VCs in Digital Health: Are You Optimizing to Make Money, or Are You Optimizing to Make an Impact?
A new interview with Ankit Jain, co-founder and CEO of Infinitus Systems, a healthcare AI company. Ankit, an ex-Googler and former GP at Gradient Ventures, Google’s AI VC fund, shares his insights.
Welcome to AI Health Uncut, a brutally honest newsletter on AI, innovation, and the state of the healthcare market. If you’d like to sign up to receive issues over email, you can do so here.
Alex Koshykov and I went head-to-head with Ankit Jain, a digital health innovator. We didn’t pull any punches—AI, healthcare’s lack of innovation, the VC echo chamber—it was all on the table.
Catch the full exchange on the ‘Digital Health & Tech Innovation’ YouTube channel or over at WellAI’s channel.
Key Highlights from the Interview
Here are Alex’s and BeKey’s bullet points of the interview from LinkedIn:
🔹 Ankit’s journey from tech to healthcare and the a-ha moments he experienced along the way.
🔹 How Infinitus is revolutionizing healthcare with AI-driven solutions and what’s next on their roadmap.
🔹 The impact of AI on back-office jobs in healthcare and how to address concerns about job displacement.
🔹 The growing gap between healthcare providers adopting advanced tech and those lagging behind—what needs to change?
🔹 The future of AI in healthcare: Will it transform primary care or remain focused on automation?
🔹 Insights into the role of venture capital in digital health and the path to profitability for startups.
Outline of the Interview
00:00 Intro
01:05 Greetings
02:32 Ankit’s journey
06:21 How did you end up in healthcare?
08:46 Tell us more about Infinitus. What else is on the roadmap?
11:47 Were there any a-ha moments once you got into healthcare?
14:53 Your product is designed to automate many of the tasks currently handled by back-office administrative staff, some of whom do not have a medical background. As the adoption of your product increases, how do you address concerns about the potential impact on the jobs of these individuals?
16:41 I have this dissonance. There are still hundreds of hospitals and healthcare providers in the U.S. that still haven’t implemented simple scheduling software. However, some have already started using such breakthrough solutions as Infinitus. So the gap between these providers is only growing. What needs to happen to narrow that gap or do you think those old school providers will just be out of business in a few years?
21:26 A person who worked for Google for a long time, what do you think about the hypothesis that AI will replace standard search at some point?
28:20 What do you think the role of AI will be in healthcare, particularly primary care, over the next 5-10 years? Will it fundamentally change how we deliver and provide care, or will it mainly be used for automation and administrative tasks?
33:06 Will digital health continue down the path of repackaging and infringement, or is there a new direction for true innovation in digital health?
37:08 There are two schools of thought on AI: 1) AI will scale and skyrocket, 2) AI costs will outweigh benefits—Jevons Paradox (Goldman Sachs, Sequoia Capital). Where do you stand between these camps?
39:47 The role of VCs in digital health
45:57 Is there a path to profitability in digital health and what are the obstacles?
49:12 Ankit’s advice to startup founders in digital health and AI
Summary of the Interview
The integration of Artificial Intelligence (AI) into healthcare is transforming the industry in profound ways. From streamlining administrative tasks to enhancing patient care, AI offers numerous opportunities and challenges that stakeholders must navigate. This article explores the journey of Ankit Jain, founder of Infinitus Systems, as he shares insights on the intersection of AI and healthcare, discussing innovations, challenges, and the role of venture capital in driving change.
Ankit Jain’s Journey into Healthcare
Ankit Jain’s path to founding Infinitus Systems showcases a unique blend of technology and healthcare. Having spent years in the tech industry, particularly at Google, Ankit witnessed the power of data and AI firsthand. His journey began in the Bay Area, where he was inspired by his father’s career in network security and entrepreneurship. This backdrop set the stage for Ankit’s eventual focus on healthcare.
After studying at UC Berkeley, Ankit joined a startup focused on natural language understanding. Although the startup did not succeed, it was acquired by Google, where he contributed to the launch of Google Play. This experience deepened his understanding of user engagement and the importance of data in driving technology.
The Genesis of Infinitus
The idea for Infinitus emerged during Ankit’s time at Gradient Ventures, where he invested in AI startups. A pivotal moment occurred when he witnessed a demo of Google Duplex, which showcased AI’s capability to make phone calls and schedule appointments. Ankit’s wife, who has extensive experience in healthcare, pointed out that similar technology could revolutionize healthcare operations. This conversation ignited Ankit’s passion for applying AI to improve healthcare efficiency.
Infinitus Systems: Automating Healthcare Operations
Infinitus Systems focuses on automating back-office operations in healthcare using AI. The platform aims to streamline tedious administrative tasks that currently require significant human effort. Ankit emphasizes that the healthcare industry is rife with inefficiencies, particularly in the coordination between various entities such as providers, payers, and pharmacies.
One of the challenges in healthcare is the lack of standardized APIs for data exchange, leading to reliance on phone calls for information verification. Infinitus addresses this issue by automating communications, thereby reducing the burden on administrative staff and allowing them to focus on higher-value tasks.
Addressing Workforce Concerns
As AI adoption increases, concerns about job displacement in healthcare are prevalent. Ankit reassures that so far, major customers of Infinitus have not laid off employees due to the implementation of their technology. Instead, the demand for support staff is growing, particularly in specialties requiring high-cost medications. Infinitus enables healthcare providers to redirect their workforce to critical areas, ultimately enhancing patient care.
