Can AI Replace Primary Care?

And there's 100 million people in this country that don't have a relationship with a doctor. And that was the earliest motivation behind counsel is how could we create a care model that was built on top of modern technology that had the ability to create a step change improvement in the amount of clinical supply available in the world.

I think the reality is that becoming a doctor is really important for a handful of things. So if I had cancer I would probably like to be seen by an oncologist. I totally believe in a world five to ten years from now where a minority of primary care is rendered by an md.

Welcome to the venture capital podcast. Today Peter Harris and I are joined by Rob Biederman and Dr. Muthu Alagappan for quick intros. Rob Biederman is the managing partner at Asymmetric Capital Partners. Rob is a conviction led investor known for backing founders transforming big complex industries. He's developed a strong thesis around the future of data driven healthcare and we're excited to dig into why he invested in Council. Also with us is Dr. Muthu Alagappan, CEO and co founder of Council Health. Muthu is a physician building one of the leading AI driven healthcare platforms focused on delivering AI enabled physician supervised virtual care through structured data. He's pioneering the idea of care at the cost of compute and we're excited to explore that vision today. So welcome to the podcast.

Yeah, thank you guys. Excited to be here.

Excited, thanks.

Do you want to start off by kicking it off and giving us a deeper dive of what your product is? I was playing with it before the podcast. It reminds me a lot of chatgpt, but I can order labs, I can actually connect with a physician.

Yeah, definitely. Happy to jump in to an overview on Council and excited to be here. Thanks for having me. Maybe to start with, the problem sort of we're tackling at counsel is that supply demand mismatches in healthcare have gotten worse than they've ever been before. So we're probably short on average 50, 100,000 clinicians worldwide. And that number is only getting bigger. And there's 100 million people in this country that don't have a relationship with a doctor. And most of the either AI or process improvements we've made in healthcare have probably increased our capacity to care by 10% or 15%. But when you're facing a supply demand mismatch as big as we are today, you really need step change improvements in clinical supply. You need a way to essentially multiply providers by an order of magnitude. And that was the earliest motivation behind counsel is how could we create a care model that was built on top of modern technology that had the ability to create a step change improvement the amount of clinical supply available in the world. Sort of a thought experiment for this is what would it look like if the best doctor who'd ever lived was available to take care of you and you alone at all times and could we productize that experience? Right. Could we essentially teach a hybrid of AI and human doctors to deliver care in such a way that it felt like you in your pocket at all times had access to the best doctor ever? So that was the high level motivation and what inspired us. What Council is today is we're an AI native virtual care company where members, patients can message us to access care at any time and that care is delivered via a hybrid of medical AI that we have sort of improved and built specifically for this purpose alongside a real physician group. So we pair together AI with human doctors in a way that feels like you get the best of both worlds. You get the intelligence and the accuracy and the instant response of an AI, but then you get the safety, the trust, the oversight of rail doctors. And that's how the platform works today. And happy to go into more detail. We've been really excited about it.

Awesome, thanks for sharing. Rob, do you want to take a moment and explain why you invested in Muthu and Council?

Of course. So look at the beginning of every early stage investment, you have to partner with somebody who's incredibly special and whatever industry you're in, whether you're in software or healthcare, hard tech or deep tech, you want somebody who has great founder market fit. And I think from the moment we met Muthu, it was very obvious that he had great mastery of the space and also great mastery of what he wanted to build. So check that off. Muthu is very talented. I think echoing all the things that Muthu said, the practice of medicine has been, I would say more or less unaffected by in the main, a lot of the extremely large data sets that we can sort of now create the intersection of electronic medical records and prescribing and sort of diagnosing. So I think one of my colleagues and I saw was an opportunity to really transform how primary care is delivered in the US and make it far more data driven, make it far more dynamic, make it something that hopefully has costs that allow us to offer it not only to underserved populations in the US but potentially outside of the US where good medical care is much harder to come by. I think we saw the person in a unusual combination of basically AI research and understanding with actually, you know, a practicing physician that felt really special to us. So within our portfolio, I don't know that we've ever had a no brainer that was quite this obvious and obviously everything that's happened in the last two years has really I think borne that out.

It's an amazing product to play with. Peter, have you invested or looked at similar opinions space? Similar space.

I mean, yeah. So I'm curious, you know we've looked at a bunch of the kind of like, I don't, I don't know how you describe them, diagnostic blood test but then they also, their primary business model is really selling DLP and testosterone drugs. Right. So I'm curious, how do you guys think about those businesses and the overall like landscape if you will. I mean I feel like longevity has really taken off over the last three or four years as a major trend and, and clearly one that is benefiting everything that you're building at council but, but feels like you guys are maybe a little bit more on the, the traditional medicine side versus the more kind of longevity side which I think in many cases is almost anti. Yeah. Medical industry. Does that make sense?

