Latest / CareTalk: Healthcare. Unfiltered. / Voice AI Is Expanding Access to Healthcare w/ Munjal Shah, Co-Founder & CEO, Hippocratic AI

Transcript

Voice AI Is Expanding Access to Healthcare w/ Munjal Shah, Co-Founder & CEO, Hippocratic AI · CareTalk: Healthcare. Unfiltered.

• 0:07
  Welcome to Care Talk, America's home for incisive debate on healthcare, business, and politics.
• 0:11
  I'm John Driscoll, the chairman of the UConn Health System, and today we have one of the absolute leaders in AI and healthcare, CEO, founder, multi-time founder, Munjal Shah, who's gonna-- who's the CEO of Hippocratic AI.
• 0:27
  W- We're gonna talk a little bit about his background, uh, where h- his founding story, and where we expect to go in the, in this, in this accelerating time of artificial intelligence in healthcare.
• 0:39
  Welcome, Munjal.
• 0:41
  Thank you for having me.
• 0:42
  I'm excited to be here.
• 0:44
  Well, tell me a little bit about your background story.
• 0:47
  I mean, how did you g- you were into AI before AI was on everybody's lips.
• 0:52
  I was.
• 0:53
  I did my graduate work in AI, um, and then built two AI companies.
• 0:58
  I'm what we call here in Silicon Valley a serial entrepreneur, so I love to build new things and build new companies.
• 1:04
  Um, and when I-- the second one I built was a machine learning computer vision company that, uh, we sold to Google.
• 1:12
  How did you get into AI?
• 1:13
  This was-- You were sort of building in AI at the end of AI winter, I mean, when, when there was a 20-year- Yeah… gap.
• 1:20
  I had actually ended up, uh, doing an internship in college, um, at a drug design company that was trying to design drugs using neural networks and what were called at the time genetic algorithms.
• 1:33
  And, um, and so I worked with, you know, some of the earliest neural network technology when, you know, today they're, they're gigantic, but back then they were tiny little models.
• 1:44
  Well, they were still pretty promising, but was that around H- HIV drugs?
• 1:48
  Was that the period where you were doing- Yeah.
• 1:50
  Actually, I, uh, did a, I was a, played a very small part.
• 1:53
  I don't wanna over-exaggerate this since I was an intern.
• 1:56
  But I played a very small part bringing, um, uh, a HIV protease inhibitor to market, um, and we built some of the simulation software that would try to predict whether a new drug candidate would bind strongly with HIV protei- uh, you know, to the target.
• 2:13
  And, um, uh, and that eventually got resulted in a drug that did come to market.
• 2:19
  Which is kind of amazing.
• 2:21
  Yeah.
• 2:21
  And, and, and, and, and now all of the large LLM companies, m-many of the large most best funded companies are focused on AI and drug development.
• 2:31
  It, it's, it'll be interesting to see how that plays out.
• 2:34
  But then you ended up founding a couple of different companies and- Mm-hmm … and, and in, in the tech world.
• 2:40
  What got you into healthcare and why?
• 2:43
  So the day I sold my company to Google, the very next day, so, you know, it was kind of the best day of my work career to that point.
• 2:53
  I was high-fiving with my wife and, um, and then the next day I was running, um, and I ended up with chest pains and ended up in the ER.
• 3:00
  So I was 37 at the time.
• 3:03
  My dad had had a heart attack in his mid-40s, so my genes weren't exactly on my side.
• 3:08
  But, um, I was about 40 pounds heavier than I am today.
• 3:12
  I'm, I'm not a very tall person.
• 3:13
  I'm about 5'6", so, you know, 30, 40 pounds was a lot.
• 3:18
  Um, and, uh, and then I just decided to take control of my health and I, I lost, you know, lost quite a bit of weight.
• 3:26
  Um, learned a ton about nutrition and diet and, um, and even took a class in endocrinology at, at… I, I lived not far from Stanford and I took one of the extension classes and I loved it.