The Growing Gap in Healthcare Technology Adoption
Despite advancements, Ankit observes a significant disparity in technology adoption among healthcare providers. Many hospitals still rely on outdated systems, while others have embraced innovative solutions like Infinitus. This gap raises questions about the future of traditional providers. Ankit believes that as technology becomes more integral to healthcare, those who resist change may struggle to survive.
To bridge this gap, healthcare organizations must recognize the complexity of patient scheduling and care coordination. Ankit highlights that simple scheduling solutions cannot address the intricate needs of patients with complex health issues. Therefore, the focus should be on integrating technology into existing workflows to enhance care delivery.
The Role of AI in Future Healthcare Delivery
Looking ahead, Ankit envisions AI playing a transformative role in healthcare. Beyond automating administrative tasks, AI has the potential to enhance patient engagement and streamline drug discovery processes. By acting as a partner to patients, AI can provide necessary information and support throughout their healthcare journey.
Moreover, advancements in AI-driven drug discovery could significantly reduce the time and cost associated with developing new therapies. By leveraging AI to test hypotheses and analyze data, the healthcare industry can enhance the effectiveness of treatments while reducing financial burdens on patients.
The Venture Capital Landscape in Digital Health
Venture capital is a critical driver of innovation in digital health. However, Ankit notes that not all venture capitalists contribute positively to the startup ecosystem. Many VCs prioritize financial returns over meaningful impact, leading to a disconnect between startup goals and investor expectations.
The challenge lies in finding a balance between profitability and impact. Ankit emphasizes that while some startups may thrive without venture funding, others must navigate the complexities of scaling their businesses to meet investor demands. The focus should be on building sustainable models that prioritize patient care and long-term success.
Advice for Startup Founders in Digital Health
For aspiring founders in the digital health space, Ankit offers valuable advice. He emphasizes the importance of focusing on simplifying healthcare processes rather than adding complexity. Each new solution should aim to enhance existing workflows and improve patient outcomes.
Additionally, founders should approach innovation with a realistic mindset. While optimism is crucial, understanding the nuances of the healthcare ecosystem will lead to more effective solutions. By considering the broader implications of their technology, founders can create lasting change in the industry.
Conclusion
The intersection of AI and healthcare presents both challenges and opportunities. As innovators like Ankit Jain of Infinitus Systems continue to push the boundaries of technology, the industry stands on the brink of transformation. By leveraging AI to streamline operations, enhance patient care, and drive innovation, healthcare can evolve into a more efficient and compassionate system.
In this rapidly changing landscape, it is crucial for stakeholders to collaborate, learn from one another, and prioritize the well-being of patients. Only then can we hope to witness the profound changes that the future of healthcare promises.
Full Transcript of the Interview
Intersection of AI and Healthcare: Challenges, Innovations & Investment. Interview with Ankit Jain.
00:00 Intro
Ankit Jain:
The number of times someone comes to me and says, “Hey, can you do this for fintech or can you do this for real estate?” I’m like, “No, I don’t have enough time. There’s so much to do in healthcare.” It was incredible to have that front-row seat to seeing what was possible with these new breakthroughs. The question of “Why me?” comes in every household. “Hey, a modern insurance company should have a different business model, or a modern hospital or health system should have a different business model than a traditional one.” I truly think in the next 10 to 15 years, that’s going to be a thing of the past. Oh my God, it all came together, all that work for 20 to 30 years pays dividends. Every major Fortune 500 company, the board was saying, “What are you doing with AI and LLMs?” There’s a whole class of great companies that should never raise an ounce of venture capital. The question is, are you optimizing to make money, or are you optimizing to make an impact? Thank you, Silicon Valley, for saving the world again.
[Music]
01:05 Greetings
Alex Koshykov:
So welcome to our digital health [Music] interviews. Today, we have a special episode. We decided to change the format just a little bit, but before we get to the interview, I’d like to remind you to subscribe to our channel, like this video, and also please leave in the comments the name of a person you think should be on our podcast next. And you can already see that today we have Sergei with us. Some of our viewers would be surprised since the interview with him just came out a few days ago or a few weeks ago, but actually today, Sergei joined me to interview our guest. So, Sergei is gonna be my colleague interviewing someone else.
Hi, Sergei.
Sergei:
Hi, Alex, how are you?
[Music]
I’m doing really well and super excited about our interview today. Our guest is Ankit Jain. Hi, Ankit, how are you doing?
Ankit:
I’m doing very well, Alex and Sergei. Nice to meet both of you, and thanks for having me.
[Music]
Alex:
Thanks for spending your time and doing this interview. I think it’s going to be great. Let’s start with your intro. Please tell us a bit about yourself.
02:32 Ankit’s journey
Ankit:
So, I’m the founder and the company lead at Infinitus. We’re a five-and-a-half-year-old startup out of San Francisco that’s focused on automating a lot of tedious back-office operations in healthcare using AI. But stepping back, I’ve been in the Bay Area for about 28 years. I moved here when my dad wanted to get into the startup ecosystem. He spent his career in the network security space, and I saw him building companies, going from zero to one over and over again. It was a big inspiration in my own journey as an entrepreneur.
After grad school, I went to UC Berkeley, studying linear algebra on multi-core machines—like a lot of the stuff that is now at the basis of Transformer models and AI in general. But I joined a search engine startup that was trying to push the bounds of what you could do from a natural language understanding perspective. The company didn’t do very well, but we ended up getting acquired by Google for the technology and the team. I joined the Android team, and on the Android team, we launched something called Google Play. So, I went from the world of natural language processing for the web to understanding the app ecosystem. We built a search engine for Google Play. We built the recommendation engine. We built a personalization platform that almost every part of Play used and really saw that drive engagement.