Yeah, I'm happy to jump in there. I mean I think we're seeing a lot of really exciting trends in healthcare and almost all of them are accretive and valuable to the end user and it's been I think fascinating to see. I think one thing to call out from kind of from the get go is that healthcare is such a massive industry. $5 trillion in 2023. The demand for healthcare is sort of unlimited. I mean it's one of the few intrinsic things about being human is that you will need healthcare in your life and sort of that demand for it will really never go away. I mean we can live twice as long as we live today and if anything that's just going to mean we spend more on healthcare over the course of 100 plus year lifespan and eventually everyone will succumb to some illness which will be very expensive and will require really good care. And so for that reason I think this is a space where it's definitely not zero sum. I think there's going to be a lot of really big winners who invest in a lot of adjacent categories, whether that's on the pharmaceutical side, the diagnostic side, the actual care delivery. And so we're excited about that. I think one of the trends we're seeing in recent years is consumers increasingly taking their hand, their Health into their own hands. So we sort of call it agency. And a lot of companies are pushing forward. This movement of healthcare no longer needs to be gated by the medical system that is serving as a gatekeeper to your ability to access care. And so companies like Function Health have done a phenomenal job of saying, what if we actually put the power back in your hands to take responsibility for your health? Where no longer do you need to schedule a visit and convince your PCP that you actually need to be tested for your iron levels? You can go and get 100 plus lab tests on your own accord. I think these movements are excellent from an agency perspective. And a lot of what we're doing on the care delivery side at council takes a similar agency first approach to care. What if you as the patient get to choose when it's an AI, when it's a doctor? And it's not up to us to sort of move those entities back and forth, it's sort of up to you. And there may be times where you want the anonymity of an AI and times where you want the oversight of a physician. What if you get to come here and ask for a refill or to manage a condition and you can sort of check in as much as you want, not only when the doctor tells you that it's time to check in. And so I think that lens on healthcare has sort of been applied across the industry and certainly one we're excited about.

Yeah, yeah, it makes, makes a ton of sense. I'm curious, why do you feel like now is the time that, that people are taking control of their own health care? Like why now? Right, why not before? And I feel like we were starting to see this occur well before AI really took off. So I don't even know that it's necessarily correlated with AI. I think AI is maybe accelerating some of it. But I'm curious to get kind of your perspective on like where. What's causing those shifts?

Hey, I'm happy to share my perspective and Rob probably has a good take here as well. I think there's two things that have happened in healthcare broadly in the last decade which has probably incited this evolution of consumers taking control of their own health. Number one is just the way health plan deductibles have gone in the last decade where there's been an increased prominence of high deductible health plans, which is effectively a form of cost sharing that puts the onus on the consumer to have to consume healthcare in more of a price conscious way. And so when you're on the hook for paying for healthcare, I think consumers suddenly are much more. They're much more intent on who they're getting the care from. Is it high quality? They're sort of shopping for it the way you would shop for any other service, where you're comparing cost and quality and convenience. And so I think maybe an unintended effect of high deductible plans has been consumers increasingly just sort of making their own decisions that they feel are best. I think the second trend is probably an increasing distrust or even distaste in traditional health care. I think I grew up in the 90s, but I think in previous decades, there was this feeling of invincibility, almost a halo effect around doctors where you took their word as gospel and you didn't challenge it, and it became really hard. There's a big information asymmetry, and physicians were on a pedestal. I think what the Internet did is it made information consumable by everyone. And so now you could actually fact check your physician, and you start to expose that the people that you're entrusting your health with are actually not invincible. They're not infinitely knowledgeable. They're forgetting all types of things that they probably shouldn't forget. They're not as up to date on the latest evidence as they should be. And so that invincibility cloak sort of floats away. And I think consumers started to see physicians as being maybe more fallible. And the Internet made it so that anyone could come into a doctor's appointment and perhaps know more about that specific condition they have than, you know, any physician they're meeting with. And that distrust in physicians has probably led people to say, hey, maybe I can do this better on my own, or maybe I should go to an online platform that specializes in this one condition and get the best care possible. And again, I don't want to. I don't want to overstep. It's not that there's this massive distrust in the system, but I think there's seeds of distrust that has probably led people to want to. Want to take health into their own hands.

I think that's also. Oh, go ahead, Rob.

