Siddhartha Ahluwalia 0:53

So I’d love to start by the letter that you wrote to your team that Pat Grady shared on Twitter, that you’re looking for people who are almost like heat-seeking designers for pain. Can you elaborate more about it?

Tanay Tandon 1:06

Yeah. We got a lot of inbound interest on our engineering team, on our sales team after that. And the core concept of heat-seeking designers for pain is you’re looking for people who work at startups really that are addicted to solving problems and addicted not just to the successes that come with that, but all the pain and the messiness of the day. Because pain is signal. At the end of the day, if there’s customer pain or if there’s internal pain, if there’s pain that you see with certain team members, it’s a problem that needs to be solved. And we create enterprise value in society by solving problems. And the ability to detect pain early and almost run after it, seek and destroy, is actually fairly meek. Not everyone has it. A lot of people try to put their head in the sand and move away from pain. And when you find someone who actually is addicted to pain, they tend to perform really well at startups. And they tend to make great leaders. They tend to grow very quickly. And our thesis has always been find these people and make them celebrities in the company, continue to grow and cultivate them.

Siddhartha Ahluwalia 2:16

And what are the other aspects that you look for while building your team, culturally and skill-wise?

Tanay Tandon 2:21

I think, naturally, some people have urgency. Some people are addicted to speed in the same way that they’re addicted to pain. One of our core company values is speed above all else. And there’s a certain type of person who can operate with rigor, with very high throughput and horsepower, while also moving very quickly. And the people that are naturally addicted to urgency and want to move fast are, again, amazing fits for companies. I think that people that are very good at questioning requirements, usually in healthcare or any industry, there’s a long list of steps before you can do anything. And with any manufacturing process or really any process out there, step one or step zero, in some ways, is question all the requirements. It’s like in the Ewan Manifesto of how to simplify manufacturing processes and make them very efficient for scale. It’s the same in software, and it’s the same in healthcare, which is question the requirements, simplify the problem stack, and then move with urgency and surge resources in a certain area and move as quickly as possible to make that happen. I also think that people that are good natural writers do well because in a chaotic environment, you need to be able to distill your thoughts, present them in written form, and then really disseminate that across the org as quickly as possible. And some people can do that very naturally, and some people can’t. And the people that can tend to also grow in an org very quickly because their way of thinking, their processes, their SOPs turn into the Bible of the org and turn into what turns into the day-to-day operating system.

Siddhartha Ahluwalia 3:57

Let’s talk about the US healthcare market. Like infinitely large market, probably a few trillion dollars, and extremely messy, and probably dominated on the technical side by one player, Epic. And Epic has been guarding the systems for a very long period of time. I think their revenue is what, like few tens of billions?

Tanay Tandon 4:23

Approaching 10 billion, yeah.

Siddhartha Ahluwalia 4:24

Big business. Approaching 10 billion, completely privately owned, no external investor. And it’s a rumor that any CIO that gets hired in large hospitals in the US is through a recommendation by Epic. So what’s your view?

Tanay Tandon 4:42

First of all, it’s a remarkable journey. You look at Judy Faulkner, you look at the way that Epic was built in the last 30, 40 years of just healthcare IT in the country, and it’s dominated by this one company. And the learnings in some ways are, there’s a lot of hate at times for the product, but at the same time, there’s also a lot of love. It solved a very specific problem. It took paper records and it digitized them. It made compliance easy for the CIOs. It made record keeping and accounting easy for the CFOs. And it provided all of those solutions in one stack. Our belief is that there’s a lot of parts of Epic, however, that aren’t built for the AI age and aren’t truly AI native, which is it’s not just people that are doing these tasks, it’s agents. And it’s not just traditional labor that’s sitting behind a screen. It’s software that’s reading large amounts of information, making decisions, and then working with health insurance companies, summarizing clinical documentation. And you need this orchestration layer that almost lives on top of the EMR that can conduct those tasks. The other point that I make is Epic is massive, but there’s so many other companies in the industry that are actually approaching similar scale or Cerner actually does more revenue than Epic. And you look at like Athena Health does billions in revenue, and then you have ECW, and then you have this long tail of EMRs. And so Epic is a chunk of the market, probably like 30s to 40%, but it’s still not a majority of the market. And the story, I think the real story of healthcare in some ways is the rest of the market that’s not even on Epic, that’s on very disjointed software solutions. That’s cobbling together patient engagement, intake, revenue cycle, documentation software. And our belief is that with language models, you can automate a lot of that. And you can almost leapfrog the standard digitization process and take something that is truly agentic, and take a bunch of labor and turn that into software as well and improve the operating margins of the system. And I think that’s what’s happening right now. That’s really the story of, you know, if digitization and Epic and the EMRs was the story of the 2010s, the story of the late 2020s in healthcare is definitely one of language models and AI replacing standard labor or augmenting that labor to make it much more efficient.

Siddhartha Ahluwalia 6:59

So the US healthcare almost spends like a trillion dollars?

Tanay Tandon 7:03

In on labor or on the way that I think about it is probably like total healthcare GDP is approaching 5 trillion. And that includes the services and the clinical work and the, hospitalizations and whatnot. And then your admin labor on that is probably between a trillion to trillion five. And then if you just look at what we call work tax or admin bloat and admin expense, probably about half a trillion dollars. So there’s like $500 billion worth of exhaust that the system creates. That is documentation, that is revenue cycle, that is claim submission, that is denial of appeals, going back and forth with insurance companies that doesn’t really exist in other industries in the same way. When you swipe a credit card, there’s an adjudication process. There is an interchange process. There’s a bank at the end of the day that’s processing that information and instantly releasing the funds. That system runs at 2% to 3% cost of transact. Healthcare probably runs at 15% to 20% cost of transact. So there’s so much room to make it more efficient and look and feel more like a credit card swipe. And that really is the work that we do and many other companies in the space do.

Siddhartha Ahluwalia 8:09

Got it. And what is the biggest bottleneck that why it has not become efficient till now?

