The Verticalization of AI in Medtech and Healthcare | LSI USA '24

Gabe Jones  0:05  
Hello, thank you for joining us this morning. I'm Gabe Jones from Proprio. As in reset, this will be an exciting one. First of all, the founder gets to interview the investors, which is fun. From us to be mostly nice. The topic is a timely one, right? It's the subject of verticalization of AI. It's aI touching everything in healthcare. And you've got investors who come from technical and clinical backgrounds and who understand both medical device and the entire spectrum of the continuum of care. So I think we're well situated here. quick poll. If you were to identify yourself as either a bowl bear or undecided regarding AI, please vote. So bowls on AI. Okay, bears. No bears? Well, no, if I believe that, but undecided, that's okay to one undecided in the back. Okay. I don't know if we're going to convert you today. But we'll try. So first a thought construct that I've been kind of mulling over in this moment in time that we're in right now in AI. And this is not my quote. So I'm stealing it. And hopefully it frames up the conversation. First, humans in terms of our brains are equipped with what you might call a Paleolithic brain, right developed for a different time in different applications. We've also developed institutions that you might refer to as medieval so think medieval institutions, hint, hint, healthcare. But now we're equipped with what you might call God like technologies, right? Just think about that context and how much chaos that could generate, depending on your position regarding AI. So in that context, I'd love to start with you, Alex, you've shared a great framework for thinking about this particular moment in AI in healthcare that we talked about a week or so ago on our prep call, you can start with that, and then maybe link that to the opportunities in the uniqueness of this particular moment in AI and healthcare.

Alexander Morgan  1:53  
Sure. So, you know, I've been working in AI and biomedicine and healthcare for 20 years now, which is saying a bit, but you know, many kinds of new real dramatic new tech innovation do take some time. And, you know, in practice, things are mostly actually an S curve. But the first part of S curve is exponential. And the early part of an exponential, not a lot is going on for a very long time, until suddenly, things really start to accelerate. And over the last couple year, or year and a half or so, things have very clearly accelerated and many things that were not working, suddenly have flipped over to being effectively fairly commoditized. As technologies, it gives you an example, automatic transcription taking a patient clinician encounter in a medical clinic visit, and turning that into a structured medical note. Things weren't quite working despite years of work. Two, three years ago, nothing really worked there, most clinicians would be satisfied with the product. And suddenly, it's effectively a commoditized technology now, and to some extent, that that is that was some of the Vantage been driven by one of our portfolio companies open AI. But they're not the only ones able to do this. It's just a it things have really converged. And that timing, and that excitement around things in AI, actually both, it creates commodity units of technology, right, because everyone's always building on the shoulders of giants for the next generation. And as soon as we can take some of these things, and suddenly say, things weren't working, they weren't working, and suddenly they are working. But everyone actually can get access to something that works pretty well or good enough for most applications. And certainly with language, we seem to be there for many applications of how people think about using generative AI, for language for 2d images were like, let's say 70 80% of AR and in certain niche applications, I think we're effectively solved to where we need to be. And then we're, I think we're going to want to you know, kind of maybe talking about an area we haven't crossed crossed over, we aren't in the world of really the the convergence of the crossover point to what we might talk about in the near term, large physical models, where you would basically have systems that suddenly connect with it in the 3d world in the kind of way that we're now seeing things like chat GPU interact with people. And that has a second aspect where that that enthusiasm and that general recognition actually creates a lot of interest at the enterprise level. So you know, just pick an area like radiology, right? AI and machine learning systems have been better than community radiologists at many tasks interpreting mammography, mammogram, for example. indentify intracranial bleed, they have been better than community radiologists for a decade. There just wasn't any real adoption recognition and then suddenly, things have really switched over and you look at a company like rad AI, they are now accelerating a ton of adoption with a tool that is supporting radiologist in their work with AI and accelerating time to to complete an A note, making sure nothing's nothing's missed. So we're in a really great period for people that are building technology to take advantage of all this this kind of crossover point we're about things that weren't working suddenly are and then building that whole next generation of tools and services and technology on top of that

Gabe Jones  5:05  
excellent summary, unless you want to go next, and maybe talk from the perspective of investor, but also a clinician here or an MD, how AI in this particular moment is impacting you, and how you're looking at investments as an investor. But also, you have that very unique and valuable perspective of a clinician, if you could maybe bridge that gap for us. Yeah,