One of the things I learned when I was doing that was the power of data on everyone’s mobile devices. Now, if we go back 12 years to 2012, mobile ads absolutely sucked. You could open any app and you saw the same banner ads everywhere; there really wasn’t personalization. If you go to websites today or mobile websites today or apps today, you see things that are actually somewhat relevant. It blew my mind because web advertising continued to grow really fast. Mobile advertising was just getting started. AdMob had been acquired by Google, but really, Flurry and the work that Twitter was doing in mobile was interesting. Facebook had a mobile ads product, but it was mostly brand advertising. It wasn’t targeted performance advertising because the data didn’t exist.
So, I left Google to start a company that could understand users on their devices, again using data from this device, and use it to improve the experience on mobile. We ended up building an interesting competitive intelligence company instead. We had data from 250 million phones around the world. That panel was one of the largest panels of real-time, near-real-time data of what people were doing on their phones, and we ended up exiting that company to a larger competitor after a couple of years.
I then came back to Google to help start Google’s AI Venture Fund. I was investing in early-stage AI companies, whether vertical full-application companies or horizontal platform companies. This was in 2017, so right when the Transformers paper came out, I think the leaders of Google rightfully said, “Well, we should invest in this new wave of technology.” It was incredible to have that front-row seat to seeing what was possible with these new breakthroughs. It was almost too good to be true, where I was like, “I cannot miss this wave of building things.” A couple of years in, I decided with my co-founder Sham to jump back into building companies and started Infinitus, and that’s where we are today.
Sergei:
How did you actually—I’m just curious—how did you start Infinitus? How did you end up in healthcare? Because I think you have a similar story; I kind of wandered into healthcare from somewhere else. So, I’m just curious about that story.
06:21 How did you end up in healthcare?
Ankit:
Yeah, you know, I think in my case, I was at Gradient Ventures, investing in a lot of AI. I remember Google I/O 2018; Sundar got on stage and showed this thing called Duplex. You could say, “Hey Google, make me a reservation at a restaurant or a salon,” and then they showed this demo of the Google Assistant making a phone call to a spa and making a haircut appointment or something of that sort. I was like, “Oh my God, this is game-changing!” I’d never seen technology like this that sounds so realistic, that is very responsive, and really understands what the other person is doing. It’s not, “If you want to make an appointment, press one.” It wasn’t like that. It was natural; it was almost magical.
So, I came home that night, and I was showing it to my wife. I talked about the speech recognition, the natural language processing, the speech synthesis, and she kind of stopped me and said, “So, you’re telling me these brilliant engineers built a machine that can talk, and they made spa reservations? Thank you, Silicon Valley, for saving the world again.” You know, I kind of took it personally. I was like, “You can’t be messing with my guys and saying, ‘What are you guys building?’” I asked her, “What would you do?” She’s been in healthcare for a long, long time—she’s been in the healthcare back office for a very long time in home health and hospice at large health systems in pharma. She said at every single one of those entities, there are hundreds, if not thousands, of people that make and receive phone calls just to make healthcare work. She said it would be a better use of this technology if you were to take it and apply it to healthcare so that all these people who joined healthcare to serve patients, to be empathetic, to be there in the patient’s time of need, could be doing that instead of waiting on hold just to check on the status of a claim or the status of a prior authorization, or whether the benefits are covered. All those things are things that machines should do, and that was the genesis of Infinitus. It was really figuring out—it was through the insights of my wife—figuring out that there’s a huge opportunity to drive efficiency in healthcare, both in terms of the dollar amount spent on these administrative tasks but also really improving the time to therapy for patients, which was being delayed because you needed a human in the loop to do all these processes.
Alex:
Yeah, you sort of told us the incentive. Could you tell us a little more details about the product itself? And how do you see the company in a few years? Where are you headed?
08:46 Tell us more about Infinitus. What else is on the roadmap?
Ankit:
So, when my wife told me about this, I said, “Well, that’s silly. You’ve got an insurance company that clearly has a plan for the patient. Why does someone from a doctor’s office need to call the insurance company to figure out what the coinsurance is or what the prior authorization requirements are? Why isn’t there an API?” Right? That’s the first question that I asked. That’s the first question my co-founder asked me when I told him about this idea. That’s the first question every investor we’ve ever had has asked us: “Why isn’t this all over APIs?” The reality in healthcare is that there are thousands upon thousands of entities that need to coordinate with each other. So, whether it’s coordinating to agree on a standard in terms of how data should exchange hands or agreeing on the timeline on which something should be implemented, everything gets distilled down to the least common denominator. And so the APIs that do exist—the EDI [Electronic Data Interchange] rails that do exist—they are there, but they’re very basic, and it doesn’t work in many parts of healthcare. It doesn’t work in the world of specialty at all. I don’t think anybody makes a phone call to figure out what someone’s co-pay for an x-ray is or for a PCP visit is. But the second someone needs to be told they need to get on Prolia, which is a specialty drug by Amgen that costs tens of thousands of dollars for osteoporosis, there is no API that tells you the prior authorization requirements, or the step therapy requirements, or the co-pays, or coinsurances. And so the second you get into any of the major specialties in this country, a phone is picked up multiple times in order to get information, in order to start processes. And so what we’ve built is a platform that can automate parts, if not all, of those back-office communications that exist, whether it’s benefit verifications or prior authorizations. And my hope over the coming years is to go back to that original question that my co-founder Sham and I asked, which is: why isn’t this an API? So, as we’ve done more of these, we’ve done millions upon millions of these phone calls on behalf of around 80,000 providers around the country. The other side, the people receiving our phone calls, go, “Why the heck are my humans talking to your machine all day long?” Because we’re calling the insurance companies, the PBMs, thousands upon thousands of times a day, if not a month, and they go, “That’s my people spending time talking to your machine,” and I say, “They don’t have to be. If you give me the data digitally, I’ll get rid of the phone call.” Right? That’s the path we want to go down. There’s a little bit of a network effect here. The more calls we do, the more incentive and value prop there is for the other side, whoever that might be, to create the APIs, which really moves the ecosystem forward.