Oh, no, no. Go for it, Peter. I'll go after you.

I was just going to say, I. I think your perspective on high deductible plans, the rise of HSAs, is. Is a really interesting one around people. Yeah. Taking kind of more control of their health care spend. Yeah. I think there's also this element of. From a trust perspective. So many. So many times we're basically told that medicine, traditional medicine is all reactionary. Right. It doesn't really step in until you're sick, till you've broken your arm, whatever it might be. And I think there's a lot more interest in preventative medicine. Right. Especially probably from your point of like, well, if I'm going to have to spend a lot of money to fix this thing later on, maybe I care about preventing it today. Right. In a way that I wasn't thinking about that before. So yeah, I think that's really interesting. What were your thoughts, Rob?

Yeah, the broader theme, and we're certainly investing against this in so many different ways at asymmetric, first and foremost with Council is we're moving from a care paradigm, at least for primary care, that was very much one to one, non data driven. A MD providing primary care through a singular annual physical and then spot visits, you know, for, you know, specific issues. If you look at the entire landscape and that's the care paradigm we all grew up with, if you look at today, there are so many other care paradigms that are not physical, human with an MD providing care, and that's urgent care, which is, you know, to a very large extent NPS and pas, it is companies like Council that are not providing human care. So if you start with the assumption that we had in the 1980s growing up that medical care was rendered by a singular human being with a md, you map to today you have non human care, you have non MD care, you have geographically dispersed care. I think the reality is that becoming a doctor is really important for a handful of things. So if I had cancer, I would probably like to be seen by an oncologist. If I need a, if I break my leg, I would probably like an orthopedic surgeon to be heavily involved. If I have a sore throat, it's not really clear that I need a again human MD to assist with my sore throat because that is something that's clearly not at the top of license. So we had a health care system and that all kind of worked well when the number of doctors and the number of patients was reasonably well balanced and people's candid expectations for the health care system were a little bit lower. Today, combination of kind of out of control costs plus physician shortage is creating this world where lots of other care paradigms that we probably all wouldn't have been that open to in 1987 or the technology wouldn't have worked for are now actually becoming very interesting. And then you can take people who actually go through medical school and residency and fellowship and all that. And you can have them working on things that clearly fall outside of solvable by either a computer or by a nurse practitioner. And I think that's maybe the most exciting thing. And it probably actually makes doctors a lot happier because the reality is they don't also want to be in the primary care business. By and large, being a PCP is not the most interesting thing you can do coming out of medical school. And I think it's actually going to be increasingly rare. And I could totally believe in a world five to ten years from now where a minority of primary care is rendered by an md.

Yeah, no, I think that's an interesting perspective. I mean, because really the primary care doc is kind of quarterbacks the health more than anything else. Right. It's like, okay, let's figure out what the issue is and then send you off to the specialist. And I think that's a really interesting place for a company like Council to step in and fill that role. Right. I also think, well, I wonder how much of another driver is just social and. And how like today, because so much is like visual. Right. And online, we're constantly consuming pictures and videos and image and content of people that are idolized for, frankly, their health. And so there's this concept of, like, health as wealth that I think has never really existed to the extent it exists today. And so I wonder if that's also like, a decent part of this. And then you layer on top of that the Ubermans and Atias of the world who are pushing a lot of content around longevity and health and preventative medicine that we didn't really get before. Right. And all of that feels like it's kind of come in the last three or four years.

Yeah. I think the social media effect is undeniable. I think there's always this element of mimetic following of people that are idolized in this space. And I think the amount we know about longevity has probably gone up exponentially in the last five years. In terms of supplement science and the things that actually make a difference, there's probably an element of this, just like Maslow's hierarchy as well, where there's some subset of people that have earned the ability to now focus on longevity because they're not having to be worried about other kind of more important foundational aspects of health and safety. And so as a former pcp, a lot of my patients had advanced diabetes and arthritis. And, like, the idea of, like, how do I live forever? Is not crossing their mind. It's like, how do I live for another five years? And you know, when you start off with a blank slate and you're quite healthy, that can be something you actually focus on quite a bit. And the last thing I'd say just on, on the longevity preventive care side is I think a lot of people do frame the US healthcare system as being very sick care, not sort of well care, and being very reactive. And in a brief moment of defense of the healthcare system, there is quite a bit of preventive care that PCPs do. I mean, everything from screening mammograms to annual physicals to vaccinations are all a form of preventive care. And I've counseled thousands of patients on diet and exercise as a pcp. But I think what software does is it allows you to take the best version of something and democratize it everywhere. So if Peter Attia is better at counseling on lifestyle than I am as a pcp, then suddenly he earns the right to do that for everyone through his book and podcasts. And same with Huberman. And so I think the interesting thing is not that PCPs aren't doing this, it's just that software makes it so that whoever does it best, in this case some of these key influencers, they have now the ability to then do that for everyone, which is again, it's a good thing, but not for a lack of trying. I think on the PCP front, just more so.