Akhilesh Pathipati  5:23  
definitely. I think Alex raises a lot of the points that we're spending a lot of time thinking about. And, you know, one of the things that I think is particularly interesting is the fact that some of these technologies do commoditize very quickly. And that's, I think, unique in the context of healthcare, because historically, a lot of innovation cannot commoditize quickly, because it's protected by part IP for devices, drugs, etc. But that's changing. And so I think that raises questions for companies that we have to answer as investors things around you, how can a company capture the value for an innovation that creates before others come along? How does it stay differentiated? What is defensible about it? One of the things that I think is linked to that is thinking about, what is the market opportunity look like? And if you end up facing pricing pressure, for example, can you still sustain an an interesting business. And then the last piece of it, which, which is also related as just figuring out how these business models work, because with the technological innovation happening so fast, it's difficult for the business models, the reimbursement, environment, all of those things to keep up. And so can a company figure that out? Before moving on to the next thing, or the or the next product? I think from the clinical perspective, it's a very exciting time, because so many of these things are transforming how care can be delivered. One thing that I think that companies and investors also need to think about is how to bring clinicians and physicians along, when you look at how physicians are reacting or you know, the surveys that are done of them now, a lot have a lot of interest in AI solutions, but also feel a lot of caution about what it's going to mean for who's ultimately making decisions about patient care what it does to the doctor patient relationship, what it means for patient privacy. And I think if if companies are not very sensitive to those concerns, there's going to end up being pushed back from the people ultimately into adopted and and supported. Let's

Gabe Jones  7:31  
come back to that adoption question, especially with clinicians. As we go back back through the topics here, Debbie, you have a unique perspective, you've your clinical bio informaticists, which is hard to say the PhD was probably harder to get, you've been on the the strategic side of the equation. Now you're at TRX. And you're somewhat on the other side of the table and some of those conversations, if you could speak to the strategics how they're thinking about this moment, as well as sort of the intersection and the kinds of companies that maybe you're seeing are possible to succeed and more likely to succeed in this environment. I think that'd be a nice bridge. Yeah, I

Debbie Lin  8:04  
mean, I was long time 15 years with a strategic with a pharma company and then ended up leading their CVC for digital health. And that time, this was five ish, you know, five ish years ago, four or five years ago, farmer I had not seen so recently, there's a news and I know that we were talking about Lily, being at the forefront of pharmacy delivering pharmacy direct to consumer with Amazon pharmacy, we would never have thought about that. In terms of direct to consumer patient delivery of medications from a pharma company at all four or five years ago. That's one second in Vidya and relation which is a drug discovery AI platform and Deerfield investing together to your field and Nvidia investigating investing together into this drug discovery. Deerfield, a very traditional biotech fund. And Nvidia normally don't see that kind of marriage but very, very interesting. So I think there's a lot there in the market signaling a lot of excitement around ai, ai, to Alex's point around radiology images. The data is very important. And I think we'll talk about that in a bit. standardized data, radiological images are standardized that comes through packs easy to kind of calculate over compute over so that cleanliness and the quality of the data is very, very important. So similar with language, right? You have conversations you we understand that and we can read in that information and, and be able to compute over that. Now when we think about making the leap over to clinical outcomes. If we think about algorithms or drug discovery that it's a little bit longer. So I think when we evaluate companies or technologies really need to look at the kind of data, the proprietary pneus of the data, as well as the capability to really deliver clinical outcomes and actionable insights in a meaningful and efficient amount of time, that can be commercializable. So that there is a valid economic value today or, you know, shortly. So that those are my main points. That's

Gabe Jones  10:32  
great. Thanks for teeing that up, Armen, you have a unique perspective in being in a larger investment group and then going out on your own and raising money to invest in space, you clearly have passionate a thesis around the area, you've also been in the or as the rep interacting with the physician and the clinical team. If you could speak to kind of the inefficiencies there in the sales operations environment in med device traditional, and how this moment might start to change those kinds of interactions. Sure,