• 3:39
  I thought it was so interesting.
• 3:41
  I said… In fact, I remember my wife was like, "They gave you eight hours of homework the first class?" You're kind of a hyper-geek is what we're, we're- … you're articulating.
• 3:49
  I, I, I just loved learning about it.
• 3:51
  It was just so interesting.
• 3:53
  And so, you know, she's like, "They gave you eight hours of homework?" I'm like,"No, I'm, I'm eight chapters ahead." Like, "This was just really interesting.
• 3:59
  I just kept turning the page." And so, um, you know, I realized that there was another whole dimension of passion that I had.
• 4:08
  Um, and maybe if I had done that first, I would have got into medicine instead of computer science, but- We're lucky you got into computer science.
• 4:15
  But you chose to go the enterprise route versus consumer.
• 4:19
  Tell us a little bit about w- uh, the founding story of Hippocratic AI.
• 4:23
  So now you're motivated to get into healthcare.
• 4:26
  Yeah.
• 4:27
  Why, why, why this direction and, and, and how did you come to found it?
• 4:32
  You know, so, um, I then spent, uh, 10 years building a company focused on kinda consumer healthcare, but it was a healthcare healthcare company.
• 4:42
  And I think I did not realize what a difference it is to build a software company versus building a healthcare company.
• 4:49
  Man, is healthcare hard.
• 4:51
  Like, hats off to anybody building a pure healthcare company.
• 4:55
  And that venture did not end up nearly as successful as my first two had.
• 5:01
  Um, but when that ended, ChatGPT had just come out.
• 5:07
  And I looked at it and said, "This is the AI I've been waiting for my whole life." Like, this thing finally works.
• 5:12
  This sort of ChatGPT-3 kind of period?
• 5:15
  Yeah.
• 5:15
  But the ChatGPT-3, you know, on where-- This is if you remember Thanksgiving weekend- Yes… of 2022 is when it came out.
• 5:23
  They opened it up.
• 5:24
  And I, I looked at that and I was like,"Oh my God," like, "AI finally works."'Cause before that it was very brittle.
• 5:30
  It worked if you only used it in a certain way and only trained it in a certain way, and now you really had general intelligence.
• 5:38
  And I said, "You know what? I want, I want another go at this." Like, I didn't achieve my mission of really helping using technology to improve healthcare.
• 5:48
  I'm very passionate about it.
• 5:51
  Let's, let's go and do that.
• 5:53
  But did you really think that was general intelligence or usef- a useful tool?
• 5:57
  You know, I think that one of the most important parts about building companies in technology is that a lot of people miss is they look at the technology today and then make a decision, versus you have to kinda skate to where you think the puck's gonna go.
• 6:14
  And when I saw it, I said, "Whoa, this is gonna continue to progress.
• 6:18
  We-- These models are gonna get smarter, and, um, this is just the beginning." And I had, you know, because of the deep work I had done in neural networks before, I knew kind of the underlying properties, and I was like, "Yes, this trajectory will continue to get smarter with more data and with larger models and with more parameters." And I said, "You know, I, I can see where we might end up," and, and that's worth kind of embarking on this journey.
• 6:42
  And so then we started Hippocratic, uh, really in the first quarter of '23, um, and have been building the company ever since.
• 6:52
  So what was the original vision of Hippocratic?
• 6:56
  Not quite what we do today, but not that different either.
• 6:59
  Um, we just set out to build an LLM for healthcare.
• 7:04
  I kid you not.
• 7:05
  Like, the, the first deck was, "Somebody should train an LLM for healthcare." Um, and, and then we quickly realized that we didn't just wanna train it.
• 7:16
  We wanted That really we should skip a step and move to the agentic world.
• 7:21
  And remember, this is '23.