You know, I'm curious, since you guys are so data driven, are there unique data insights that you've found, you know, in terms of the demographic that the council really works well for or that finds it really interesting and signs up for, does that make sense? Like I'm part of, part of the reason I'm asking is I'm wondering if, you know, a younger population is more willing to engage with this, more interested in engaging with this, or if it's really spread across all generations?

Yeah. I think briefly on the data piece, we view data as a really important input to the type of care that's being delivered. And if you think of care as just sort of being generic, the more data you have, the more personalized that care can become and the more that care varies in style based on who you're giving it to. And we think there's broadly two categories of data that matter. One are just broad based clinical data. So like what are the right guidelines for ankle sprains or what is the right step wise progression of medications for someone with asthma? And that seems obvious, but there are a lot of PCPs and providers in this country that probably are not as up to date on the guidelines. And AI with the right techniques can be very, very good at adhering to instructions and protocols and guidelines. And so, so that's one type of data is how do we just feed in the latest evidence for a clinical scenario? The second type of data is personal data. So what do I know about you and your health history and your lab tests and your sleep scores and your biometrics, but even deeper than that, what do I know about you as a person? Do you prefer capsules or tablets? Do you prefer to get blood draws or do you hate needles? At the end of your life, are you someone who's going to want maximal intervention or minimal intervention? And that type of data, it's almost like learned preferential data about a person is maybe the most valuable type of data. And that's where PCP is accrue, is you get to know someone over 20, 30 years. That's why a lot of patients will never leave their PCP because they actually sort of feel like some part of their personality has been embedded in that person's ability to care for them. And so that type of data we value quite heavily as well.

Yeah, that's super interesting. How does, how does CAPS or council work in terms of, you know, I come in, I've got this issue, it's not something that's solvable necessarily through a chat. How do you guys handle kind of next steps and while still ensuring kind of that, that personalized experience?

Yeah, I think it's helpful here to think about council as a care, you know, council being like messaging based AI, native care as a care setting. And you can think of care settings as a kind of place of care that has its, like what it's good for and what it's not good for. And so the emergency room is a great place if you're treating a heart attack and it's probably a terrible place if you're treating and if you're giving vaccinations. It's just not suited for that. And we think of messaging based care as its own care setting. And so messaging based care is really good for a subset of conditions, but just like any other care setting, it has its upper and lower bounds where you have to escalate to a different care setting. And so someone comes to us and they're saying, hey, I'm having significant chest pain and my jaw hurts and I feel short of breath. There's a bunch of things we do, but the most important Thing is we escalate them to a different care setting, likely the emergency room and all of that. To say, when you're designing new care models, I think the key is to know where is that upper bound? At what point does an escalation become necessary and then how reliably can you make sure you're escalating when you're hitting that upper bound? Right. The worst thing you want to do is sort of miss those escalation points because that's what leads to patient harm and safety incidents. And so I think we've spent a lot of time and investment knowing what those upper bounds look like. What can we treat safely via asynchronous care? Where can we not? And then how do we programmatically escalate so that we're not missing any opportunities.

To move someone to higher care? Makes sense. Do you feel like, who do you feel like are your main competition? Is it Google and WebMD? Is it ChatGPT? Is it the prime, you know, the PPC?

Yeah, I think, you know, going back to an earlier point, I think healthcare is, it's, it's hard, I think for the human mind to even grasp the, the size of the market demand here. I mean, if you're in a situation where you've got supply demand mismatches by a factor of 10x, like there is almost no such thing as competition. Like you can flood the market essentially with like folks who are willing to deliver care and there will still be enough demand just because of, you know, the nature of the population and the need for health care. So, you know, that doesn't. Not to ignore your question, of course, we're a startup company in a space where others are probably building along similar contours as us in terms of AI native care models. And you could from a VC perspective consider those competitors because you know, they're building with a similar approach and they're going after a similar end user. But I think our mindset is very much, look, the demand is so high and there's enough room for a lot of people to get the model, get the care model really. Right. And kind of going back to our earlier, earlier discussion, every care setting and model has its own strengths and weaknesses, just like having different types of screwdrivers. And so just because someone else is building, let's say, a video based AI care platform, it doesn't mean that both of these can't coexist. In fact, there may be some subset of conditions where we're really well suited and another subset where brick and mortar Primary care or video based primary care is actually really good and the key is then how do we coexist to serve the patient? So yeah, I don't know. I know that doesn't directly answer the question, but that's sort of how we've thought about it.