Tanay Tandon 8:17

It’s a great question because there’s so many reasons. And one, it was sort of a tragedy of commons or almost like a bystander effect where as the bloat accumulated, it was never in someone’s total incentive to solve it. There were always more urgent problems. When a pandemic happens, it’s go, go, go. Just use whatever we got in the system and pile on top and just get the job done. And I think our health system has now been under duress for 20, 30 years. And there was never really time to stop and think and try to fix these problems. And it was just between a pandemic, a chronic disease crisis, the obesity epidemic in some ways, the growth of an aging population, chronic disease, the system has just been under stress for so long. And COVID in some ways was this breaking point where we really burnt out our providers. And I would go so far as to say one of the reasons that healthcare has been the fastest adopter or amongst the fastest adopting industries of AI is because of that massive workload that was placed on providers during COVID. And now it’s just, there’s no option. You either adopt AI or you lose your physicians, you lose your nurses, you lose your admin staff. So we’re excited because it’s an amazing point in time where you can actually give relief to the system using a tool that was just created in the form of artificial intelligence and language models. And in some ways, everything just lined up perfectly for the healthcare problem to be attacked.

Siddhartha Ahluwalia 9:51

Well, let’s say on an average, a healthcare provider, physician, or a nurse usually works to 8 to 12 hours a day, right? On how much of their time they are spending in these admin staff filling out the forms, filling out patient records?

Tanay Tandon 10:05

Yeah. I think in some cases, I would say 8 to 12 hours is a very good day for a nurse or a physician. We see sometimes these 14, 20-hour shifts and it can get pretty brutal. And in the conversations that we’ve had and the customers that we’ve worked with, you’re spending almost 20, 30, 40% of your time. So hours every day on documentation, on paperwork, on filing appeals, on responding to an insurance company. So it is a huge chunk of time. And I would say in the worst case, in some of the highest acuity cases, you’re seeing almost 50% of the time being spent on administrative work, which is just unsustainable. That’s not a good use of… The thing about clinicians, these are some of the smartest, most trained people in our country. They should be spending every minute of their day face-to-face with a patient, taking care of a patient, doing the thing that they were trained to do and love to do, as opposed to this administrative exhaust.

Siddhartha Ahluwalia 11:00

Yeah. And let’s say there are so many solutions today. You said COVID was an acceleration point that accelerated the solution. But still, entrepreneurs say that healthcare is one of the hardest market in the US to get demos in, to get pilots in, to get paid contracts in.

Tanay Tandon 11:18

Right. I think it is a hard market. And in some ways, it’s the only market I’ve spent time in. So maybe I just don’t know any worse or better. But the learnings, I would say, if you contrast healthcare sales in the late 2010s, like 2017 through 2020, to healthcare sales post that in 2021 onwards, two completely different markets. One where you have a buyer that is really just feeling the pain viscerally of the administrative tax and is ready to make changes. And I’d say technology capabilities in the form of language models that are also just so magical that there’s almost no option but to adopt these new capabilities. So it’s a very different market than the one that came before. It’s still hard. And it’s hard because it is relationship-driven. It is distribution-driven. It is very top-down. If you know the CIO and have cultivated a strong relationship with him or her, very different process than if you don’t have that. And so, I think in some ways, counterintuitively, my recommendation to early-stage founders is, unless you have that distribution advantage, healthcare is still actually a very hard market to get into. And you’re better off either working with a larger company or distributing through a relationship somehow. Because if you go into it raw, you will lose to the players, the legacy players that have the trust in the distribution. Because healthcare isn’t really for experimentation. At the end of the day, they need solutions, but they want them from durable, long-standing providers who’ve been in the space for a while.

Siddhartha Ahluwalia 12:55

And overall, what will be the software spend in the U.S. on healthcare?

Tanay Tandon 12:59

That’s a good question. There’s many ways to size it. And I think about because you have like clearinghouse software, and then you have like straight software, and then per-seat licenses, what is clear is that in aggregate, the spending on software and software-related expenses is probably like somewhere between like 70 to 80 billion bucks. That itself is not massive. And if you’re just going after that spend, I actually think the problem is less interesting. What makes it interesting is the half a trillion in labor spend that is attached to it. You have $500 billion worth of administrative labor that is sitting on that software, moving their mouse, moving files from one place to another, fixing claims. And that’s really what we get excited about and what our customers get excited about. Because if you solve that, one, you get paid faster. The claims come through faster. But two, there’s a ton of labor in your system that you can reallocate to clinical problems or revenue-generating problems, as opposed to purely back-office problems. So the $70-$80 billion pure software number, interesting. I would argue that the half a trillion dollar labor number is just much more compelling and is also the real problem. These health systems are not becoming unprofitable because of their software spend necessarily, unless it’s a massive line item or a big migration. It usually is because of the associated labor spend or the dollars that are just leaking through the funnel when insurance denies.

Siddhartha Ahluwalia 14:23

So when an entrepreneur is trying to find a building point solution versus building a platform, what would be your recommendation?

Tanay Tandon 14:33

It’s hard because you have to start somewhere. And when you look at our journey, we started with what effectively looks like a point solution. And to this day, many of our hooks into these systems are specific point solutions like ambient scribing or autonomous coding or intake automation, or the voice agent platform that’s one of our fastest growing products ever. It takes over the front office and call center. The issue in some ways, though, is if you’re a one-trick pony, you’re not going to get the biggest customers in the country. The biggest customers and most important customers in the country want a platform partner. They want someone who can come in, customize the solution to their needs before deployed engineering, and then really tie that solution to a suite of other products to automate more than just one thing. Because the implementation time is kind of fixed. The relationship building time is kind of fixed. And you don’t want to put all that time in for this one little solution. The ROI case will never make any sense. The CIOs and the CEOs and the CCOs that we work with are all thinking about this problem over five years. I’m making an investment in a partner. And then over five years, they’re going to take over bigger and bigger parts of the system. So often when we’re demoing, we’ll engage on one or two product lines. But the roadmap is always, we’re going to take over the call center, we’re going to replace the revenue cycle, the back office. And so the ACB expansion story for us is fascinating. And more importantly, the labor ROI story for the customers is remarkable because they put in the fixed cost once and then the software platform expands and augments more and more labor.

Siddhartha Ahluwalia 16:02

So I assume today there will be many customers who would be paying you eight or nine figures.