Armen Vidian  10:58  
yeah, we are truly at an inflection point. And I say that because way back in the day, if I had said that an algorithm could demonstrate clinical significance, people would have laughed at me, in fact, I think they did. Not with me at me. And, you know, we are now seeing that real artificial intelligence can demonstrate that it can detect disease, produce images reliably, and guide physicians to better clinical outcomes. That's great. The challenge that we see, you know, in attending a lot of the sessions here this week, in hearing about how the med tech industry is embracing innovation and AI and that is great. But to really achieve the breakthroughs in the alar, to have people picking these this up at a at a faster clip, we have to restructure really rethink about how it is we're delivering these technologies to patients. And by that I specifically mean, people like me who were standing in the O R, because people like me, cost an annual estimated about $35 billion a year in sales operations cost. That's a figure in full disclosure of several different med tech companies estimates of what they spent $35 billion is what Chase just a rather Capital One just paid for Discover. Right? That's, it's a lot of money. And we as an industry, I say we from I used to be in it really have traditionally thrown people at the problem. Right? How do we have somebody standing in there in the alar, training somebody to do a procedure. And I think that we've really got to think about how how artificial intelligence, and other technologies associated with it can really enable the physician or other caregivers like nurses, medical assistants, and so on, to more effectively have a clinical outcome. And a get there on their own without that. But that really requires us rethinking how it is a medical device company is structured, its business model, where that revenue comes from, and some of those underlying structures. And that's why I, my firm, and others and my peers in the industry really believe that a lot of the breakthrough companies are going to come from startups, unencumbered by that previous model.

Gabe Jones  13:43  
That's a great segue. Alexei, I'm going to tee you up here and put you on the spot. If you could talk a little bit about from how your interaction and discussion with strategics is different. Now in this moment, that's taking off of what Armen just said about the likelihood of a large strategic being able to transform into an AI company quote, unquote, and the unique opportunity that startups have to position themselves at this moment for potential partnerships in the future with strategics.

Alexander Morgan  14:10  
Well, in every large strategic organization, right now has some kind of spun up an AI strategy group or an innovation team at this point. Now, but they're large organizations and so they do we do they do you actually, everything that you ask because they do reach out to us because we were big investors in things in AI. But if I were to, you know, some groups will very quickly, clearly Microsoft has really leaned in. Other organizations are not moving nearly as quickly. And we are at one of those major periods where there is a kind of tidal wave of tech innovation happening, and many things are up for grabs. You know, the future is unwritten. And it's people in this room who are gonna actually determine how the future bakes out, and it's gonna take a lot of work, but one of the advantages being a small company is you have the agility and the ability to try something and the ability to move quickly and you don't have to have of C suite initiative and report and then a decision to allocate funding and resource it, you can just start doing things. And as Armen mentioned, one of the key things is the scaling property of software, right. So traditionally, in healthcare problems have been solved by throwing labor at the problem. And that scale sub linearly in that quality doesn't quite scale linearly, you have to have other managers, there's a lot of organizational any change then becomes harder and harder to roll out organizational change, because it requires behavior change and training. Well, technology, of course, is much easier to scale hardware, there are economies of scales and midday manufacturing, great properties, although there are huge upfront costs and tooling manufacturing, and then it's can be quite hard to pivot too much once you have hardware technology, software can still have significant innovation costs, upfront costs to build, but then the marginal cost of scaling is usually near zero. Now currently, with a lot of AI technology, it's expensive to do compute, but that might be 10 cents per episode of actual scanning, as opposed to a hardware that again, it's if some, you're trying to roll out billions of an item, if it costs $1 $2 $5 to manufacture. And then of course, the all in cost of shipping and distribution and so forth, software has very easy distribution, very easy marginal scaling cost. And also quality improvements can often be very easily pushed out. And so your product gets better as a function of scale, which is a really unique thing. So you got again more that greater scale, you're you're at the Greater rationale you have for additional investment in improving it. Whereas for human infrastructure, trying to improve, it gets harder as a function of scale software, it actually your product, it gets better, the more it's distributed in us. And that's a very unique property. And so, you know, most of have much of healthcare is, you know, to Harmons point involves interaction it is, you know, human to human late, you know, all kinds of stuff where the scaling properties are terrible. So the quality at scale is terrible. The cost that at scale is terrible. All kinds of things get inverted. In a world where a lot of that quality and value creation, Sif shifts to AI and software.

Gabe Jones  17:06  
Love it if you could continue on, from that clinicians perspective, what moves the needle for adoption to drive that initially inefficient process, but then skip much more scalable processes Alex alluded to.