Yeah, just really big.

How did the two of you look at the.

Go ahead John.

How did the two of you look at the future of integrating data into council? Maybe an example of one of the things that's been on my wish list is ordering a genetic test because a lot of the, the issues people have, I think they're expressed in the genes and then manifested by lifestyle, diet, et cetera.

I don't want to dictate this is Muthu's business, but I mean what I think we saw the big big here, there's like the small, there's the small, which is what we're currently doing, which is a very cool business. There's the big of we own all of the data for us kind of outcomes and diagnoses. The. That's also a pretty big idea. And I think the ultimate vision of what you can do with AI healthcare is the machine can almost predict you in advance of you having a problem. So we checked your wearable, we noticed that you've slept way less than usual. Plus you have two or three. Your Apple watch has been in these six different time zones. You know, we recommend you take these three supplements, get this many number of hours of sleep. And so you know, I think like the ATIA book is great, but one of the problems with atia's philosophy is that you have a human that has to drive compliance with all the good things. And I think the more that you can have a always on 247 physician in your pocket for the cost of compute. There's lots of things and there's obvious things like you know, we noticed that you swiped your credit card for two packs of cigarettes this week. What can we do to make that fewer packs of cigarettes. But then they're also just lifestyle things that at the intersection of sort of preventative health and big data, I think can create a really, a much more dynamic health care paradigm than the one we currently have. But Muthu probably has his own vision for his own business.

No, I think it's super exciting. And so it's like Rob's point is really, really great in that we started off talking about this thought experiment of what if we could take the best doctor that's ever lived and make them available to you at all times. I think Rob brings up a good point that maybe that's not good enough. Maybe there's actually a way to create something that's far better than the best doctor that's ever lived. Someone that is far more intelligent, more personalized, more proactive in their outreaches, more continuous. You know, an analogy would be, you know, there was a point in which, you know, the way you took care of your car is you come in for an oil change every six months or a year, and now most cars have built in, sensing that sort of tells you how everything is doing. It's sort of like having a mechanic in the engine at all times, giving you the green light that the car is working fine. And we don't really have the equivalent of that in healthcare. But to Rob's point, I think that's on the horizon is that you have an entity that looks far more capable than even the best doctor that's ever lived in your pocket at all times. And the key part being that it's effectively the cost to compute and one that can truly be available to everyone.

So one of the big things, one of my big frustrations in healthcare is that so much of the data is so siloed and doesn't talk to each other. And I'm curious why, from your perspective, why is that the case? Is it solvable? Can you pull all the data together? Because I feel like until you can do that, it's really hard to provide kind of the vision that you're laying out there. I need to know that you were in six different time zones, for example. Right.

So, yeah, this is another one where I think the criticism is fair of the industry that, you know, the data is siloed and it's hard to access. But if you were to just take a snapshot of what it looked like 10 years ago to today, I think we've made huge progress as an industry in that data sharing and aggregation. And so, like, right now, on council, one of the first things you can do when you sign up for an account is you can literally just select every health system you've ever been to sign on using your MyChart password or whatever the health system is, and within five minutes have all those records pulled in. And you can do that without a single fax machine, without a single phone call, like literally just in the app, by clicking and logging in. There's other data access techniques that exist today as well that we could go into, but it's gotten a lot easier now. There are still platforms like wearables, that have their own sort of data environments that they try to keep quite shielded and largely for kind of competitive reasons. You want to have some lock in for the member. But yeah, I think even Apple Health has done a phenomenal job of being an aggregator of both biometrics plus health record data. I think we're going to see a lot of progress here and I think it's already gotten significantly better than it used to. And the other thing I'd call out is that I think there's an over emphasis on structured data like I need to know your lab tests and your vital signs and your, you know, your last chest X ray report. And I would just emphasize again that there's a lot of like learned personality data that is so important that is really only generated from repeated engagement. Like I have to take care of Rob a thousand times as a human being to probably know enough about his preferences to deliver the best care. And one of the things you learn in med school is most care decisions, or at least a lot of care decisions, are not black and white. They're very much no, wrong answer. What does the patient want to do? Do you want to try third line chemotherapy for your refractory cancer? Do you want to go on dialysis because your kidneys are failing? Or would you rather live in peace for your remaining days? Do you want to take this medicine twice a day and have slightly better blood pressure control or once a day? And these are all preferences, they're not actually right answers. And getting good at preference based care requires knowing someone quite deeply.