Tanay Tandon 16:08

We have customers that are paying us tens of millions of dollars. We have a few customers that are approaching $100 billion a year. And then we have a lot of customers that are in the low single digit millions. And then hundreds to thousands of customers that are below that. And I think the beauty of our platform is that it serves small practices, which are called like one to 20 provider groups that are physician owned. And then those same software solutions scale to large massive health systems. And there’s more customization required and more implementation time required when you work with a large system. But our long term belief is that the best software products, you look at Microsoft Word, you look at Google Docs, you look at many of the things like cursor and code editors, they really do scale across the gamut of organization types. And with LLMs, I think there’s a once in a generation opportunity to reimagine the operating stack for all healthcare practices, small ones and large ones. And that Microsoft Office suite is being built now. And we really think that our suite is the one that will be that answer.

Siddhartha Ahluwalia 17:16

For today, how much of the market you would have, I would say, captured or taken or serving?

Tanay Tandon 17:23

Yeah, it’s, we have, you know, these numbers up in our office. And I think about, so the first number is what percentage of all US healthcare appointments run on a platform or, you know, a tooling of ours. So for ambient and charge entry and charge capture, it’s about 7% of all healthcare visits in the US run on our platform. Now, the more exciting metric in some ways, but also at times, the one that my team gets more upset about is the number of dollars in terms of full cycle RCM, or we consider our software solution to be fully penetrated, is like 0.1%. And that’s exciting, because we’re a company that’s doing hundreds of millions of dollars, it’s almost unheard of to be doing hundreds of millions of dollars in annual revenue, and only be 0.1%, 0.1% penetrated in your in your total time. And so the exciting part of that is a great well, 7% of us appointments run on ambient and some of our other tooling, that’s a great launching ground to then hook and attach the rest of the revenue cycle platform onto. And just in our current customers, you know, if we, we stopped selling any new logos, and all we did was continue implementations and continue expanding those accounts, probably like 10 to $15 billion worth of software revenue for us to go implement and expand into. And that gets me excited. And then if you play it out, and you acquire more net logos, like we will, of course do, that number just continues to grow.

Siddhartha Ahluwalia 18:50

Got it. So, I would love to learn about your journey, right? You started as a medtech device, went to YC, and, you know, then, you know, got some of the legendary investors on your cap table.

Tanay Tandon 19:07

You know, it’s been a very interesting journey. So, I grew up in the Bay. I was a research intern at Stanford AI in high school, and I worked under Richard Socher, who’s a professor at Stanford. And then he spun out and started his own company called Metamind, which was eventually acquired by Salesforce. I was a software engineering intern there, and then a research intern again at Stanford AI in the summers after. And I learned so much in, you know, from him and from the teams at Stanford, because, you know, really, I tried to be a sponge, just come in and see what was happening. And I remember when I was an intern at the Stanford AI lab, it was really when this like first sequence of GPU explosions was happening, like 2013, 2014, 2015. And it was clear that accelerated computing would change the world in terms of self-driving, in terms of image recognition, and even like early examples of language modeling. And I, you know, this was fascinating. And one of the first applications that we saw was computer vision in pathology and diagnostics, because computer vision models actually got really good at tasks and, you know, human equivalent or superhuman well before language. Language took a lot longer and some algorithmic improvements and some scaling improvements, but imaging got there early. And our belief, you know, me and Deepika, we used to compete in high school science fairs, and our belief was that you could take a blood test, the complete blood count, which was done through flow cytometry or microscopic pathology, and replace that with computer vision. And so, we built the prototype, and it was actually quite simple. It was like an imager, like a microfluidic test strip, which was a channel with some, you know, staining in it. And then we’d take hundreds of images, CN convolutional neural nets would recognize the cells, classify them, count them, spit out a result. And that was the prototype, and that was the device. And we were like, great, we can, someone might have use for this. Like, it was a blood test. Like, why wouldn’t you use it? And that kind of kicked the journey off. We demoed the, we got into YC based off of this prototype, and then we demoed the product to Alfred at Sequoia and the team there. And I remember Alfred said something like, this is fascinating. Like, you know, as a team of two, you guys have accomplished a lot. We don’t really know that much about the med device space, and Sequoia isn’t traditionally a med device investor, but we really believe in you guys and the concept of applying artificial intelligence to healthcare. And I remember at the time, I was like, sounds like Alfred doesn’t believe in my medical device. Like, it sounds like he’s telling me that we might have to pivot or turn into something else. And Alfred was right. The company transformed in many ways over the coming years. But that core thesis of build products that take artificial intelligence and automate rote tasks in healthcare was the core of the business. We got the med device cleared, Sequoia invested, let our seed round. We got the med device cleared, launched scale to millions in revenue, and then decided that once we were in these practices, once we were serving these doctors and in these clinics, we saw the real pains of running a healthcare business. And yes, diagnostics was part of it, but the clinical side was actually well covered. In general, the clinicians had the tools they needed for the average patient to get the job done. The real problem was administrative. It was in how they ran their telehealth operations and how they did their billing and how they fought insurance companies. And I think as a team, we decided, hey, we did something really hard. Let’s go do something harder. Let’s actually build the whole suite. Let’s build the rest of the solution set and bring it to all the practices in America.

And that’s how the solution expanded from one device to the software platform. And then as the journey progressed, Hemant from GC, General Catalyst, led our series B. We became very close with the fund as they introduced us to a lot of customers and partners. And then we had this very interesting opportunity to take this Commure, which is a GC incubation focused on health system transformation, and merge it with Othellus and turn it into really this larger company focused on bringing AI into all healthcare practices and all hospitals in the country. And that merger is one of the best decisions we’ve made so far. It’s accelerated both businesses and it’s brought our technology to some of the largest hospitals in the country. And that’s really been the summary of the journey so far.

Siddhartha Ahluwalia 23:32

Yeah. What have been some of the highs and lows during the journey?