Akhilesh Pathipati  17:19  
Yeah, I think there's a there's a couple things. You know, I think the reality is that the adoption doesn't move as fast as the technology does. And there's a couple of reasons for that to think one is building trust. So making sure that the clinicians feel comfortable with the new technology as it comes along. I think the second thing that is going to become your is something that needs to be figured out more and more, some of these models come along as the regulatory environment. So I think the FDA has interacted with a lot of algorithms at this point, but they do tend to be close to algorithms, you kind of get what the inputs and outputs are going to be, as some of the systems advanced beyond that, there's going to be need need to be new frameworks that are developed. And then the third thing I think, is payment. we've alluded to this already, but how some of these technologies get paid for, it's still a work in progress in a lot of cases. All of those things, I think, are very addressable. But they're, they're the pieces that I think, you know, cause cause adoption will be but be a bit slower. Once you work through them, I think that there is going to be rapid adoption and scaling. You know, the value that these things create in terms of workflow efficiency, clinical outcomes is apparent. And so they're all there'll be a lot of interest.

Gabe Jones  18:38  
Someone from Mayo yesterday, a physician on a panel said something to the effect of AI doesn't drive change management and healthcare data does. That's a clinician saying that, I think that's an interesting frame of reference for what you just said, which is, you know, clinicians want to see data and proof to your point as well before I bring that in product into my workflow. And I think that's what that person was alluding to. But from a technologist perspective, we would say great product, whether AI is under the hood, and does this or that or the other thing is what really drives change. From a user behavior adoption perspective, that's at least our opinion, Anybody got an opinion on driving adoption among clinicians, with AI driven products going to open up to the panel to talk about that?

Akhilesh Pathipati  19:21  
I think it goes back again, to clinical actionable, meaningful insight, at the time of at the point of care if you can create with AI this capability to make a decision quicker or and help a patient faster, and they can see it, and people want to pay for it or the payers want to pay for it. Absolutely.

Gabe Jones  19:42  
Could you extend that thinking because of your unique perspective into the actual product development process? So whether that's new drug discovery and development companies like recursion come to mind where they've got both public and private datasets and now they're running their own internal sort of foundation models to simulate the drug development process. faster so that different employees within their ecosystem can sort of, you know, run simulations of new molecule development much faster and cheaper.

Akhilesh Pathipati  20:09  
Yeah. So I was asked when a company says, I can do that, and I have those new algorithms show me. And a lot of times, it's very hard to show me. And they'll tell me, okay, well, I can make this incremental. I can, I can validate incrementally, I can make molecules that look like a set of molecules that was made before and I'm like, Well, okay, so So what, show me something different, but then you need to take it to clinic or you need to take into a mouse. And that takes a long time. So I would say that, though, there's still question marks around it. Certainly, there are combinations of molecules you could have you can make now that probably never before have been made because of the capability to compute over that space. But you still need to put in a mouse and you still need to put in a human that takes that data

Gabe Jones  20:58  
improve? Yeah. I mean, could you talk a little bit to the founders, and people running early stage companies in the audience, and this will also be watched by other people post facto. So think about that audience how they should be. Founders should be targeting datasets, products, applications, that will drive the kind of adoption that then becomes very valuable in the marketplace, both for m&a, acquisition and exits. How should founders be thinking about using these tools to establish that initial foothold that Alex was was alluding to? Sure.

Debbie Lin  21:27  
So I think that when I meet with a lot of founders, one of the questions that often presents itself is you have something that's really great for a physician to use, or it's got some novel aspects to it, there's a lot of whiz bang factor, but really nailing on the on the on the head, what is the problem exactly at Salt, we see a lot of things which would be incrementally nice, which some physicians would find that could be a nice to have in their day. But you got to think about physicians amount of time per patient, especially when you're looking at GPS, it's eight minutes they have per patient. Various other specialties have limited interaction with the actual patient, or it's a person lying in front of them. So you got to really think about how is that going to a produce that clinical outcome. But second, how's it going to be transparent to the physician, as to how you're getting what you're getting? A lot of the hesitation I saw early on in adoption, was because it really wasn't obvious how you got the answer. If you diagnose somebody, or there was an outcome you were trying to guide them toward, they didn't know why. And they felt that they were driving blindly with whatever it is the robotics they were given. So it's got to be transparent. Thinking about things in STAPs, augmented reality, before full on robotics, to judge you just trust, thinking about presenting an image that is subject to an interpretation by the physician that they like to do, because that's what they went to school to do, as opposed to, we're going to take that away from you. So really think about their day, their needs in the product. And how it is that it's going to make their jobs more fun.