What are like some of the things you're most excited about? Being able to build with council and just even outside of council when it comes to healthcare?

Yeah, we're, I mean we're excited about a tremendous amount of things that are on our roadmap and we'll be announcing some of them in coming months. And so I'll sort of save it for the announcements, I suppose. But before we started I was mentioning when any sort of AI lab leader gives a talk, they often make a point to mention healthcare is one of the undisputed goods of AI. It's one of the few maybe elements of AI where it's not a double edged sword. I mean this almost surely in healthcare will have mostly positive benefits. And so I think we're just so excited and grateful. It feels like we have a front row seat to getting to take whatever the top AI companies release. And then we immediately figure out how do we use this to make the care of people more effective and lower cost and safer. You can really take any sort of AI trend whether it's observability or using MCP servers or how do we do better rag and instruction following, how do we incorporate voice and video? Like any of those things you could immediately then ship to say okay, what does this mean for patients? And we get to do that. And so there's quite a bit coming that we're very excited about.

I'm excited for the MyChart data to sync. I didn't see that on my side when I started.

Yeah, if you go to your health profile and then you click on Connect my Providers, you'll see you'll get a chance to do that.

Oh, there it is. I missed it.

How about you Rob?

Glad you asked.

What gets you excited about this space?

Look, I think echoing something Muthu said there and making it more explicit, we get a lot of questions are you worried about this latest OpenAI or Gemini or whatever release? And the reality is it's fantastic for us the faster and harder that the foundation model companies run, the better our product gets for approximately free. So I think I've been in a weird way, when we met Muthu, it was a business that I was thinking about trying to find a way to build for a really long time, which was the connection of large data sets with healthcare outcomes and credit to a practicing physician and AI researcher who figured out how to combine that. But I think across our portfolio we have a somewhat schizophrenic view on AI at our firm, which is that almost every single company currently being founded is going to be worth absolutely nothing, both foundation model companies and application layer companies. But there will be a very small fraction of application layer companies that end up being worth a tremendous amount. And I think those will be the ones that whatever space they're in, healthcare or otherwise, are able to combine a proprietary data set and access to really interesting third party data with the benefit of the foundation model. Because I think the reality is I don't think I may be proved to be wrong but I'm willing to say on November 18, 2025 I do not think a foundation model company is going to win AI for healthcare. I think it's going to be a purpose built application layer company like Council and I think that's true in every space. And so I sort of see the foundation model companies as kind of the railroad tracks or the kind of core infrastructure for a lot of very specific use cases. And I think they're also really happy playing there. I think they enjoy messing around with, with specific use cases because it's interesting and useful. But the right way I think to build here is for them to put out the best possible kind of engine for everybody else to use and just sort of clip a toll. And if they do that well, they'll have a really interesting business. And I think that, you know, the competition in that space may go on to be pretty severe, but in the end there's, it's possible you could have a lot of winners in that space and there might even be one that's particularly good at healthcare. And so when I look about what we're doing in our portfolio in healthcare, it's a combination of this. It's a company called Head Start Health, which is a business in a box for autism care providers. We had a company called Miga Health, it was remote patient monitoring for cardiac care. I think a lot of the well intended healthcare technology innovation from kind of 2010 to 2020 was a little bit ahead of its time and I think now we have the ability also mixed with beneficial regulatory changes. I think one of the most exciting things is nurse practitioners being able to provide more of care and having reimbursements. Obviously the backbone of the healthcare system is that, you know, excluding very rare care settings. Reimbursement either from various government entities or commercial payers is a core driver of how the system works. And I think regulation and legislation is changing in good ways for that. At the same time that technology is making new things possible.

Yeah, makes, makes a lot of sense.

I'm.

Yeah, I worry just like you about every company in terms of like, are you just one demo day away from being obsolete? Right. From these big LLM models, but certainly unique data sets and I also think like unique user experiences are areas where you can create somewhat of a moat from these, from these large players. So. And I think this like healthcare is one of those things where like that data just builds over time as you build a deeper relationship with, with the patient. Right. Which I think is something that makes council really interesting is if you're able to like engage with the patient on a very frequent basis over a very long time period. I think that's, that's where you get some really interesting unlocks that we've already discussed. So yeah. Do you, are there any things that kind of keep you up at night other than, you know, chatgpt making a big push into your space?