Alex:
Once you got into healthcare, did you have any “aha” moments, like, “Oh, that’s how it works in healthcare; this is completely different from any other verticals”? Did you have any of those moments?
11:47 Were there any a-ha moments once you got into healthcare?
Ankit:
You know, it’s interesting you ask that. I think there are many “aha” moments. The biggest one is I’ve not had a moment to think about how it would work in a different industry. The number of times someone comes to me and says, “Hey, can you do this for tech, or can you do this for real estate?” I’m like, “No, I don’t have enough time. There’s so much to do in healthcare.” So, the truth is, I don’t know how the same problem plays out in other industries. What I have realized—and I don’t think my co-founder or I ever thought about this when we started—was the fact that he was an early member of the Google security team or Snapchat security team has actually played to our strengths because of how important information security and data security are. And so what he was doing because he was just, his muscles were built that way from a software infrastructure perspective, when we went through and got our SOC 2 Type 2, or when we go through all those third-party risk assessments or information security assessments from prospective customers, things that other people trip over or say, “Oh, we’ll get back to you in six months when we’ve done that,” we’ve always had the ability to say, “Oh yeah, we’re already doing that,” and that’s made it a lot easier for us to build our business.
The same thing, by the way, is very true in the world of large language models. If I go back two years, when there was that ChatGPT moment, everyone got excited by it, but within six months, every major Fortune 500 company, the board was saying, “What are you doing with AI and LLMs?” The information security team was saying, “Ah, this stuff can go all over the place. We need all these guardrails.” And it was roadblock after roadblock after roadblock, which honestly, most companies weren’t able to get through, all those machine learning review boards, the AI review boards. And credit goes to my co-founder and our security team for being way ahead of that and saying, “Before we put this into our environments, let’s design the right guardrails so that our customers, who rightfully want to make sure data is protected, want to make sure the correct information makes it to the right person at the right time, have that comfort knowing that yes, this is a new tool in our toolkit that we use to drive efficiency, but in the right way.” So, the importance of security and data privacy—like we all talk about HIPAA, but what does HIPAA really mean versus how does it play out in a live environment, which is the sales process and ongoing controls that are in place—that’s been probably the biggest “aha” moment for me, which is some of these things that seem like blockers are really good to protect all of our data.
Alex:
Very interesting. And so your product is designed to automate many of the tasks handled by back-office administrative staff, and some of them don’t have a medical background even. As the adoption of your solution increases, how do you address the concerns of potential impact on the jobs of these individuals?
14:53 Your product is designed to automate many of the tasks currently handled by back-office administrative staff, some of whom do not have a medical background. As the adoption of your product increases, how do you address concerns about the potential impact on the jobs of these individuals?
Ankit:
Yeah, I think it’s something that we think about pretty actively. So far, we haven’t seen a single major customer of ours let go of people as they’ve adopted our technology, and I think that’s because of a couple of trends that are kind of tailwinds to this industry. One, the amount of specialty drug usage is only going up, so the number of patients that need these high-cost specialty medications or therapies is increasing. So, the number of support staff needed for the entire patient lifecycle is increasing. So, technology like this is allowing our customers to redirect that staff into other roles that they’re not able to backfill. And there was a report by CAQH [Council for Affordable Quality Healthcare] that came out a few weeks ago, and they said the biggest problem in healthcare continues to be a workforce shortage. And so it’s allowing the employers to plug the gaps from where they have a shortage but also be able to support more patients as they come into the system, especially in the specialties that we work in.
Alex:
I also want to say that I have this dissonance because I’ve personally experienced many times when a lot of hospitals and healthcare providers have not implemented just simple scheduling software, but some have started using some breakthrough solutions, like yours, for example. So, the gap between those providers is growing. What do you think should happen in order to narrow that gap? Or do you think those old-school providers are just going to be out of business in a few years?
16:41 I have this dissonance. There are still hundreds of hospitals and healthcare providers in the U.S. that still haven’t implemented simple scheduling software. However, some have already started using such breakthrough solutions as Infinitus. So the gap between these providers is only growing. What needs to happen to narrow that gap or do you think those old school providers will just be out of business in a few years?
Ankit:
So, I think it’s an interesting hypothetical question. The reality is that there are two kinds of scheduling, and scheduling, I think, is a great example, right? There’s very simple scheduling, and there’s very complicated scheduling. For people who are really sick, when they go into the health system, their day is packed. They go in for blood work in the morning, see Specialist 1, then they go for some other test, then they see Specialist 2, who’s going to review the blood work. Then they go do a couple of other procedures. And being able to do that jigsaw puzzle with the travel requirements and coordinating all those pieces isn’t something that simple scheduling can do. And this is another one of those things for me in healthcare, which is, for someone on the outside, a lot of things that seem easy, the second you put it into the workflow of a healthcare entity combined with the complexity of human life, suddenly something simple becomes very complicated. And getting that right is important because otherwise, we’re not going to make forward progress on that person’s health or that family’s journey in what is probably an anxiety-driven process in their life.