• 7:24
  Everybody else was talking about copilots, and we said, "No, you won't get enough leverage unless you build an autopilot." So the LLMs are really the large models that sort of can contextualize, answer, and explain everything theoretically today that around which the hyperscalers are building.
• 7:43
  And that would've been quite a, a step, but the next step, I think na- narrowing that would be to build that just for healthcare.
• 7:53
  And instead, you're no longer… You're already thinking ahead for what people are talking about today- Mm … where you create, you know, sort of endless amounts of low-cost agents that do stuff.
• 8:06
  Do you wanna just explain what that agentic step means?
• 8:10
  Because you're, you're going from sort of a, a, a call and response Google-like- Mm-hmm on steroids engine to sort of support, advice, extension- Mm-hmm … augmentation.
• 8:23
  Do you wanna just explain that a little bit so that just folks c- uh, folks who aren't necessarily have your kind of a background can understand- Yeah, yeah… what that could mean?
• 8:32
  So, um, these healthcare agents largely call patients on the phone and talk them through their care instructions, make sure they've gotten their refills, um, you know, help verify that they've stopped their medications before their upcoming surgery.
• 8:50
  Like, they operate within the scope of a registered nurse, w- and registered nurses don't prescribe and they don't diagnose.
• 8:57
  So there was this Goldilocks zone that we picked.
• 9:01
  We said, "You know, if you build an AI doctor, you're probably gonna kill somebody." And like, I was like, "I, I don't-" That's not reassuring, Munjal.
• 9:09
  No, I st- I still think you will.
• 9:10
  I, I do not think the technology's safe enough to truly diagnose.
• 9:14
  It, it… Down the middle, it's fine, but what happens is medicine is a giant game of corner cases and exceptions, and when you hit one of those, that's where you have a very high risk.
• 9:25
  But I think, on the other hand, we didn't just wanna do admin agents that are just, you know, gathering your insurance information and, and doing simple things like that, because that's not gonna change healthcare, right?
• 9:38
  Like, that, that's gonna make things a little more efficient, but it's not truly going to improve healthcare.
• 9:43
  And so we said we wanna do something clinical, but we don't wanna diagnose.
• 9:46
  And there's this nice Goldilocks zone in the middle, which is the scope of an RN, where you're doing care activities.
• 9:52
  You have to do them safely, um, but you are really helping people to manage their condition.
• 9:57
  And if you look at the healthcare burden in this country Like it's diagnosing our l- our large chronic diseases dominate that burden.
• 10:07
  Diagnosing them's not that hard, actually.
• 10:10
  Like, "Oh, you have diabetes.
• 10:11
  What's your A1C?
• 10:12
  Oh, okay, it's an eight.
• 10:13
  Okay, you have diabetes." Like, you know, "Oh, w- what's your blood pressure reading?
• 10:17
  Oh, okay, you have high blood pressure." Like the diagnosing is not the hard part.
• 10:21
  The managing is the hard part. And so managing your diabetes, managing your high blood pressure is where the difficulty comes in, and we said, "You know what?
• 10:29
  Let's build agents that help patients do that." Now, one last part that I'll highlight is we then realized that there's going to be a voice renaissance That the idea that these agents just chat with you over text is actually where digital health went wrong.
• 10:44
  Because in healthcare, reach is not our goal.
• 10:48
  Persuasion is our goal.
• 10:50
  And text messages and apps are not that persuasive because you just ignore them when you don't.
• 10:56
  But a voice conversation is w- more persuasive.
• 10:59
  And of course, a face-to-face is even more persuasive.
• 11:01
  But we moved everything from the most persuasive channels to the least persuasive channels, and that this was the opportunity we had now with voice AI was to bring that back.
• 11:12
  In just contextualizing this, you're selling these agentic nurses to enterprises, healthcare companies like Cleveland Clinic and others.
• 11:23
  Um, just to contextualize that, I think, like something like a third or half of all nurses wanna quit 'cause they're burnt out and close to 20 to 30% of doctors.