You know, I don't think that keeps us up as much for the reasons we talked about where I think purpose built application companies, especially in Healthcare, I mean, we're taking so much clinical risk every day. Like, like there's a lot of companies out there that you wake up and you go to sleep and like the worst thing that can happen is like your platform goes down and like someone can't make their PowerPoint slides or I don't know, can't record their meeting notes or something, which is fine. But like that, that, that is, you can probably sleep easy knowing that like, you know, an outage is not going to ruin someone's life. But yeah, I mean as we're talking, we're doing lots and lots of medical visits behind the scenes and these are people that are, you know, sick and could get worse and we're making, you know, judgments and decisions on their health. You know, patients sometimes with suicidal intent, with cardiac emergencies and. Yeah, so like, I think especially as a physician, I think I'm sort of humbled to what could go wrong in the clinical side and just us wanting to do the best job there. And so yeah, I think that's where our focus always is, is the patient getting the best care. And have we orchestrated that the right way? We often tell engineers we're recruiting as well. I think this job will make you more proximal to impact than probably any other engineering job because a few lines of code can actually affect someone's life in a really big way. And if that's intimidating, then that's intimidating. If that feels like an opportunity to really democratize care in a huge way. I think Council is an exciting place. And so yeah, I think the main thing is, yeah, how do we keep our clinical group kind of continuing?

How do you protect against some of the risks that of using AI, Right. You think about like the people that have engaged really deeply with AI and has resulted in, you know, self harm, suicide, et cetera. And then you layer onto that like hallucinations, you know, all of those kinds of things. How do you guys think about that? How do you protect against that, those types of outcomes?

Yeah, I mean I think the first thing you acknowledge is that there, there's clearly a double standard here where AI is held to a higher standard than human doctors. And I think that's fine. I'm not saying there shouldn't. This is the same thing in the self driving space. I think self driving cars are held to a higher standard and it's a different standard. And I think that should be the case for a bunch of both moral and pragmatic and philosophical reasons where you have to hold them a bit more accountable and So I would just say it's sort of that off the bat. And so, yeah, how do we effectively ensure there's a 0% error rate or miss rate from the AI? And there's a number of things we do before we deploy the model that we do in real time and that we do retrospectively. All that are safety oriented, some of them we published on, others we don't talk a lot about, but they're very important to the platform. But yeah, we think of safety in a very sort of. It's the first and maybe only thing that matters in every interaction. And so before the model's deployed in real time and retrospectively, we've got a variety of techniques and parallel agents and other safety harnessing approaches to make sure the AI is complying. But the biggest thing to call out is that we are not an AI chatbot. We have rail doctors to backstop the care. And so at any time, and John, if you use the platform more, you'll see this at any time. If the conversation is becoming clinical in nature, we will suggest a handoff to a real doctor and one of our doctors will join the chat, usually in under five minutes. And so you can go from chatting with a personalized medical AI to chatting with a human in the same exact thread on the same platform in a couple minutes. And so we've essentially taken all of the friction out of accessing a doctor. And so that is by and large the ultimate safety stopgap. Which is also why we don't anticipate a lot of the foundation model companies getting into the space directly. Because do you really want to hire and manage a medical group and carry the clinical liability that comes with it of delivering care? A lot of these companies, the downside risk is way too big to sort of have on their balance sheet.

Your system seems much more thoughtful than ChatGPT and it hasn't been hallucinating on me.

Yeah, I'm glad you found that. And that's. There's a lot that goes on behind the scenes to making that experience come through.

Yeah, this past week I was messing with ChatGPT and I was asking it. I uploaded my lab results and he said, what's my testosterone? There wasn't a testosterone test performed, but it came back and gave me numbers and ranges and it said, I'm doing great.

Yeah, there was actually a really important nature paper that came out earlier this year on medical hallucinations from off the shelf foundation models. So like Quinn, you know, Mistral, ChatGPT, and yeah, if you ask things like, you know, there's some great examples of you can make up a medical term like black blood cells. It's actually not a real medical concept, but you could just say like, hey, what is a black blood cell? It'll make up a definition, it'll make up a range, it'll make up a value. And there are ways to make it not do that, but that's sort of. A lot of that work is being done at the application layer.

Awesome. Well, as we wrap up, this has been a really interesting conversation. I'm curious, if you had to bet in 10 years, will AI be a copilot to clinicians or will it frankly replace them, at least on the primary care level?

Rob, do you want to go first?