Gabe Jones  23:32  
Like that. That's actually a good segue. Alex, if you could talk a little bit about sort of your thoughts and framing up the entire industry and how things are moving with respect to that human in the loop. Be at the clinician, perhaps in this case, or another key contributor to an actual decision in the healthcare healthcare economy. Where Yeah, it's great to take a foundation model and throw it at a data set. But the human loop reinforcement learning human feedback, or lhf, might be the most value added. Also, we want a clinician in that workflow, like calculation, validating that decision from a reimbursement perspective as well, we have to have that. Can you talk a little bit about the intersection?

Alexander Morgan  24:15  
Well, I you know, the the last point that you made about reimbursement does presuppose a bunch of things, which you may not necessarily be taking a typical reimbursement route for many of the things you deliver and there are a lot of other paths, right? self insured employers in the US have been very early adopters of a lot of digital innovation in more and more that's becoming AI technologies, right? So we have a company sword that does that home physical therapy, and they have written their, you know, mostly getting most of their traction with self insured employers. They are getting, they're really leaning out on how you use AI to improve that that product experience. And, you know, it's easy to focus on physicians as the group that you're trying to interact with. But of course, Is there only a tiny fraction of the people human beings involved in the biomedical industrial complex? Right, Armand mentioned, you know, drugs. Basically, drug development companies spend 1.7x on sales and marketing they do on developing. Okay, well, then there are companies like Komodo, which is maybe again, I'm not involved with Komodo, but I'll just use that it's a gamble. It's kind of an old generation AI company, but they've been able to really move the needle and say, We will help with reducing cost or making the ROI on your sales and marketing spend much more efficient telephony, you know, still, much of much of medicine involves telephony, right? People call and try to do scheduling. If you look at what a Polly AI can do with a an AI agent that you interact with, that seems very empathetic, very patient will spend lots of time talking you, most people can't actually tell that, you know, we've, a couple years ago, you know, you would call up FedEx to ask a tracking number, it's very clearly you're talking to a machine, there are systems now where the average person cannot tell that it is not a human being that's interacting with them. And those there are lots of opportunities in healthcare. Where there, when you have an industry, which is 70% of the economy, there's room for 1000s of billion dollar businesses to provide a ton of value. And you know, many paths, we don't have to necessarily think about traditional reimbursement routes, there's lots of money available, there's lots of that, well, another way to look at it is there's lots of value creation and value providing opportunities. And if you position them in the way that Armen described, you really have to make sure that you're actually helping someone and that you're you realize that there's a human decision maker on the other end that you would need to make it feel like they are helping their job and their position in the in the organization improve by bringing a product to them, if you can solve that there's tons of opportunity. Now, there are cases where you know, there are digital diagnostics, digital therapeutics that are taking an FDA regulated approach, right, you know, companies like robots doing clinical trial for postpartum depression with a chatbot. There are many applications where people are taking MIT AI is medical device route, you know, alive core in IDX are some of the earliest companies to go through FDA with AI as a diagnostic tool, and all that that's an opportunity to, but there's not just one template for any of these things. And there's just a Gatling, you know, as, especially for an early stage company, you don't have a major large incumbent labor force, you typically don't have a lot of CapEx and in real estate, you have a lot of opportunity to to try to move quickly and take advantage of any innovation and build on top of that. And there's lots of opportunity to build and create value for stakeholders

Gabe Jones  27:26  
love it, if any of the rest of you could speak to how you're seeing proprietary data, as we discussed before the panel, plus these kinds of tooling actually linked together to add differentiated value. So whether that's like use the recursion example of public patent data on existing molecules, plus private data cameras and sensors, on assays on microscopes, etc, literally at the benchtop to enable a bench scientist to perform like a Chief Science Officer, if you will, in terms of simulating molecule development and applications like that. Obviously, I'm biased because I think proprietary data is very valuable in the clinical settings, so much so that built systems to go get those data to build other things. Any interesting ideas or companies you've seen lately, who are leveraging data, in particular data they are acquiring or accessing, in unique ways to