Tanay Tandon 23:35

So many. I would say that it’s almost daily. You’re just taking punches. And that’s part of the reason I wrote the heat seeking missile for pain essay is that you really do have to normalize pain. You have to normalize what it means to be in a state of pain very, very often, because even in your best quarter, and when you’re putting up the best numbers, there are more bad days than good. And the good days are these flashes of brilliance where sign a big deal. And then you realize, well, there’s 10 other deals that are stuck, or this deal is going to be really hard to implement. And I think running a business and building products is such a hard, painful endeavor that it makes sense to codify that as like, well, we have to be heat seeking missiles for pain. And everybody in the company needs to be a heat seeking missile for pain. So some of the lowlights in particular, it took a long time to get our FDA clearance. You know, it took almost two years longer than we initially expected. In hindsight, that was how long it was going to take from the beginning. That’s how long the FDA takes to review processes. And I just assumed that our first application would be perfect, we’d get an approval, and then we’d be off to the races. Wasn’t the case at all. It took so many iterations back and forth, visits to DC, engaging with the regulators, redoing the clinical trials four or five times until they finally approved it. So that was every time they rejected it, it was painful. The first time they rejected it was super painful because we got this like 40 page document telling us about why we were done. Later on, I think the way things got painful or hard were, you know, if you have a tough quarter where like sales just the engine isn’t revving, and you have to really diagnose the problem is product, is there something wrong with, you know, our the way that we’re marketing and top of funnel. And that happens, you know, as recently as like a year ago, like, we’ll, we’ll, we’ll feel those pains. And, and then I think some of the other lowlights are just, sometimes you ship the product, and there are serious like gaps between what the customer needs and what you’re what you’re doing today. And the only way to find that out is to be on the market and take that punch and then work with the customer to solve the issues. And, you know, I really credit our work with HCA is one of the largest systems in the country for being willing to work on the hardest problems with us, like co develop an ambient solution for emergency departments and hospitals in acute settings that are not the easy, you know, low hanging fruit. But because of that, I think that’s why we have the best ambient platform in the country and the fastest adopted ambient platform in the country. It’s because of those investments in the hard things.

Siddhartha Ahluwalia 26:04

But healthcare in general in the in the US, right, has not been a very attractive space for entrepreneurs to build. Low number of IPOs, for example, in this space, low number of acquisitions. Right. So, so why do entrepreneurs and investors keep on investing?

Tanay Tandon 26:26

I think the one it is, it’s important, it’s one of like the, you know, if you think about one of the persistent problems in the human experience is health, and the way that we deliver healthcare, and it’s a massive chunk of the economy. And so I think it’s a noble place to spend time and invest in, because eventually, we’re going to nail it, and we’ll get it right. That’s just the way the way these things work. I also think that, historically, the reason some of the outcomes have been relatively smaller, is they were going, they were only able to really go after that 70 $80 billion time that we talked about in aggregate, which is the software spend in the segment. And now the rest of the time has opened up. And so it’s a very different ballgame when you’re going after half a trillion dollars to spend, maybe even a trillion dollars in spend, versus really being pigeonholed in that 70 $80 billion software segment. And number two is, it’s a low operating margin business historically. So the customers, unlike, you know, tech companies, don’t have a lot of dollars to reinvest in, in, you know, their own operational productivity. So that flywheel has never existed, where we have a lot of cash, we’re going to invest it in more tools and technology to make ourselves more efficient, and then we’ll grow and grow and grow and then expand. And because of that, most of the growth in the healthcare sector has come from PE software, which is go build these, you know, empires by buying up hospitals and buying up practices, roll them together, and then internalize the, you know, cost savings at the operational level. You can now run health systems like a technology business, which is super exciting, because the labor, the big chunky labor piece of it can turn into software. And when that happens, the flywheel emerges, and the flywheel emerges, spending behaviors change, the TAM increases, the quality of life for the patients and the providers improves. And I think that will be the trend of the next 10 years, where like health tech companies really become technology businesses, as opposed to complex operational businesses that they’ve historically been.

Siddhartha Ahluwalia 28:25

Yeah, we need few of the companies like Commure to become 100 billion businesses, as Hemant called it, you know, for the new innovation to come up for it to become attractive for entrepreneurs and investors to put more effort and money into.

Tanay Tandon 28:38

100%. You need big wins in the space to kick that off. And, and also just like show the way and show the path. Because what once the other truth is, like, as we expand in these, you know, in these large health systems, we become an attaching ground for net new software. And that’s not been the case with the historically on prem solutions that are very hard to integrate with and play ball with. Interop is a core thesis within our product, anyone can build on top of our EMR, anyone can build on top of our ambient product, versus the historical tools that these hospitals.

Siddhartha Ahluwalia 29:10

Yeah. Our portfolio companies have really found it hard to integrate with Epic. APIs are not cleanly available, transparent. Why do you think is the case? Is there not enough incentive for companies like Epic to open up to collaborate?

Tanay Tandon 29:26

I think it’s a question one of incentives. But at the same time, I also think it’s priorities. If you think about where does where does Epic spend their engineering resources and dollars? What do all EMR spend their engineering resources and dollars, they have 100 requests from every customer that needs something really specific built out a certain type of report a certain type of tooling. And so these are often like life or death, either for a patient or for the system itself on a financial from a financial standpoint. And that’s what they’re going to prioritize instead of building out a perfect interop suite. And so you almost can’t blame them because they’re just serving the needs of their customer. Now, there’s some good tailwinds in the space. One is regulatory, which is there is a lot more regulatory pressure to make EMRs interoperable and a lot more requirements that EMRs have these open accessible APIs. And that’s good because now it’s a kind of standardizing enforcing function. Number two is, I think the concept of agents where you don’t really need a clean API per se, you can make do with unstructured data and take inputs and shoot out outputs that are processable, human readable, but also machine readable. And that to me is going to be a big megatrend where the concept of an API is changing very quickly. And so those two things paired together create a very strong forcing function for faster interop adoption, and then also just a flurry of companies in the space that can build on top of EMRs that couldn’t do so before.

Siddhartha Ahluwalia 30:56

Got it. So you think there is no lack of intent from their side, it’s just that their bandwidth or the priorities on the engineering side has been dedicated to completely something else?

Tanay Tandon 31:10

I think so. And I like to assume good intent at the end of the day from the players. And I do think every interaction we’ve had with an Epic engineer is one of someone trying to solve a problem. And every interaction we’ve had with an engineer working in the CIO’s office in the health system is someone that is burdened with a thousand tickets and is just trying to solve problems but needs to prioritize. It is a little bit of a question of prioritization. And then I think it’s also, Epic is very responsive to their customers, and their customers are the health system. So when a health system needs something and asks for something, they make it happen. But they’re not necessarily, historically, been accountable or responsible to some startup out there. But with interop standards, that changes the game a little bit, where now there is a maintained set of APIs that people can submit apps into the garden, get approved, and go from there.

Siddhartha Ahluwalia 32:05

So is the government regulation trying to create an app store for each EHR?