• 11:32
  And so the-- what you're also doing, I think, Munjal, is, is, is helping leverage the nurse base and the doc base that we've got today to actually enhance the amount of labor or, or, or patients we can see and support, uh, if you're successful at a time when we really do have a significant labor, a demo-- a labor s- shortage because of post-COVID burnout and a n- a, a looming labor shortage based on the mismatch of how many doctors and nurses we're educating and how many people are getting old.
• 12:06
  Yeah.
• 12:06
  So, so first thing there, um, we do not call them agentic nurses.
• 12:10
  The word nurse is a protected term, and even spiritually, I, I think there is something special about that term that should be reserved for humans and the care that they can provide.
• 12:20
  Um, we do call them agentic care managers and other things like that to differentiate, um, and still communicate the concept of providing care management.
• 12:30
  Um, second, I mean, you're correct.
• 12:33
  You know, we are short so many clinical staff in all of healthcare.
• 12:40
  The problem's only getting worse.
• 12:42
  The number of folks in school does not even equal the number that are gonna be retiring in the case of nursing.
• 12:48
  So the problem's, A, getting worse because of, you know, the rate of graduations to retirement.
• 12:53
  It's getting worse because the aging population's gonna increase demand, and a 70-year-old doesn't go to twice the number of doctor's appointments as a 35-year-old.
• 13:02
  They go to 10 times the number.
• 13:04
  So the healthcare intensivity is so high and about to get a lot higher with the aging population.
• 13:09
  So, so with all of that, why has it sometimes been a challenge for enterprises?
• 13:16
  I know you've had s- you've had phenomenal sales and success with the large enterprises, but sometimes nurses have pushed back Folks have sort of been a-a-at, not just for you, but for any AI, there's a, there's a, there's a, a fear there.
• 13:30
  How have you-- What have you seen, and how have you dealt with it?
• 13:33
  The most important part is we're all here to support patients, and if patients aren't getting the attention they need, they're not going to get the care they need.
• 13:45
  And I think that everybody intuitively understands this, and we're seeing it.
• 13:50
  And so actually, you know, we've signed on so many health systems, almost now over 50 clients, um, that are typically health systems over two billion and up.
• 14:02
  And everybody gets that this really is about extending care and expanding care, and that we need to do that desperately.
• 14:10
  And so, you know, there's pockets of conversation about it.
• 14:13
  But I will tell you in, in all of mine with senior healthcare leaders and many of our clients and others is everybody gets that somebody needs to be there.
• 14:22
  You know, my mom was diagnosed with high blood pressure five, six years ago.
• 14:26
  She did not take her meds because she-- it made her dizzy.
• 14:30
  They made her get, you know, have all these other side effects she didn't love.
• 14:35
  She s- didn't keep taking them.
• 14:37
  Fast-forward five years, we end up with two twenty blood pressure one night in the ER, and she now has the beginnings of congestive heart failure And I thought to myself, you know, she didn't tell me, of course.
• 14:49
  She said, "Hey, um, the, I saw the doc and he said I'm the healthiest…" You know, at the time she was 77 year old, she's now 83, "… that he knows." And so I couldn't help her.
• 14:59
  Um, but I know her doc knew that she wasn't taking her meds 'cause she didn't get a refill for five years, right?
• 15:07
  But the system isn't set up for him to reach out to her.
• 15:10
  Maybe once, "Ma'am, you didn't do your refill," but not continuously bring her in, titrate her medications, try to change to getting a f- a, a formulation that works for her.
• 15:21
  What the promise of an agentic support for the humans can be is s- building some intelligence and some learning and personalization for your mother, my- Mm-hmm my 93-year-old mother, who's, you know, struggles with a bunch of different things.
• 15:35
  Y- the, the, the thing that, the great thing about the, the promise of AI, I guess for me in what you're building, is that it can learn and personalize to the patient and their challenges, you know, social, economic, as well as healthcare, and cultural, which is harder- Mm-hmm I think, to scale in an environment where we cut off immigration and we limit the number of docs and nurses.