And so, the answer for most of these has traditionally been people, right? You have people like case managers or patient navigators, whose entire job is to help choreograph the journey of a patient through the system. And when a lot of startups get started, they’re creating a point solution, and they get initial traction in some of those early adopters that you mentioned. But in order to be able to really become pervasive and used by everybody, they have to think about and create the platform for injecting themselves into the workflow to make the workflow that much more seamless. You have to not only build the right technology for that feature, but also think about how it fits in from a data perspective. How does it fit in from a workflow perspective? How do you train the staff in change management? Someone once told me, “If you change the screens of the EHR in a pretty dramatic way, you can’t just go for a week and train everybody at a hospital on those changes because chances are, at any given point, 30% of the staff is not there that week.” Because someone’s on sabbatical, someone’s on a research leave, someone’s on PTO, someone’s sick. And so, when you change the user interface, just as one example, you have to constantly do education over a six to nine-month period, sometimes even longer, in order to get everybody there. So, the cost of change management is something that more and more people are becoming aware of and are pricing into the adoption of technology. And when they say, “What are the pros and cons of moving in this direction?” they’re pricing that cost of change management, and that’s why they’re focusing on the things that have the biggest bang for the buck. They’re able to take on the things that they say, “You know what? I’ve got one person who does scheduling for my office that costs me somewhere between $45,000 and $90,000 a year. If I move to a technology that on paper costs me $25,000, the argument is, ‘Ah, you’re going to save between $20,000 and $70,000 a year,’ but really I have to train somebody. I can’t get rid of my one headcount, so I still have to pay that person. I have to change all my systems. There might be things that break. Now, I need to have an IT person.” So, the total cost of adoption is a lot higher than what most companies think the cost of the platform is.
Sergei:
Yeah, that makes sense. Just switching gears a little bit. When ChatGPT was launched in November 2022, my first thought was, “Oh, you know, that’s gonna be tough for Google. It’s basically the end of search.” And then, like recently, maybe like a year and a half into this story, I looked at statistics in terms of market share; it didn’t make a dent for Google. You know, people are still Googling, even with the proliferation of LLM models. I’m just kind of curious why that is. Your experience—you obviously have expertise in building search engines. I’m just curious about your perspective. Do you think that one day AI-based search engines will replace traditional search?
21:26 A person who worked for Google for a long time, what do you think about the hypothesis that AI will replace standard search at some point?
Ankit:
Yeah, so I’ve got a few different thoughts on this. I spent a lot of years in search. I think there’s a technology discussion to be had, and there’s a distribution discussion to be had. So, let’s first talk about distribution. Google is the default search engine on billions of devices, whether it’s on Android, iOS, or even many desktops that are sold. And so that switch isn’t going to happen overnight. I worked at a startup where we tried to build a Google killer—a search engine to take over Google—and there were places where we were significantly better than Google, but it’s very hard to change someone’s behavior, especially when it’s the default behavior on their primary devices. People don’t clamor to change the default unless something else is significantly better. So, that dynamic plays some role here. I think whether Apple and others slowly unbundle Google and provide alternatives in a way that is seamless is to be determined. I think that has the ability to change market share. Otherwise, I think it’s very hard to change market share.
The second one, and I think this is the more interesting one, is technology. ChatGPT stole the moment in 2022. I think they have some incredible large language models, and they just announced SearchGPT, and they’ve been adding more and more real-time data around what they have. But that doesn’t mean Google’s been sitting around for the last five years—not two years, but the last five years—doing nothing. I think we assume that Google search today does not use a language model, and it’s traditional search versus LLM search. I don’t think it’s like that. If you go to Google, they’ve been having those one-boxes that understand what’s there in the content and give the answer as needed. Increasingly, you’re seeing those LLM-like conversational answers in Google search as well. If you go back in the history of search in the ’90s, Lycos and Yahoo had a list of web pages you could go to. It was a directory, right? There was your door to the internet. What we saw in the next generation of search engines, whether it was Excite, AltaVista, or Google, was they said, “No, no, no, let’s put a search box and let’s have the people search for what they’re looking for, and let’s give you ten blue links of what you’re looking for.” The generation after that, and this is the generation I joined search in, we said, “Well, search shouldn’t just be text. It needs to be multi-modal. You can search for text, you can search for videos, you can search for images, you can search for news. And the search results page doesn’t have to be just links and a snippet; it can also have images, because that will help the person know where they want to go and what they want to look for.” And all of that came into being in the late 2000s, early 2010s, but Google really took advantage of all that and built it into its own search page and made it richer.
Now, the question that we should ask is—sorry, after that Amazon did something very special, and I think most people don’t give Amazon enough credit for this, which is, Google was the one place we went and we searched products. Most people don’t go to Google to search for products anymore; most people go to Amazon or one of the other places to search. So, Amazon did build vertical search for shopping, which was better than Google’s, and that’s where a lot of people default to. So, the question that comes to my mind as we think about this new generation of technology being injected into search is, is it one that empowers the current incumbents, the Googles of the world, to ingest that new technology, the new kind of interface, conversational UI, into their current product, or does it empower a newcomer to say, “For these kinds of search results, come to me. For others, keep going to Google,” and that’s how you steal market share away? I don’t think that answer has quite been clarified yet.
Personally, I use ChatGPT probably a dozen times a day, but the kinds of things I’m using it for are very different from what I still go to Google for. When it’s a navigational query, when I’m trying to search LinkedIn, I don’t go to LinkedIn. I actually go to Google, and I search for that to find someone’s profile. It’s the easiest way to do it. On the other hand, if I want to take a message that I want to send to my team and write it in a specific way, I go to ChatGPT, and it helps me wordsmith parts of it, which, by the way, Google couldn’t do in the past. So, the question to ask is, are we seeing the dawn of the expansion of what we can do with these thinking machines, if you will, or are we moving market share? Does the market get bigger, and then the share changes, or is it the same size market, and you move market share?