Tanay Tandon 32:12

I think we’ve spent time with the administration on this. And what is exciting and promising is, as opposed to forcing the EMR itself to, say, create an app store and regulate how that’s set up, they’re much more focused on ensuring that the data that lives in the EMR itself can be accessed in a smooth way. And then what emerges from that emerges. And some EMRs have taken the approach of having sanctioned app stores with clear libraries in terms of how you integrate and what you pull. Other EMRs have taken the approach of, here’s the data. You have standardized HL7 or FireFeeds. You can pull in right back in. You can pull out right back in. And the end result from that is, I would say, in the last three years, the volume of apps and types of tools you can ship on top of the EMR has increased by tenfold. And I think in three years from now, it’ll have increased by tenfold again, which is really good for the healthcare economy.

Siddhartha Ahluwalia 33:05

Got it. Hemant has said that unless Commure becomes a $100 billion company, it wouldn’t have created any impact.

Tanay Tandon 33:13

Sure.

Siddhartha Ahluwalia 33:13

Why do you think this?

Tanay Tandon 33:14

Well, I think it’s true. If you think about the scale of problem that we’re trying to solve in healthcare, I wouldn’t consider it a success. We really wouldn’t have penetrated much of healthcare at all if we weren’t a $100 billion company. If you look at the breadth of our tools in the suite, they can create real operating profit and improvement in operating margin at these health systems. And one, we have to do that at scale. And two, we have to continue to do that well, such that the partners see that operating cash flow increase and use that to reinvest in their own growth, in more of our tooling. And so, a $100 billion company is a company that is approaching $10 billion in run rate or whatever multiple it is that you want to use. That’s still a fraction of a fraction of total healthcare expenditure and total transaction volume and all these numbers. And so, it still wouldn’t be a very penetrating company in healthcare. And so, it’s a good benchmark just from, if we do our job reasonably well, and we continue to grow in these customers, there’s line of sight to being a $100 billion business.

Siddhartha Ahluwalia 34:18

And you are now trying to expand also outside the US, which are the focus geographies and why?

Tanay Tandon 34:23

Yeah, it’s very interesting because one, we were actually initially brought into these regions through our EMR partner, Meditech, who’s been an amazing partner in the space. And between the Middle East, Australia, Europe, a couple of regions in Asia as well, there’s just been so much interest and pull that initially, it was more of a, we’re going to learn about this and go into these markets and have early meetings. But there were fast acting customers. And ultimately, as a technology company where you see fast acting customers who have real pains, that’s signal to jump in. And the problems were also more standardized than we initially realized, where documentation is a problem, or even parts of the revenue cycle stacking of the pairs are different, they look quite similar. And so, adapting our tools to the needs of the region was an order of magnitude easier than we initially thought. And language models have made localization and workflow optimization for a region much easier. It’s just the volume of code you can write and the volume of, the speed at which you can adapt software is increased as well. So, that’s been an exciting part of the journey. We have multiple partners in the Middle East now, multiple partners in the UK and rest of Europe, and that segment will continue to grow.

Siddhartha Ahluwalia 35:38

What is the hardest resistance that you are getting still now from healthcare providers in the US?

Tanay Tandon 35:44

I think there continues, and I see it as, we have to do a better job of educating in terms of what, when you see the, like level one of adoption is pretty simple, which is like, great, here’s a note taker that integrates to your EMR and can automate a lot of documentation time, improve the quality of your documentation, and as a result, improve the dollars that you get back from insurance companies by reducing denial rates and increasing procedure capture rates. And so, that’s sort of everyone believes. I think the thing that many health systems haven’t realized is just how eugenic some of these models have gotten in terms of solving tasks end to end, getting a claim back, reading the contents, reading the patient’s demographic file and their chart history, looking at previous denials, and then formulating the fix on the fly can happen in a matter of seconds. And I think continuing to educate CFOs and CIOs and CMIOs about those capabilities through good demos, through good integrations is part of our job. And that’s really the resistance in that people want to make sure that when they hand something off to an autopilot and an agent that’s going to take on all these tasks, it has good error bars, good error handling can escalate when it’s confused and not go rogue. And that continues to be part of the narrative and part of our work to make sure the product has that built in.

Siddhartha Ahluwalia 37:06

One of the resistance of healthcare systems in the US of adopting agentic technology has been, ultimately, if something goes wrong, lives are at stake. Who takes ultimate accountability?

Tanay Tandon 37:18

Yep. Yep. I think the legal structures today are pretty clear on this. And I think the right approach is, ultimately, if you’re the entity that has the malpractice insurance, the entity that is rendering the care, the entity that’s making the final call still needs to be a human. And that’s why purely AI-driven care, in my opinion, is not the right approach. You need to have a human in the loop. And that’s what physicians are trained for. They’re trained for accountability. They’re look at a set of symptoms, analyze them, consult with a second opinion, third opinion where needed, but ultimately make a call. And the model can help with that and can help synthesize information. But I think it’s very important that humans are still in that loop and are making that decision. That’s how the healthcare organizations that we work with are set up. And that is also how most patients want care ultimately rendered. So, the risk and part of the risk is also, it is taken on by us as a technology provider. We need to render a good output, just the same way that if your car malfunctions, there’s some maybe amount of blame on the truck driver. But ultimately, the truck and the vehicle automation software or whatever is running needs to deliver. And so, there’s some of that that we take on as one, often on a contract-by-contract basis. But the provider is the provider. And we want to guide them, but we never want to make a call for them.

Siddhartha Ahluwalia 38:41

So, it’s not fully FSD mode right now?

Tanay Tandon 38:44

Not fully FSD, particularly in clinical care. Now, I think it’s a different story in the back office. I think the back office where it’s claims adjudication, and it’s financially driven, and it’s much more focused on completing a series of tasks and clearing admin tasks, that’s an area where FSD is very alive, very real. And for us, 85% of our claims don’t touch a human. They’re entirely autonomous. And everything from the packaging of the claim, the submission, the coding of the claim, and then the back and forth with the payer is done by a model, which is very exciting.

Siddhartha Ahluwalia 39:16

So, for a patient who’s still in the back office, for a patient, when can a healthcare experience in the US become like Waymo?