Yeah, look, I, I think this fits into my broader view on AI, which is AI is a wonderful triage mechanism and there's situations where AI can take you all the way home and there's situations where AI hopefully can and should recognize the limits of its own omnipotence. And so I believe that I don't, I don't think in my lifetime we'll see completely fully automated. Sorry, not in my lifetime. I don't think in my tenure as a venture investor, but possibly in my lifetime will we see fully automated care all the way through the whole process. But I think whereas 5 to 10 years ago AI was 0%, I think we'll see that it takes over 95% of physician patient interactions over the course of 15 to 20 years. But I also think that at least for a while, that setting broken bones and advising on cancer diagnoses is probably still going to be humanoid.

What's your take?

Yeah, I largely agree with that. I think if we restrain it to just primary care, I think my confidence that there will be more AI only workflows that happen goes, goes higher. I think subspecialists and specialists are much harder to. A lot of those are very hands on, requires a lot of robotics intervention as well and stuff like that. If you take just like the cognitive elements of primary care, I think we'll make a lot of headway in AI doing many of those workflows autonomously. I think there will be places where humans still jump in and I think which places those are will sort of vary over time. It's probably a lot like Tesla's full self driving endeavors too, where it can maybe do well on the highway, but still you have to park it yourself. And then over time you're chipping away at more and more sort of road scenarios. I think we'll see something similar where hopefully over time more and more of this can be done in AI first manner. But I do imagine there's places where human touch and involvement is going to be really necessary.

So one of the things that I feel like doctors is really important for doctors to do well is this whole concept of bedside manner. Do you think that's something that AI can replace? And part of the reason I ask is because I also think that that's one of the things that venture, venture capitalists can do well in terms of being, you know, there during the, the challenging times is not just a, a source of funding but also a little bit of a therapist or somebody to hear you out and support you as a founder. Right. So there's also an element of like does AI replace VCs? Does it have that potential? Right.

I think the difficult thing about AI replacing VCs and probably doctors in kind of high density, high intensity care situations is empathy. The best empathy from a non human probably doesn't feel that good. We all get the messages on, you know, united.com I'm so, you know, I'm so sorry your flight was canceled. Phrases like I'm so sorry your flight was canceled only have value insofar as they're sincere and something that's to going created by a computer model is per se not sincere. So color me slightly skeptical that we're ever going to want bad diagnoses rendered by physicians. And I think in the same way there will always be a role for providers of capital. I think the role of my industry is going to change. But at the end of the day, and I think this is one of the biggest misconceptions people have about venture do. I think that AI could replace a lot of private equity for sure. But venture capital is far more art than science because in the main when we back Muthu, there were no numbers for this business. There was a market opportunity which Jim and I could have written a really good summary of. But I think and by the way, we get it wrong a lot. We happen to get it right here. Why we got it right here as we saw. And Muthu is kind of very difficult to underwrite and explain human beings. Somebody that would be successful at doing this and somebody with his exact resume but not his habits of mind and Persona would have been less well shifted. And you might say, well, you create some sort of AI exam to interview Muthu and Bizarro Muthu and pick the one that was going to create the good company and Maybe I'm. I don't want to sound like one of those VCs that doesn't think their job is disruptible, but I think that that's something that's uniquely and innately human.

Interesting.

Yeah, love it. What are your thoughts?

I will not comment on the venture side because I don't have the qualifications to. But on the physician side, I think, yes, I think there are certain forms of just human to human connection that are impossible to replace and that sort of evokes something like at the hormonal level in people that just you will probably not get from like a robotic or AI kind of counterpart. I think it has been eye opening how quickly people have taken to relationships with AI. And I mean relationship in the broad sense of the term, which is people are trusting of it, they enjoy its company, they enjoy talking to it, they feel they can be vulnerable. A lot of things that people would have maybe not guessed would have been true. And so, yeah, I think there's been a number of studies on this too that people have viewed AI messages as more empathic than human messages from a clinical perspective. And so, yeah, I think we'll see. Like AI actually have quite good bedside manner and for some people, they actually prefer talking to an AI because it, it again has a bunch of characteristics. But yeah, to Rob's point, there's probably something innately human in some cases that it's just impossible to replicate.

Awesome. Well, thank you both so much for joining us today. It's been, as I mentioned before, a really fun conversation. I'm really excited about the future of healthcare and AI and this whole concept of being able to provide more, more personalized health advice and counsel at scale. I think it's going to just, you know, tremendous opportunity, not just on a business side, but also on, you know, the overall welfare of the world. So thank you for all of your work that you guys are doing and for joining us on the episode today.

Thank you guys.

Our pleasure. It was awesome.

Great.

Yeah, thanks. Thanks guys. And I appreciate it. Sarah.