Sergei:
Yeah, a lot of interesting thoughts there. I have my own thoughts. I have a Substack where I express sometimes random thoughts, so this is very interesting. I agree with you; Google obviously is not just going to give away its market share, and it has its own very powerful AI. I want to go back to healthcare, to AI’s role in healthcare. That’s kind of the reason I started. I came from areas like fintech and academia because I just see so many problems, particularly in primary care. When I started about five years ago, I was also excited because I thought I knew all these models and what they could do, but it’s a tough area. The applications I’m seeing are kind of similar to Infinitus. Correct me if I’m wrong, but there are a lot of things to be done for automation and streamlining administrative tasks, things like that. I’m curious if you think that AI could go beyond that, let’s say in the next five to ten years, if it could maybe fundamentally change the way healthcare is delivered and provided. What are your thoughts on that?
28:20 What do you think the role of AI will be in healthcare, particularly primary care, over the next 5-10 years? Will it fundamentally change how we deliver and provide care, or will it mainly be used for automation and administrative tasks?
Ankit:
Yeah, listen, I think—and I’ve talked about this and written about this in the past—my daughter, when she was born, had a very rare disease, and I think one of the things about healthcare is that it ends up being personal to everybody because either you or someone in your family or a friend goes through the system. There were dozens upon dozens of angelic people that helped us through those times, but there were still times when we felt lost, and we felt like we were the only ones. The question of “Why me?” comes in every household, right? When you find out you have a disease, it doesn’t matter if it’s as common as hypertension or diabetes or as rare as the hyperinsulinism my daughter had; we always ask, “Why me?” It’s almost like nobody else is going through it. I think AI and technology have the ability to be that partner to all of us as we go through the healthcare journey. Today, when we think about the generation of technology that exists, it’s almost always targeted at the business side of it—the provider, the payer, the pharmacy—because they’re ready to pay for it. As this technology matures, I think the ultimate impact it’s going to have in healthcare is to be that partner to patients. It’s to be that partner to their families as they’re going through this immense change in their life, to make them feel better, that it’s okay. It’s okay to go through this, and there are a lot of resources there. That’s just one that I think we’ve not even seen what that can do on the patient-facing side.
The other thing, and we’re seeing this—it’s not being talked about as loudly as it probably should be—is all the work happening in the drug discovery world, whether it’s digital twins, the ability to test hypotheses and different reagents, or being able to search through literature. There’s so much stuff that is being unlocked from an AI and technology perspective to reduce the time to more effective therapies, which I think will fundamentally change how medicine and drugs are in this world. If you can reduce the investment needed to come up with the next successful therapy, then you can reduce the cost of therapy. The reason the cost of drugs is so high is because billions upon billions of dollars go into R&D, and 95% of experiments fail. You have to make the money back on the 5%. But if you can reduce the cost of experimentation, you can get a higher success rate at a lower cost. That’s a win for everybody. I think there are patient-facing unlocks that are happening. There are drug discovery-facing unlocks that are happening. In the traditional healthcare system, I think there are going to be a lot more resources to help reduce burnout, support more patients, and just make clinical processes more efficient.
Sergei:
Makes sense. I have this question, and again, we’re on this podcast; we’re all brutally honest, so I’m just gonna throw some questions out there. You know, sometimes they’re not easy questions. The one that I got a lot of heat on social media for was this post, which I think was completely misinterpreted because, remember, you and I and Alex are, I think, in this industry to make things better and to actually advance innovations. But I said something along the lines of, “When we look at the history of innovation, there was nothing as big as AI, or, you know, you worked for Google, like search engines, or self-driving cars.” In digital health, there are a lot of companies doing great things, and they’re helping patients and customers, and that’s fine. But I haven’t seen—I guess I looked back maybe years, even decades—something really big. Again, you worked for Google; you know how those big innovations work. My question is, yeah, for instance, even with ChatGPT, I see a lot of companies, for example, in AI scribing, doing their APIs to GPT-3.5, and they repackage it, and voila, you have a product, which may be fine for some segments of the population. But my question is, do you think this is kind of the path where digital health is going, or could there be a different path to innovation in digital health?
33:06 Will digital health continue down the path of repackaging and infringement, or is there a new direction for true innovation in digital health?
Ankit:
I think your statement is true until it is proven false, and that’s true in any innovative space. I think the truth about innovation and truly transformational companies is that they’re built on the shoulders of other giants. Let’s look at fintech. I think one of the most meaningful fintech companies in the last 15–20 years has been Robinhood. What Robinhood fundamentally did was say, “I will build on the modern rails of fintech and offer something that none of the incumbents can dare to do.” We know the modern rails are more efficient than the ones in the ’70s and ’80s, so why is the trading model still monetized the same way as it was in the ’90s? They said, “Our cost of doing a trade is not much, so let’s offer trades at zero.” Now, this was very controversial, but they started getting a lot of users, and it forced all the incumbents to go, “Oh, if we don’t do this, we’re going to lose the market.” And so it changed the entire industry because of the courage of one founder, and two, the rails that were built over 20–30 years in fintech, right before it was even called fintech. I think the same thing is happening in digital health now. A lot of those rails are being built. A lot of companies are building building blocks, and it’s going to take someone who brings all those pieces together and says something like, “Hey, a modern insurance company should have a different business model,” or “A modern hospital or health system should have a different business model than a traditional one.” But that’s not going to be possible until all the rails are built. I think we’re in this exciting part where the rails are being built; no one knows how it’s all going to come together, but it’s being built together, and that’s what makes it exciting. But there’s going to be that moment when we go, “Oh my God, it all came together, all that work for 20–30 years pays dividends.” Self-driving cars—we know about Waymo and Google’s self-driving car project for 15 years; it’s not overnight. It was 25 years of research before that brought it to where it is, so it’s 30–40 years. I think digital health is still 8–10 years old in that, and now we’re getting the investments in these things. Along the way, you’re going to see 90% of efforts fail, and that’s normal, right? I think one of the things that Silicon Valley gets right is that they’re okay with failure. It’s okay to fall down and get back up and try again. That’s unlike anywhere else in any other ecosystem, where people really take success and failure to heart. Here, the most successful people try again; they’re not satisfied because that’s just in their DNA. And the people that fail don’t just go—you know, some of them get into downward spirals, but most people really get back up and keep going because they go, “The world’s not ended; I’m going to try again.” I think that’s what’s special, and I think we’re very early in digital health, so I think it’s fine. You want the detractors to be sitting on the sidelines and throwing popcorn and eating it, but I think there’s going to be that magical moment and that magical year when things come together, and we’ll all look back and go, “Do you remember the time when we had to fill out a clipboard every time we went in, and the doctor was always running late?” I truly think in the next 10–15 years, that’s going to be a thing of the past.
