

Ambar Bhattacharyya: And I think a lot of a lot of the fears are twofold. You know, one is replacing labor, big, big fear, right? And like, not just in health care, but overall, uh, and two is, you know, You know, with this, um, commoditizing my salary. So even if I have a job, that is somehow, you know, technology being able to do this, take down my salary because.

Ambar Bhattacharyya: You know, the, the nurse, the phone call by the nurse, you know, the nurse may make X dollars an hour. You know, the, the, the, the phone call done by, uh, open [00:39:00] AI API is going to be 4 cents a minute, right? You know, and it’s just, there’s like a very, very big difference, you know, on an API call basis. And I think what in the place I’ve been trying to, you know, intellectually explore, I don’t have an answer yet is, is there a way where you can enable the existing health care system to embrace it?

Ambar Bhattacharyya: Where. They are actually extending themselves and every nurse probably and doctor every but many nurses and doctors probably do feel overwhelmed, do feel burnt out. And wish they can make a clone of themselves in some shape or form to help more. I mean, that’s the mission driven world we’re living in. So if you reframe it that way of like, this is a clone of Marcus, a clone of Vic, clone of Bombar.

Ambar Bhattacharyya: And Oh, by the way, this is actually helping, helping reach more people. And also, by the way, there are more people that you’re able to reach. This is actually going to help you, you know, monetarily help more people from a mission perspective. Um, I, I, I wonder if there’s a surface area there that, that could be a win win.

Ambar Bhattacharyya: I mean, it’s very against like quarter [00:40:00] baked at best, but that that’s kind of where my mentally I’ve been exploring.

Vic: Yeah. I’ll be an early investor in that.

Marcus: All right. So, uh, we’re, we’re, we’re coming up on time, but I, I can’t help because there were several things that were said that I’ve just want to, I want to dig in on a little bit more.

Marcus: Um, And maybe just leveraging that last conversation. And talking about not antagonizing the existing industry, right? It’s like, I feel like when you say the existing industry, it’s not entirely clear who you’re talking about there, right? Because, um, you could be talking about, uh, health system at the highest level.

Marcus: You can be talking about health systems or you could talk about payviders. And, and I think those are two pretty different players in the incumbent see today. Um, I, I think one of the, the things that has, uh, Really irked me, uh, [00:41:00] this year as I’ve been deeply embedded in the healthcare system is that, and this is going back to your points of when you first use Shopify and maybe tied it up with, with Zapier to, you know, Pinterest board or whatever.

Marcus: You’re like, Oh my God, this is like incredible. There’s nothing incredible or remarkable in healthcare technology today. Right? I mean, it’s, it’s actually, Unbelievably frustrating, um, as a, as a patient and as a customer, because we’re paying, you know, at the end of the day, we’re paying and it’s, it does seem to me that in my experience, the standards and the expectations for where we should be are generally speaking, pretty low.

Marcus: Right. They’re pretty low. And I feel like AI is one of those technologies. People have, you know, likened it to the internet itself. People have likened it to the iPhone. And I think what’s, what’s, what’s so, um, [00:42:00] what’s so transformational about these kinds of technologies is they change the way we work.

Marcus: Like they fundamentally change the interface. You know, I’m thinking about like the case manager or the clone, you know, these are situations that. could, because they’re capturing data, much like Gemini or Chad GPT does when you’re talking to it, radically change how we capture data from patients, right?

Marcus: Because it’s, it’s voice, you can catch inflection. There’s all sorts of, you know, you can get color and tone in addition to actually transcribing the actual words, right? Um, and it just feels to me that The overwhelm and the burnout and the existing dysfunction is so severe that it impairs the ability to dream, to envision like this bright new future and that in and of itself, that cultural malaise could potentially be the biggest threat to AI actually, um, [00:43:00] showing up in the, in the healthcare world over the next five years.

Marcus: And I’ll just make one really quick point. You know, Vic, Vic and I, we invested in a, in a company that started as a medic med adherence, and then moved into doing remote patient monitoring, um, was Pilsy. Then it turned to an optimized health, uh, and, and then us venture partners ended up, you know, placing a large investment and they were kind of off to the races.

Marcus: Um, and I remember like, as that transaction was happening and the space was, you know, being filled with RPM, everyone was just covering the areas. Oh my God, another RPM company, blah, blah, blah. I don’t want to hear about RPM. And then, you know, my dad gets sick. And he needs his blood pressure monitored, you know, three times a day, and I love his PCP, but they send me home with a Walgreens Omron blood pressure cuff, and they tell me to measure his blood pressure three times a day, and write it down on a piece of paper.

Marcus: And take a picture of the paper and then upload the paper [00:44:00] to my chart, dude, like, you know, no, we do not have enough remote patient monitoring. No, no, no, we’re not even close to enough remote patient monitoring people, you know, and so I guess, like, what would you say to that? I know that this, this is not this cultural issue in health care.

Marcus: Innovation is not lost on you. Um, you know, you came from Bessemer, which, you know, between. Uh, everything they did in, in, in consumer and SAS clearly sort of revolutionized spaces that had higher standards where competition rose, you know, raise the standards up in healthcare. We don’t have that. So how do we overcome that?