Akhilesh Pathipati  28:13  
build things. Right full time I work for Karis life sciences, I think you might all no Tempus as well, as a company, former Groupon founder. So same space, they started as IHC companies, molecular diagnostic profiling companies, over 40 years, I'll just speak for Karis sitting on a mountain of data. And so now it's okay. It's no longer while it is still molecular diagnostics, bread and butter, but you've got a whole set of data that you're sitting on that you're leveraging and productizing and selling the farm on all of that. So it's these companies evolve. I wouldn't say this, I don't call I wouldn't call it a strategic it's pre IPO both of these companies. But he says companies evolve and they change and you know, when you can sort of leverage the data set that you built, and then productize it in an experience or a product that that people would pay for. I think that that would be an example. I know that

Debbie Lin  29:13  
sent a med tech example. But I think it's still as illustrative for the audience anyway, I had a colleague, a lead an investment in a company in my prior firm, called recursion, which uses machine vision to watch cells retreat from diseased into healthy states. And when we took a look at them in their earliest stages in a small lab in Utah, one thing that we noticed was with extraordinary attention to detail. They had these plates on which they conducted all of these experiments. And they were initially starting with monogenic diseases and they watched how it is that impacted certain cell functions that they expect To see and one day, their experiments were off, inexplicably. And they went through everything and they couldn't figure it and they said, You know what? I think something about the petri dishes we're using is different. So they called the petri dish manufacturer, like imagine, you know, a Petri dish manufacturer getting a call from a startup in Utah, asking if they did something differently with the petri dishes. This is a crank call, right? And they said, sure enough, yes, the dye we're using is different. And they made them give them the old petri dishes. It was in that moment that we knew that we wanted to invest in this company. So it's that attention to detail around understanding what you have. Getting to the bottom of why it's so important that you have data that is really great and consistent and pure. That led to later on an investment from Nvidia. Because who else's data would you use? I think in med tech, what you need to know about that data, the question is, if I'm going to say the data is valuable, I need to know what the data is going to be valuable for. Right? How is it going to either steer me to? If I have a bunch of surgical data? is it applicable to a clinical circumstance? What's my proof? What is my audience gonna ask of me to do it? So that's how I think about it. Love it,

Gabe Jones  31:36  
Oculus? Could you speak to data, proprietary data, the value of it and at nvm? Now putting on your investor hat? How you guys think about those opportunities for companies claiming to build something valuable with a proprietary data set, for example?

Akhilesh Pathipati  31:51  
Yeah, no, absolutely. I'll make a statement that is maybe obvious. But I think still worth saying, I think that a lot of the companies that ended up having very valuable proprietary data, are companies that were able to collect that data, because they were doing something else. Maybe it was putting sensors and or so they have very unique surgical data.

Gabe Jones  32:09  
I appreciate that. Thank you. Yes.

Akhilesh Pathipati  32:12  
I think that where we see a lot of companies trip up from the domestic mindset is they go looking for existing datasets, and try to make those proprietary. And so we'll hear companies say things like, I have exclusive access to data from x large health system. So that's a proprietary data. It may be proprietary, but it's not unique, because there'll be lots of other large health systems. And so I think that when companies are thinking about what makes data truly proprietary and special, they need to think about what do they have that no one else has? And as Armand said, How's that? How's that useful?

Alexander Morgan  32:49  
Can I add something because it's even there. In some cases, it's worse than that, in that the value of data for things in machine learning is typically logarithmic, right, let's say you have 10 data points to get the same left, you then need 100. And then you need 1010 1000. And if you were in any way, spending linearly to collect that data, that is a divergent function. And that is not a path to success. Now, there are special cases where you get a monopoly of all of the data of a category, and that has some advantages. But that's a very important property that many people forget when they're trying to obtain a data moat. And, you know, the other key thing is many data types also massively depreciate very quickly. So for example, if you're very cutting edge with a new sequencing modality, if we're now we're back on the cost cutting curve for sequencing, so that data asset starts to depreciate very rapidly foundation models that people are investing in why Nvidia has this money to invest in recursion, those foundation models that are being trained depreciate incredibly rapidly. And so if you're not able to catch the winning edge, and builds the kind of winner take all scenario, you are burning money without really creating intrinsic protectable value for your company. And I think that's where some companies get their data acquisition strategy, rah, rah, you know, incorrect, it is great if you're, for example, making being paid to collect proprietary data, like you're doing something and you're collecting data in the emergency in the operating theatre, that has value. You know, Google's the classic example of a company that built a great data mill, because they were making money with everyone clicking in the early days and giving PageRank feedback. And so they, you know, got an advantage that companies like, you know, Bing and Microsoft couldn't catch up with, but that they're not all data acquisition and strategy are the same. And there is very clearly a huge economics to data in its value and its cost. Love