Alex:
Yeah, I hope we’re going to witness that moment, hopefully within the next five or ten years. I was going to ask—you probably know that there are two schools of thought on AI. The first one says that AI will scale and skyrocket. The second one says that AI costs will outweigh the benefits, the so-called Jevons paradox. Which side are you on?
37:08 There are two schools of thought on AI: 1) AI will scale and skyrocket, 2) AI costs will outweigh benefits—Jevons Paradox (Goldman Sachs, Sequoia Capital). Where do you stand between these camps?
Ankit:
I’m on both sides. I think in the short term, AI’s costs are very high, but I think with all things technology, over time, costs come down. And I think at that point, AI will scale and skyrocket. I think we’re still, again, in the very early days of figuring out what these models do. You know, for what it’s worth, I don’t think AI has—it’s not a moment in time when it changed; it’s been improving year over year over year. You had the breakthroughs with the image models with the convolutional neural networks, and then you had the larger language models. You’re now seeing new sub-quadratic models that are able to do things in ways and at costs that weren’t possible in the past. And I think that innovation will continue. It’s not like we’ve come upon the final model that’s going to rule them all. I think innovation is going to continue. I think in the short term, in many places, the costs of AI are very high, whether it’s from a climate perspective or from a real dollars-and-cents perspective. But I think it’s important with any problem to have a stake in the ground and say, “This is where we are now,” because the second you put that, then everyone figures out how to optimize it in the right direction, rather than talk hypothetically and not make progress.
Sergei:
I want to ask a question about venture capital in digital health. Vinod Khosla recently said that 90% of VCs add no value to startups, and 70% even harm them. I think he was almost—just from my own research—he was almost talking about digital health. Just in my very general observation, and of course, there are exceptions to this, but I think my personal opinion is that a lot of venture capitalists look at healthcare just the way they look at manufacturing or finance. It’s just another area to maximize ROI. And in my opinion, there are a little bit more dimensions to that, and as a result, I believe we see a lot of what I call pump-and-dump, where VCs push one particular startup and then exit, leaving customers sort of in the dark. So, I’m just curious—this is kind of maybe a little critique of venture capital in digital health, but I’m just curious what your thoughts are about the role of VC in digital health.
39:47 The role of VCs in digital health
Ankit:
Listen, as a former VC, I appreciate the nuance and the feel-goodness that a lot of people try to portray online, how venture capital is a way to change the world and X, Y, and Z. And I will say, the top 10% of venture capitalists truly are in it to change the world because they’ve achieved a lot of other things. They’re trying to go and change the world. Along the way, they want to make sure that they also have tremendous financial returns. The other 90% are doing their job as a capitalist—let’s be very real. Their job is to take in a dollar and return significantly more than a dollar. So, whether it’s a pump-and-dump or backing their portfolio companies in ways that some might question, that’s kind of their job. They’re being given management fees and carry in order to go create capital out of capital. I wouldn’t blame them for doing that. I think some are more enlightened in some ways than others—they understand the dynamics better. I think Vinod, in that statement, is probably right in that a large percentage of VCs probably harm their portfolio companies more than help them in achieving true impact on the world. The question is, are you optimizing to make money, or are you optimizing to make an impact? Very few can balance the two and have both of that impact. So, I think it’s important to figure out where on that line you are.
The other thing—I think if 80–90% of startups fail, it’s probably a good thing, right? It sounds painful, but if 100% of startups are succeeding, then we’re not taking enough risks. Part of the reason we’re able to move the world forward is because people take risks. The fuel of venture capital is meant to enable people to take risks, be okay with failure, experiment, and then move forward with the successful ones. Double down on the successful ones and back off from the experiments that aren’t working. So, like any other industry, there’s 10% of absolutely exceptional venture capitalists that not only add value but really add jet fuel to companies to move forward. There are probably 20–30% of average ones, and then there are others that, you know, every ecosystem has those people hanging around there.
I think we’ve had quite an exceptional set of venture capitalists that understand not just technology and AI but also healthcare, be around our table. I would think our extended team—not just me—is thankful for the value they’ve added to us and the direction in which they’ve pushed us, the way they challenge us. It’s a very healthy discourse that we have both in the boardroom and between the team and our investor base when we ask them for advice or help, that conversation. I think it’s important to have a partnership. It isn’t about a hierarchy of VCs or startups being at different levels. I think that’s one of the things you often imagine after reading Twitter or reading some of the posts that are online. You almost feel like one side is more important than the other, but the truth is, the two sides very much need each other. VCs and startups need each other to survive and need a healthy partnership to create value.