Tanay Tandon 39:23

Yeah, I think that’s a great question. And for us, that’s the goal. That is, when we’re going to health systems, that’s what we’re pitching, which is, it should feel like a Waymo for the provider and for the patient, which is everything but the most essential bit, which is the care itself, is stripped away and abstracted into beautiful software. And we’re heading in that direction very much. So, I think some of our practices today, especially in outpatient settings, literally, the provider will walk in with a microphone or maybe something running on their Apple Watch, and they will see the patient, they’ll converse with the patient, the documentation will automatically be generated, they won’t even look in their EMR, and the claim will be submitted before they walk out of the room. And to me, that, and the patient will have called and a voice agent will have helped them schedule, reschedule, intake and eligibility information will have been handled, and everything will have been kind of loaded into as context for the model. And that seamless experience is getting very exciting, where you just render care, you just focus on the care, and everything else is handled. And the best software experience, the best product experiences feel like that.

Siddhartha Ahluwalia 40:32

You have mentioned that your sharp learner, you know, that you’re looking for to hire, are inspired by companies like Stripe, Rippling, and Palantir. What among these, or from each of these organizations that you have observed in your hiring practices?

Tanay Tandon 40:48

I think the, when you look at folks that come from Palantir, they tend to be relentlessly resourceful, they’re willing to, and they have a very good instinct for adding value to a customer’s life, in that they will show up, they’ll assess and understand these problems, and then they’ll attack the problem head on and try to add value very, very quickly. Palantir’s time to value for an organization of its scale is so impressive. And that’s something that we try to bring into our, you know, we look for people that have those same instincts. It’s a blend of technical and non-technical being able to really listen to the needs of the customer. You know, when you look at a company like Rippling and how it was built, they were very intentional about building this multi-product platform business from the get-go, which, you know, has worked well for them in their segment, but I actually think works even better in healthcare, because the needs in healthcare and the patient story and the insurance story and documentation are so interconnected that you really can’t just do one thing. It is, you know, seductive to try to just do one thing, but you have to build the rest of the stack. And we’ve learned a lot from folks at Rippling and really how they’ve approached building a platform business from the get-go, making sure that these things are modular, they listen to each other, they speak to each other. I think when you look at the urgency also at both of these companies, they operate with a very high velocity and just quarterly urgency in terms of shipping product, in terms of getting things out of the door. Frank Slootman businesses as well, Snowflake ServiceNow, you know, we’re big believers in the amp it up approach and, you know, it’s like required reading for senior leadership at the company. So, again, he talks about speed, he talks about urgency, he talks about raising the bars, and just always trying to find a way to get the thing done faster and questioning the default timelines. Startups are all about that, questioning default timelines with your customers, your partners, with your internal teams, and you just have to find a way to make things happen faster. If you stick to the defaults, you will die. There are larger businesses that are equipped to service customers at the defaults.

Siddhartha Ahluwalia 42:58

And if you reflect back, what have been some of the reasons why Commure became a success and one of the leading companies in healthcare?

Tanay Tandon 43:07

We’re, you know, I think we’re still on the journey. It’s a long way to go before I would say that we’re one of the enduring businesses in healthcare, but I think we’re well on track. And what it’s come down to is there is a level of customer obsession and urgency. When you go to the office at 7 or 8pm on any day, there’s always a huddle of people that are working on some important customer problem. And, you know, the scale of thousands of customers, there’s always going to be a small percentage of them where maybe there’s a red alert or something, an integration broke because of their existing EMR, or, you know, there’s some confusing aspect, their dollars are down because they stopped performing XYZ procedure and they need us to help debug that for them. And there’s a real care for the outcome of the customer itself, as opposed to, I think historically, software companies have been, we sold you the tool, figure it out, like go, you know, use the tool, figure it out. And you can scale fast with that approach. There’s no doubt, like the beauty of software is it is really cookie cutter, like you get it, and then you’re required to figure it out and use it. I think with language models and the age that we’re in, customers expect and demand more, which is they really want to see outcomes, they want to see real hard dollar success. And we can use LLMs as a weapon to go make that happen. So those things have been super ROI driven for our customers and not allowing ourselves to turn into a, you know, per seat salesman has been huge. And I think we’ll continue to play to our advantage in the coming years as well.

Siddhartha Ahluwalia 44:38

And till now, the enterprises in the US have complained that with agents and AI, they haven’t seen the cost come down. Obviously, there’s increasing amount of work, productivity gains are there. That’s why, you know, hundreds of billions of consumption of AI tokens up there. But the spend hasn’t really come down. What’s your view in healthcare? Has your customer spend come down?

Tanay Tandon 45:03

I think what we’ve seen in healthcare very, very tangibly, and HCA talked about this in their earnings call, I think a couple of our other customers have now talked about this earnings calls as well. They’ve seen the productivity lift very clearly in terms of the number of appointments that a physician can do, and the quality lift in terms of denial rates going down in appointments that are touched by AI. And so because of that, you see that, some of our customers are driving to historically high profitability and historically high revenue targets in a very difficult healthcare environment. And that’s a synthesis of strategies, their own growth, their own operational improvements, but AI is 100% part of that story. And premier AI tools are 100% part of that story. So we see the healthcare story a lot clearer. I think the ROI story is much clearer. The challenge is that if you’re an org and these new AI tools come around, and you basically have everybody spending now an extra 10, 20 grand on tokens a year, you can quickly lose many of the benefits. And I think that’s what’s happening in some of these larger companies. But if you’re a tech company that prints billions a year in revenue, you can kind of afford that experimentation tax. You actually want to overspend for a little bit, understand the limits and capabilities, find where the gold is hidden, and then follow that down. And then there will be a time obviously for austerity and fixing, reducing these costs, and there will be entire companies built in that category. So the other final thing I’ll say about healthcare is there’s also such a supply-demand imbalance, where there are more patients to care for and more appointments to see than most hospitals can take on today, or most ambulatory settings can take on today. And because of that, AI has allowed them to service the needs of even just their existing customers. So we’re seeing outcome scores improve, we’re seeing overall patient NPS improve, and then the dollars and the top-line metrics improve as well in tandem for these groups.

Siddhartha Ahluwalia 47:00

But have you seen patient outcomes improve, let’s say, less wrong number of diagnosis, like patient recoveries happen faster, or surgeries taking less amount of time, or surgeries being more effective?