• 15:56
  Um, maybe talk, walk us through a little bit of y- what a, an enterprise relationship is like.
• 16:01
  You've got some amazing hospitals, Northwestern, Ochsner, Cleveland.
• 16:06
  Y- when you implement your system, what happens to a nurse at the Cleveland Clinic in Ohio?
• 16:12
  I'll give you an example of, um, a couple of different products.
• 16:15
  But one, what, what we have found is, so there's two ways you can apply this.
• 16:21
  You can do things you're doing today and offload work, or you can do things that you should be doing but you never thought to do until you have kind of an infinite supply of, of clinical labor.
• 16:34
  When I look at those two things, the second group is actually where things are really succeeding.
• 16:40
  So things you never thought to do today.
• 16:43
  So I'll give you an example of something we did for a health system.
• 16:47
  They said, "You know what?
• 16:48
  We're onboarding you.
• 16:49
  W- we wanna do this, um, quick and easy while we're waiting for the full integration to happen.
• 16:54
  Um, we will do the scheduling integration first.
• 16:57
  Um, and then can we give you 1,700 patients lost to follow-up for low-dose CT scans for lung cancer?" Great.
• 17:02
  I'm like, "Well, what, what happened?
• 17:07
  You tried." They're like, "We tried.
• 17:08
  It's lost to follow-up." That means we sent a text message, we sent an email, we sent a physical letter, and they still didn't schedule this low-dose CT scan.
• 17:18
  And so the AI got on.
• 17:20
  The AI just persuaded, not all of the 1,700, but they were gonna be at zero, so frankly anything we get is upside.
• 17:27
  And we got 252 to schedule.
• 17:30
  We got 180 to actually show up and do the scan.
• 17:33
  Um, we found nine significant findings and three cancers.
• 17:38
  And so hopefully we saved three lives in that.
• 17:41
  In fact, one of those cancers I heard we found so early they only needed to do a resection.
• 17:46
  They didn't need to do chemo or radiation.
• 17:49
  So that's amazing.
• 17:51
  But the most important thing just i- is that you're f- I- is that you're, you are bringing h- healthcare, uh, sort of the, um, the exiles of the system who would otherwise fail health back into a system that- That's right can help them.
• 18:06
  And, and when you listen to the call, you know, the… There's this little old lady who gets on, and the first thing she says when the AI's like, "Hey, I'm calling to talk to you about your low-dose CT scan for, for potential lung cancer risk," she's like, "I feel fine.
• 18:19
  I don't need it." Sounds like my mom.
• 18:20
  She literally says those words.
• 18:22
  Yeah, and this AI's like, "Ma'am, um, you, you, you wouldn't feel lung cancer until it's too late." She's like, "Oh, really?" "Yeah, ma'am.
• 18:30
  That, that's how lung cancer works." Um, and then she was like,"But I don't like that machine.
• 18:35
  It made me really claustrophobic," you know, the CT scan.
• 18:38
  And it's like, "Ma'am, we have a new machine now that's more open. You might find that one okay." And there are.
• 18:43
  There's these new, you know, larger CT machines, op- open MRIs and things.
• 18:48
  And so anyway, so, like, it went through her specific issues.
• 18:53
  So you've got… You've had 250 million-ish conversations now with patients?
• 18:58
  Yeah.
• 18:59
  I think we're at… I just saw on our little dashboard this morning when I walked in the office, 279 million, uh, patient interactions.
• 19:05
  What's, what's the big- Mm-hmm … what's the biggest surprise in those interactions or things that you learned that you didn't expect to learn or, or saw that you, or that the systems and the doctors didn't expect to see?
• 19:16
  I think it's, it's this notion of the abundance use cases, that so much of what… Like, h- if you think about healthcare, it's always we've designed it premised on clinical scarcity Everything about healthcare assumes scarcity.