Sergei:
Yeah, I agree with many things you just said. Like with failures, for instance—absolutely. You miss every shot—what is it? You miss 100% of the shots you don’t take, right? So, it’s absolutely part of the process, and it’s actually part of innovation. What I’m kind of—and again, I don’t want to dwell too much on that—but my problem is that, again, from kind of an outsider’s perspective, but also from doing some research on that, is that in certain cases—and again, the examples you’ve shown are definitely credible, and there are great people out there, and digital health obviously needs capital. We need to finance those great ideas. My only problem is that I feel like oftentimes, a venture capitalist would take a startup knowing that the product has issues, even outside of being non-profitable, which is a completely separate issue, and they try to get to the exit, even with the knowledge that it’s too early to get to that IPO. But for them, it’s an exit, for them, it’s ROI. So, I think there’s a bit of a conflict sometimes, or oftentimes, especially in digital health, between what’s good for the company and what’s good for the VC firm. But I take your point that there are obviously VCs who are thinking about healthcare, who are thinking about the medical community, and there are exceptions to the rule.
Just a quick follow-up: what do you think about this whole issue that—I think I was looking at statistics, not for digital health, but just for technology in the ’90s. In the 1980s, about 90% of IT firms that were going IPO were profitable, and now it’s the other way around. It’s actually only 10%, and I think the number is even lower for digital health. Now you just basically find an investment banker, an underwriter. Venture capitalists can approve that because for them, it’s an exit, oftentimes. Is there a path to profitability in digital health, and what are the obstacles that you see?
45:57 Is there a path to profitability in digital health and what are the obstacles?
Ankit:
So, I think in the last couple of years especially, there’s been an increase in thinking through what is important on the financial side of most startups and how you get to that profitability. In most boardrooms that I’m aware of, that conversation is pretty actively happening, which is: how do we get to that point of profitability and self-sustenance, so that you can think about venture capital or additional capital, either through PE or through an IPO, to be jet fuel to grow inorganically rather than growing organically? I think it’s very healthy that we’re having those discussions increasingly. I think that the biggest challenge for digital health startups is that most of them never get to a scale that matters from a profitability perspective, from a venture profitability perspective. If you’re making $10 million and profitable and growing at 5% year-over-year, that is a great business for a founder, maybe even for the team. You can dividend it out. But from a venture capital perspective, if you’re investing out of a $500 million fund or a billion-dollar fund, it’s never going to have the ability to make a difference to your fund. So, this is why venture investors are pushing for the fewer companies that have the potential to grow into hundreds of millions of dollars of revenue to get there and then eke out that profitability. Because if you don’t get to the hundreds of millions of revenue, profitability doesn’t matter as much to the investor ecosystem. It can mean very different things to the founder ecosystem, and I think this is where there’s a lot of debate out there about whether every company should be venture funded or not. I’m a big believer that there’s a whole class of great companies that should never raise an ounce of venture capital, or only raise a seed, pre-seed, or Series A, and then say, “We’re now on the path to profitability,” and grow that company in that manner. I think the second you raise hundreds of millions of dollars, it’s very hard to say, “I’m going to become profitable at $10 million and then grow at 10, 15, 20% year-on-year,” because it just won’t—the expectations and the return profile won’t make sense. I agree. In the last three or four years, there are a lot of companies that have raised more money than they should have, which are now in this weird spot where they still believe it’s growth at all costs without thinking about profitability. But those are two very different classes of companies, and the public markets, for what it’s worth, still reward the companies that have the right metrics. So, you’re seeing the digital health companies that have good economics being rewarded, the few that there are, and the ones that are losing money and not growing fast enough really getting hammered in the public markets.
Alex:
Yeah, and I think you’re right. Conversations like this should actually bring some change in that space as well, and hopefully, it’s going to shape for the better. Traditionally, at the very end, we ask our guests to give some sort of advice to startup founders, and in your case, specifically startup founders who got into AI in healthcare in the last few years. What would you recommend them to do or not to do?
49:12 Ankit’s advice to startup founders in digital health and AI
Ankit:
It’s really easy to get excited by the new shiny object, whatever it may be. Every couple of years, there’s something shiny that the ecosystem sees, and everyone runs in that direction. It’s important to do justice to new ideas by letting them prove their merit before getting too excited. As a founder, it’s important to be optimistic; it’s important to want to change the future. But it’s also important to be realistic and understand how it fits into the workflow and fits into the total ecosystem that this solution has to be a part of. I think too many people jump to, “I’ve got another solution,” which actually just adds to the complexity of healthcare rather than simplifies it. I think we would be doing each other a disservice if we added more complexity to an already over-complex healthcare system, and we should really focus on simplifying things in an effort to have that long-term impact on healthcare. Simplification is probably the best product that we can, as an ecosystem, have for the industry.
Alex:
Great advice. Thank you so much, Ankit.
Ankit:
Well, thank you for having me.
Sergei:
Thank you, Ankit.
[Applause]
👉👉👉👉👉 Hi! My name is Sergei Polevikov. I’m an AI researcher and a healthcare AI startup founder. In my newsletter ‘AI Health Uncut’, I combine my knowledge of AI models with my unique skills in analyzing the financial health of digital health companies. Why “Uncut”? Because I never sugarcoat or filter the hard truth. I don’t play games, I don’t work for anyone, and therefore, with your support, I produce the most original, the most unbiased, the most unapologetic research in AI, innovation, and healthcare. Thank you for your support of my work. You’re part of a vibrant community of healthcare AI enthusiasts! Your engagement matters. 🙏🙏🙏🙏🙏