Ambar Bhattacharyya: I, first of all, it was really well said and like, That’s that’s a harder situation that you had to deal with. So that’s just first of all, just my heart goes out to you for that. You know, secondly, um, related to your question, I had the reason I spend, you know, part of my time outside of health care is to make sure I can imagine what is possible, [00:45:00] um, and to make sure I, I, I’m able to understand those wow moments when they occur, because, you know, Over the course of time, uh, those, at least in my firm belief, those will raise the bar for what the patient experiences, what the consumer experience expectations are.

Ambar Bhattacharyya: Um, because there, there, there’s no choice, right? We were just used to certain, certain things. Um, I think, unfortunately, healthcare is, is, it’s, there’s a myriad of different issues that have to be solved for there. I think if the, the, the, the word that gets used a lot is leapfrog, right. You know, can generative AI technology leapfrog an existing, uh, uh, technology that exists today?

Ambar Bhattacharyya: And I think if you Listen to a lot of the, um, [00:46:00] uh, early pitches from some of these scribing companies prior, uh, you know, payment integrity type companies. Uh, there’s like a short term plan and there’s a long term plan and in the long term plan, it’s always about the leapfrog, right? Um, it’s about, you know, how do we, you know, reinvent the system and really simplify it.

Ambar Bhattacharyya: There’s a lot of, there’s a lot between here and there. Like there’s no question in my mind. Um, but I, but for first time in at least my professional career, at least I think the probability of that is above zero, you know, and, you know, and, but if you would ask me like, you know, 30 months ago, what’s the probability that I’d say zero, you know, and how you, one can debate whether it’s 0.

Ambar Bhattacharyya: 1 or 10%. Um, Or 99%. I mean, you got to have different perspectives on it. But at least, at least, I think there’s a credible argument to say that [00:47:00] the patient experience in 2034 won’t be what you just described. Um, and, and I think, and, and that, that, that at least gives me some hope. It gives me some energy.

Ambar Bhattacharyya: Hopefully find these entrepreneurs, you know, find, you know, find, you know, find these early adopting customers, find these technologists, um, these doctors, nurses, um, that want to build towards that, um, because it wasn’t possible before. And I think now it’s possible. And so, you know, we should, we should get, we got to get to it.

Marcus: That’s, that’s, that’s a pretty, uh, optimistic note to leave out on. I mean, um, or anything else you want, you want to leave us with that? We didn’t ask you or we didn’t touch on before we, we wrap and let you go.

Ambar Bhattacharyya: No, listen, I think I mean, the only other thing that’s important as I think about all these different areas, we’ve talked a lot about a lot of wide ranging conversation, which, which is important.

Ambar Bhattacharyya: I think it’s the last thing I think about is that it’s really easy to think this is all one size fits all. And I think if there’s any 1 thing we’ve learned over the course of course of [00:48:00] history, much 5 years. Uh, the world is not one size fits all, and I think there there’s a there’s a relationship in health care.

Ambar Bhattacharyya: Um, there’s a, oh, there’s always a tendency to think of what is the company or the technology or the model or whatever. Um, and I tend to have a much more nuanced view, which is I think there may be, you know, you know, different strokes for different folks, you know, like, and you know, there may be something better.

Ambar Bhattacharyya: You mentioned elder care. There could be a phenomenal type solution for elder care versus for pediatrics, you know, just to call one out, right? Um, and it could be really, really different for, uh, specialties versus primary care. You know, you can just keep cutting, slicing and dicing in different ways. Uh, could be different for for virtual versus in person.

Ambar Bhattacharyya: You know, um, we haven’t talked about robotics, right? You know, and generate AI and robotics. Um, that’s another space exponential [00:49:00] type improvement where, you know, you’ve got open source models, you know, zero shot training, meaning like there’s no fine tuning able to program, you know, and video releases blog posts last week, you know, you know, it’s a robotic dog walking across the street on top of a yoga ball.

Ambar Bhattacharyya: Off is zero shot training. I mean, that means like it’s just they just put the model into the dog and the dogs are doing that. You know what? What is the implication that in health care as it relates to elder care 10 years from now, you know, um, as it relates to, you know, private duty home care type situations, taking a blood pressure medication.

Ambar Bhattacharyya: You know, our blood pressure cuff measurement in exactly the right way, taking a photo of that, sending it to someone or analyzing it themselves. These things just aren’t possible today, but like for the first time, we’re seeing the glimpses of this begin to happen. And that’s like, That’s what gives me hope.

Ambar Bhattacharyya: And I think it’s like it’s hope with like segmentation and with [00:50:00] nuance. Um, and so, you know, last but not like empathy, you know, empathy means something different for Vic, Marcus and me, right? Like, how do we personalize it to that level? Um, you know, and, you know, and to make sure it’s there. Um, And then like, can AI models be good at that?

Ambar Bhattacharyya: Can they be better at that? Or will they be worse at that? I mean, these are the open questions, but I’m hopeful that we as an industry, like tackle them head on.

Marcus: Amazing. Um, Bart, thank you so much. And I hope, you know, we, we, we have guests that are so important to our ongoing conversation and narrative that we do ask them to come back.

Marcus: You know, with the exponential rate of change in a year, it’ll be like a brave new world. So hopefully we can get you back on in a year and talk about, man, what a weird conversation we had a year ago. Cause look where we are today. Anytime, anytime. Awesome, man. Well, thank you so much, man.