• 19:33
  Word triage assumes scarcity, right?
• 19:37
  Population health always assumes scarcity because you don't really tackle the whole population.
• 19:42
  You do risk stratification which is I pick a sliver By scarcity, I think you mean everything is tied to a budget.
• 19:50
  A budget has limits.
• 19:52
  A limit is what, what defines what you prioritize and what you can and can't do.
• 19:56
  A- and, and just physical staffing.
• 19:58
  Like, there's only so many people.
• 20:00
  Like, I'll give you an example.
• 20:02
  Um, in the heatwave in New York about two and two, two and a half months ago now, we called 50,000 people at the hottest two hours of the day every single day of a heatwave and did a heatwave education, um, about how they could stay cool and avoid ending up in the ER.
• 20:17
  Like, you can't get enough people at a five-day notice of a coming heatwave to call 50,000 people every single day.
• 20:26
  Even if you have the money, you won't find the staffing.
• 20:28
  And so this is the big idea is that we can rethink healthcare assuming abundance.
• 20:36
  And healthcare is one of the few areas that can absorb that abundance'cause the ideal staffing is kind of one clinician and one person.
• 20:42
  You know, only education and healthcare have this property'cause ideal staffing in education is one teacher to one student.
• 20:48
  And so we can really use AI to do things that we don't do today or we gave up on because, to your point, we have limited staffing, um, and limited budgets for that.
• 21:00
  But also we have just limited labor availability, especially for the surge use cases like the heatwave.
• 21:07
  So you make a compelling point about how AI can help us not just bend the cost curve, but, but actually engage people and, and more effectively.
• 21:16
  Th- that's very powerful.
• 21:18
  What are people getting wrong about their expectations about AI in healthcare?
• 21:23
  You're in the center of this.
• 21:25
  You've talked about the abundance- Yeah … hypothesis.
• 21:28
  That's really fascinating.
• 21:30
  Wh- what else, what else are people missing about the promise of AI right now in healthcare?
• 21:35
  Most of the use cases that try to do what people are doing today cheaper are not working.
• 21:40
  This is why I think the labor replacement thesis is false It, I can see why people are nervous about it, but I think that they haven't understood the secondary impacts of it or complexities of it.
• 21:53
  And when you finally do them, we've rolled out hundreds and hundreds of use cases now.
• 21:57
  We've seen all kinds of things, and it is much-- It's not impossible.
• 22:02
  It, you, you can actually roll it out for, for some of the labor savings elements, but man, is it hard.
• 22:08
  And it's hard for-- because it's, it's very glass half empty.
• 22:12
  If you don't do it just as good as the human, it's like, well, you know, it's not as good.
• 22:17
  Let's not roll it out.
• 22:18
  But when you're doing something that we don't do any today, then like the low-dose CT scan example, it's, it's upside.
• 22:27
  I didn't get all 1,700, but the humans didn't get all 1,700 either.
• 22:31
  In fact, those were the ones the humans couldn't get.
• 22:33
  And so there are so many… Everybody thinks you should take the things you're doing today, try to save some money, get some efficiency out of them.
• 22:43
  It's actually significantly harder to use AI in that way compared to using it in these, uh, new emergent use cases.
• 22:50
  You've raised a titanic amount of money.
• 22:52
  I think my last te- Well, my last number was, like, four hun- over $400 million in capital.
• 22:58
  You've, you've scaled pretty effectively, and yet the hyperscalers, the large LLMs, y- oh, whenever they talk, they sort of mention healthcare in the con- without really defining it.
• 23:10
  I, I think the most recent interview- Yeah … with Greg Brockman I saw, again, he threw healthcare in as, as almost like a, a, you know, p- potentially an enterprise strategy for hospitals that OpenAI is working on.
• 23:23
  How do you compete in a world where the Anthropics and the OpenAIs are, have infinite money and they appear to, although they don't necessarily show up in most pitches and, and, and, and programs, they talk a lot about getting into healthcare.