Gabe Jones  34:34  
it. If we're about five minutes left, I've got two different clocks, and they're off by a minute. So we're just gonna go by the average here. Could you talk a little bit about for the startups in the room and checking out the feed later, and the strategics on the other side of that, what is it startups can do at this moment with these technologies, these data opportunities to position themselves want to be able to establish that initial position in the market to show that value, be it with a clinician or in a workflow or with a sales operations for was to take enough of a position of the right kind of position in the market, such that one of two things can happen, right? The strategic then goes, oh shit, that's going to impact our existing business in a way that we should have a reaction to that either partnering with that group or maybe acquiring, or that you can carve out enough of a niche to be a sustainable and growing business, despite the downward pressure from, you know, really powerful incumbents in the space who have massive work, sales forces or forces open to the panel, how do you do that, at this moment in time,

Alexander Morgan  35:33  
that your product market fit, right. So that's very different from products and categories. And, you know, they're, you know, that that's a classic issue and, and most the way, the other frame that implicitly, you're saying it's actually most small companies really struggle to get a product that the consumer that is has the ability to pay, and you know, money to spend, and the biggest decision maker to actually be like, this is the product I need. And if you know, and you know, it is true, many companies struggle with adoption. But if you crossover and figure it out, and good product design, many cases, people will say, Oh, you'll never get off, you never get adoption. And suddenly you do like I mentioned rad AI because we were investors 10 years ago, when a company called Zebra, one of the first AI radiology companies they really struggled. Times change that I figured out exactly the product they needed to build zips along its ton of adoption.

Gabe Jones  36:23  
Yep. That's actually maybe I'll reframe the question better. Thank you for that. What is product market fit? How is it different now? And how are you measuring it or saying it in this context?

Debbie Lin  36:34  
Yeah, so it was something I always say is, who is the user buyer? Who's the financial buyer? And who's the technical buyer? And are those three people often the same person. So when I think about product market fit is the person benefiting from your technology, the person who's paying for it. I saw this recently in my own family, where I could have benefited from a lot of these information, workflow, things that I see coming my way, but I'm not the one paying for it. So no, absent that, no one's going to do it. And I just get to continue to bang my head against the wall. But Alex is completely right, that it's all about that product market fit. And I think that if you're thinking about opening up some medical devices to new users, or allowing an existing user base to do something a lot more easily, and you know, have a seamless experience, and get to be something that's their essential tool and partner in the alar, then you have something that people are going to want the same things that apply to consumers apply to physicians and hospital systems as well.

Akhilesh Pathipati  37:54  
I have just a bit to add to both of my colleagues here. product market fit, but also to Alex's point look for non traditional buyers. And you need to paint a vision or an experience. That might not I mean, he was saying, you know, you don't have to go the traditional reimbursement model life insurance, for example. I mean, there were test cases or pilots that if you're a smaller startup, you might might want a partnership, paint a vision with potentially I mean, the reason why I talk about life insurance is traditionally it wasn't the sort of payer that one might go to, to sell a digital health device or a wearable device, but they care about people's health, and making sure that people are well in the course of their life. So potentially, you could pair up with life insurance, in helping with diabetes, a pharma company, digital health, wearable, whatever. And these are sort of stakeholders where you might not necessarily pair with a payer that you might not necessarily have thought so before, so just be creative as my my point.

Gabe Jones  39:03  
Okay, let's bring us home. Yeah,

I think that I think that there's all good points. I liked what Alex said about how data gets old. I think it's characteristic of the fact that there is a condensed lifecycle to innovation in a lot of these technologies. And what that means is that it's faster iteration. Companies need to stay ahead of things and move a little bit quicker, perhaps, but ultimately, the business fundamentals of finding product market fit, demonstrating value. I think all those things remain the same with these technologies. And yeah, it's always worth remembering that

perfect this clock is to be believed that we are right on time. If that one we got 30 seconds, but thank you all very much.