Tanay Tandon 47:15

We will start seeing that in the next year or two. And the reason is that the adoption of AI was kind of… Step one was get these tools into the physician’s hands, and get the productivity improvements, and remove the need for them to be typing all day long. One, you have some immediate improvements from that itself, which is you have a higher-quality appointment. You have the physician that’s paying more attention, that’s making fewer mistakes. And the number of times that we’ve seen, because the AI is running, if it detects there might be a mistake, it services that as an issue to the physician, and it does a care queue, and the physician can quickly make a change, or an intervention, or get back on track. But I think as these tools become more adopted, the concept of a care queue, or we call it a nudge, will become more commonplace as well, where it can actually read clinical guidance, help and guide the physician in terms of identifying the issue, or even making a suggestion. And then we’ll see the bend in the cost curve, and the bend in the outcome curve in terms of those improvements. It’ll take time though, just because one, these are also large populations, and these are large cohort studies that need to be run. But the anecdotes are there. And I think the anecdotes have made it very clear that the care quality does improve, especially in these high-acuity settings, because you have a more attentive physician, and you have an AI assistant pulling the right clinical information at the right time.

Siddhartha Ahluwalia 48:35

And let’s say for India now, they’re adopting systems, so it’s obviously learning from the US, but it wants to avoid the mistakes that US has made. What would be your guidance on that?

Tanay Tandon 48:48

I think it’s a very exciting opportunity, because there’s an opportunity to leapfrog, to your point, many of the mistakes that were made over the last 20-30 years in the US health system. And I think the biggest, I have some very specific recommendations. One is, you want to keep and allow care to be as simple as possible, and also at the same time, avoid weird supply-demand mishaps that emerge from regulatory burden. And so, in the States, I think an insane policy that we have today is that a physician licensed in Kentucky can’t serve a patient that’s in California. And so, basically, you don’t have liquidity in terms of your supply pool of clinical labor. And so, same thing with a nurse, and there’s all this documentation, and recredentialing, and contracting that needs to happen. And basically, you want a very simple federal-level licensing and credentialing process. And that should hold true for the payers as well. Often, what happens is if the payers have their own providers, their own provider groups, they have these incentives to direct care in certain directions versus other directions. And there should just be more transparent pricing. So, pricing transparency in the US is a great thing. I think pricing transparency should be the default, and should be where things start in India. And good regulation there can make a big impact. Because once you have pricing transparency, you have pricing standardization and normalization. And care across the street is going to be roughly the same amount, and you can actually have competitive dynamics emerge. So, the broad recommendation is allow competitive dynamics to emerge by not overtly regulating licensure, and overtly regulating how credentialing and contracting is done. The other piece is you really don’t… There’s this big disparity today in the States in terms of how hospital and inpatient care is paid versus outpatient care. And because of that, there’s a huge drive now to transition more and more care to outpatient settings because it’s so much cheaper. And the same procedure in one setting gets paid three times or four times more than a procedure in another setting. You want to normalize that. You really want… One, it creates a competitive playing field for the startups and the smaller upstart physician practices, and allows a 20, 30 provider group to emerge and compete with the hospital that’s been around for 30 years. But at the same time, it also ensures that you’re not overpaying for the same procedure, which happens with insurance a lot of times. And then I would say embrace an AI-native system from the get-go, which is tasks like documentation, claim submission, the appeals denials process, it should be agentic. There’s no reason for a human to be reviewing standards and saying no. And LLM can actually do a better job of that and be more transparent about what needs to be fixed. Ultimately, you want the claim to get paid, even as an insurance company. If it is not a fraudulent claim, you want to get it paid so that the dollars continue to flow through the system. And I think LLMs really help in that endeavor. So that’s the high level of how I design the system.

Siddhartha Ahluwalia 51:56

But today, are you only on the provider side or also are you providing solutions to the insurers?

Tanay Tandon 52:03

We exclusively work for providers. So our entire tech stack is directed at providers, is all the engineering work that we do, claims processing, and documentation is for the providers. And I think that lets us keep our incentives very much in line with them. And we want them to grow and want them to flourish. And in some ways, it creates a bit of an antagonist with the insurance companies. But that’s good. I think every good company needs a great enemy, and insurance companies are a wonderful enemy to fight against.

Siddhartha Ahluwalia 52:31

And why would you take open battle against insurance companies?

Tanay Tandon 52:35

Well, one, I think it makes it more fun. And I think it’s a worthy cause to come battle against, which is we want… The fact that insurance companies have such a… They’re basically monopolies. So which is why when you look at US policy around monopolies, the fact that they’re coming after Facebook and Google makes no sense. It’s so easy to switch between Google and some other search engine if you want. It’s almost impossible to change the hospital that you visit because it is quite literally a regional monopoly. And it’s much harder to change the insurance company when there’s only two options in the entire state. So as a broader point, I think many of these insurance companies have settled into being regional monopolies. And we have to break that in some way. And technology is one of the approaches that can aid that. And so they’re worthy opponent because of the fact that they’ve created so many steps that make it hard to practice medicine. And the real value in healthcare is the practice of medicine. If we can eliminate all these other steps, really insurance companies should be profit minimizing exercises. You don’t want a lot of glut and profit to accumulate there. Because ultimately, the premium should just be covering… The aggregate premium should just be covering the total medical claims. And that’s gotten out of whack in terms of how the system is set up today.

Siddhartha Ahluwalia 54:02

Yeah, because for a patient, the cost of surgery, let’s say in the US, is maybe still 20x expensive than India.

Tanay Tandon 54:10

Yeah. And I think a big part of that is the administrative state is the fact that it takes so many steps to build the patient’s insurance, to work on denials, to the back and forth. And then also the regulatory costs. It’s expensive to build a surgery center today. And we have to eliminate some regulation to make that easier and easier so more of them can pop up. And with any market, if you create competitive dynamics, the pricing will go down. And it’ll come to its fair market value.

Siddhartha Ahluwalia 54:36

Because ultimately, whatever happened in the healthcare system is coming out of the patient’s pocket.

Tanay Tandon 54:43

For sure. 100%. And it’s turning into one of the largest line items for employers as well. It is a massive line item for most families. And so the problem needs to be fixed. The dollars are becoming unsustainable.

Siddhartha Ahluwalia 54:56

So why not build another insurance company to solve that?