• 23:37
  How have you differentiated and, and what's sustainable for Hippocratic in that world?
• 23:42
  We picked the hardest problem.
• 23:44
  We said we're gonna go after clinical patient-facing conversations.
• 23:48
  So there's three things in that: clinical, patient-facing, and voice, i.e., the conversation part.
• 23:55
  So there's three things in that: And each of those has provided a different need.
• 23:58
  So there's three things in that: First, on the clinical side, you've gotta be the safest.
• 24:02
  So there's three things in that: So, you know, let's say I'm building voice AI for your Taco Bell drive-through, and let's say I'm 2% better than the OpenAI version, but they're half the cost.
• 24:10
  So there's three things in that: I lose every time then.
• 24:13
  So there's three things in that: But let's say I'm building clinical AI and I'm 2% better, and they're still half the cost.
• 24:19
  So there's three things in that: Nobody buys it because you can't tell your boss, "Oh, I saved half the money, but I, I killed 2% more people." It's a bad day at the office.
• 24:28
  So there's three things in that: Like, you can't.
• 24:28
  So there's three things in that: It's a bad day at the office.
• 24:29
  So there's three things in that: So incrementalism matters in clinical safety, and we built safety system after safety system after safety system and got every last corner case and made sure they're right on every last agentic call.
• 24:43
  So there's three things in that: And there's a l-- And we're the only ones that do massive output testing where we have 7,000 clinicians calling the AI.
• 24:50
  So there's three things in that: Now they've called it over 800,000 times, marking every error it makes, and we do not sh- And then one out of 50 times they get a human clinician on the other line and mark the errors they make, and we don't ship till we're safer than the human.
• 25:02
  So there's three things in that: We don't use one LLM.
• 25:03
  So there's three things in that: We use 31 LLMs where the other 30 are supervising the main one to make sure it's not doing anything wrong.
• 25:09
  So there's three things in that: Like we, we built so many redundant safety systems and went after that hard problem, and that's what healthcare wants.
• 25:17
  So there's three things in that: It wants a system that is truly safe.
• 25:20
  So there's three things in that: Um, there's also unique things we've done on the voice side, um, that ensure safety there, which, uh, you know, are more technical.
• 25:28
  So there's three things in that: But we've ensured that we can deliver the best models as fast as we can.
• 25:33
  So there's three things in that: And the, the analogy I always draw here is OpenAI is like General Motors.
• 25:37
  So there's three things in that: They make all kinds of cars.
• 25:39
  So there's three things in that: They make trucks, light trucks, and they make, you know, convertibles and they make SUVs.
• 25:43
  So there's three things in that: Um, but despite General Motors being there, there's still room for a Caterpillar.
• 25:50
  So there's three things in that: Why?
• 25:50
  So there's three things in that: Because Caterpillar's design requirements need a level of reliability and, you know, kind of my analogy to safety that, that is second to none and makes no sense to put into a normal car because you'd over-engineer it and the car would cost too much.
• 26:06
  So there's three things in that: But this is where we've realized that there are going to be segments where things like safety matter so much that you can build a specialized product.
• 26:16
  So there's three things in that: Well, I was gonna ask you a final question around why we should trust Hippocratic and w- how you convince people, but I think you just wrapped with that answer in the last question.
• 26:26
  So there's three things in that: So, uh, Munjal, it's been an absolute pleasure.
• 26:30
  So there's three things in that: Thank you for joining us today.
• 26:31
  So there's three things in that: Yeah.
• 26:32
  So there's three things in that: Thank you for having me.
• 26:33
  So there's three things in that: If you like what you heard or you didn't, we'd love you to subscribe on your favorite service.
• 26:38
  So there's three things in that: I'm John Driscoll, the chairman of UConn Health.
• 26:40
  So there's three things in that: Thanks for listening.