Tanay Tandon 55:00

I think it’s a fascinating approach. To build a… And there are companies in the GC portfolio and the Sequoia portfolio that are working on the insurance side of the problem. I think there’s so much low-hanging fruit on the provider side of the problem in terms of, let’s just make a practice run more efficiently. And if profit emerges there, then you can reinvest that in better care. It’s closer to the types of problems that excite me personally, which is the actual rendering of care. And the insurance side, I think, from an incentive standpoint and a regulatory standpoint, building a good insurance company often comes down to building a very efficient denials machine or also you have to get very large populations be able to predict… Oscar Health has done an amazing job of this. Basically, predict medical loss with a very high degree of specificity and have enough patients such that your modeling works at scale and actually applies at scale. So in my opinion, the LLMs are uniquely situated to create operating margin improvement for providers first and clinical and care improvement. And that’s where our team enjoys spending its time.

Siddhartha Ahluwalia 56:12

What are some of the other problems that you are not attacking right now on the provider side that could be opportunities for other entrepreneurs and probably then you can partner with them?

Tanay Tandon 56:21

Yeah. I mean, I’m fascinated by… When you look at drug discovery, it’s not quite a provider problem, but it is ultimately a care problem. And it feels like the moment is now ripe. There’s been so many companies that have been built over the years and blown up and run out of funding. And I think now with the types of things that we can do with Transformers, the amount of compute that we’ve scaled up to, the amount of dollars that pharma has generated as well. I mean, you have trillion-dollar pharma companies now because of GLP-1s. And I think that’s a good thing. Pharma companies produce a real product and that real product helps very real people. So it’s great timing to make these large compute investments to develop interesting simulation and drug discovery companies. I also think that when it comes to the provider stack itself, the physical parts of the provider stack, the medical devices, the robotics, the surgery instrumentation, it’s an entire part of the economy that is run basically by four or five companies that operate, again, as monopolies. And there are opportunities in parts of that to go disrupt with faster engineering timelines, cheaper devices, and just reintroduce competitive dynamics into the space. And so we’re very excited about medical robotics and medical suppliers. Some of the largest companies in healthcare, strikers, $150 billion company, Medtronic, Abbott, $100 billion companies. And no one really talks about them, but they’re big parts of the economy.

Siddhartha Ahluwalia 57:46

Yeah. So you have done 15 acqui-hires and how many acquisitions?

Tanay Tandon 57:50

We’ve done, I would say like five meaningful acquisitions and then 15 acqui-hires. Ex-founders are some of the best talent in our business. They’re aggressive. They know what it takes. We really do believe that we’re a great training ground for founders as well with the intensity that we operate at. So acqui-hire has been a great source of talent for us. And I always tell our team and people outside of the company, if they’re running a company and they ever feel like they want to join a larger ship that has a ton of distribution, we should talk because there’s a conversation to have for sure.

Siddhartha Ahluwalia 58:22

What are the incentives for investors and let’s say founders to be acqui-hired?

Tanay Tandon 58:28

I think for founders, the incentive is, let’s say you took a stab at an idea and it kind of worked, kind of didn’t, maybe some aspects worked out and you got stuck in the distribution rut. Maybe you had an interesting product, but it just wasn’t able to take off, which happens a lot. It is useful, whether it’s Commure or another business to see what excellence looks like at scale. What does great distribution look like? What does great implementation look like? What are some of the problems in healthcare? You can’t see that unless you really work at one of these businesses. So come work here for four or five years and then go start the next thing. And I think for investors, we’re often a good soft landing because Commure will be a big business. We’re on the path to be $100 billion plus business. Their equity will appreciate. And so when we structure these deals, there’s a way to make it as a win for the investors where their prep stock converts in a favorable way. And the existing founders get a meaningful compensation where they’re incentivized to continue to grind for a couple of years and build product with us. And then when they start their next thing, their investors can take another swing or we can take another swing. And that makes it exciting.

Siddhartha Ahluwalia 59:41

Have you incubated other companies inside Commure?

Tanay Tandon 59:43

We’ve never incubated other companies, but we’ve obviously over the years had folks that have gone on to start their own companies. We’ve had folks start companies and then come back and boomerang back to us where we back wired those businesses. And so, you know, ultimately, I think we’re good founder training ground because every engineer spends so much time with the customer. Every salesperson spends so much time with the engineers and we’re very aggressive in office culture. So you get to feel and see that in person all the time.

Siddhartha Ahluwalia 1:00:12

For a fast-growing startup like Commure, when a founder is considering an opportunity versus let’s say some of the larger labs, how do you win their talent?

Tanay Tandon 1:00:24

Talent wars are more aggressive now than ever before. I do think there’s a certain type of talent that we attract that the labs don’t. And it might not even be good fits for the labs, honestly. Which is, we’re addicted to gritty people. People that want to see problems through, finish hard problems. And the labs, candidly, right now are attracting a lot of people with shiny object syndrome. Which is one, the compensation packages are wonderful. And get one of these crazy compensation packages. You should… It’s rational to take it. But a lot of people are going there for the wrong reasons. I mean, Dario himself said that it’s turned a little bit into a mercenary culture. Whereas here, you get direct customer impact every day. You were in front of the customers. You’re solving problems for the customers. And I think a lot of value will accrue at our layer. Which is like that final application layer where the last inch or the last mile, whatever you want to call it, is how these models get developed. And we’re in pricing bake-offs with all the models every day. We pick the best model for the job. And we’re extracting every last bit of margin from them as we can. And the minute we detect that there’s a cheaper model that can accomplish that task more efficiently for our customer, we’ll pass those savings on to our customer and take advantage of that. And so the foundation model companies are serving a very important role. And they are some of the most fascinating businesses ever built. But I think there are more interesting problems being solved when you’re a domain expert in the arena working with the customer and actually adapting the software to the needs of that customer.

Siddhartha Ahluwalia 1:02:02

Thank you so much, Tanay. Love the conversation. Love learning from you.

Tanay Tandon 1:02:06

Thank you for having me. It was a lot of fun.

Siddhartha Ahluwalia 1:02:08

And specifically, I come from a background of building a healthcare software company in India. So I can imagine how hard it is to build a company of this scale.

Tanay Tandon 1:02:16

And you did it a while ago. And so kudos, because it sounds like it’s scaled to millions of patients. And it is very hard. So a lot of respect for anyone that’s ever built a healthcare.

Siddhartha Ahluwalia 1:02:27

Thank you so much. I appreciate it.

Tanay Tandon 1:02:28

Amazing. Thank you.