Latest / Powering the Future Within / Episode 22: Healthcare's AI-Native Infrastructure Layer: How Autonomize Tackles Problems at the Root
Transcript
- speaker-0: Welcome to Powering the Future Within, the podcast where we explore the breakthroughs shaping the future of healthcare. I'm Omari Bachnight, CEO of ResidentLake Medical. Each episode we speak with a leader who is driving the next wave of medical innovation, advancing technologies that are smarter, less invasive, and more connected than ever before. This is where the future of health takes shape. Let's dive in. Today I'm excited to be joined by Ganesh Padmanaban, founder and CEO of Autonomize AI, a company using agentic AI to health care help healthcare and life sciences organizations turn complex fragmented data into actionable intelligence. Ganesh is a longtime entrepreneur and technology leader who has spent his career at the intersection of AI, data, and business transformation. He brings a particularly thoughtful perspective on how AI is evolving, what it means for organizations and leaders, and where this technology is ultimately taking us. Ganesh, welcome to Powering the Future Within. It's great to have you. speaker-1: Thank you for having me, Amari. It's great to be here. speaker-0: absolutely, absolutely my pleasure. So Ganesh, you grew up in Kerala and I understand you were quite the cricket player, is that right? speaker-1: Well, I'd like to think I was. No, I did actually. I spent probably most of my childhood playing cricket than any other sport. I did play a lot of other sports too, but it was always the I mean, y if you if you I grew up in India, so India, one point five billion people, I'm pretty sure seventy percent of them want to play cricket or want to play cricket at a level that they all because that's what you that's what I grew up with and I'm pretty sure it's still probably the same. But yeah, I did I did play a lot of cricket growing up. speaker-0: Yeah. Did you have dreams of of taking your your your l to taking it to the next level in terms of playing at a professional level at some point? speaker-1: Yeah, no, I absolutely did. And I think, you know, i i back when I was playing, this was in late nineties and early two thousands, there there was no pro path for cricket at that time other than just play really well and you get selected for the national team. And then there's like, you know, you go for the p district, the state, the zones and all of that kind of stuff. So I went through the whole ladder, played the nationals and stuff, playing cricket. So I was just set in for like, okay, I'm gonna go do this and stuff. And then Do you realize the economic reality of being a cricket player in India at the time was not very, you know, it was very bleak. This was pre what we now have as IPL or which is a Premier League, which is they got money making machine and stuff like that, and people players get played paid really well. Almost everybody who played at the national level in cricket, when I was there, when I was young, when I saw these players out there, they had to have a real job to support themselves in addition to playing cricket. So it was almost like a you know, I had to make a choice. I it was less of a choice than like, okay, you can continue to play cricket or you can focus on getting a degree and getting a job and stuff too. So and I like I I try to play as much as I can after, but then I found software and computers and I got into, you know, my engineering school and you know, never re I did play a little bit after, but then cricket wasn't as big of a part of my life after after that. speaker-0: Right. Right. And so then how did you find your way to the to the US? speaker-1: So I first came here through so I you know, come did my undergrad in mechanical engineering. wanted to go at the time I only wanted to go design the fastest car possible using you know, that was all I wanted to do. And then I realized that mechanical engineering was beyond just car design and you know, learned a lot. but then f I found computers during college. So I had aspirations to come in to the US for my masters that didn't work out. It was when I graduated right around 911, so that didn't really help. and then you know, but later on I I took a job with a company called Adaptech at that time. It was a dot com darling. And they first, you know, moved me to the US. I was in Orlando, Florida. So I spent, you know, time there working with, you know, I was a software engineer doing work with intersection of hardware and software and very data driven work. so that's how I originally came to the US and then, you know, during a series of moving back and forth and I went back to India after a failed attempt to actually go get my PhD, different story. But then I joined Dell and I ended up spending eleven years at Dell, the first year and a half of which I was in India in Bangalore. And then in tw early, I think it was 2008 that Dell was putting together a team for a software group and a growth team in in Austin and Dell is headquartered here in Austin. So that's you know so they asked me to move here And then I never left. So being here in Austin for almost eighteen years. speaker-0: Wow, eighteen years. And in the that eleven years at at Dell, you find yourself at this business and it went from nearly zero to one point three billion. I believe it was like the youngest business within Dell that was at that size and scale. You end up in the position of the the managing director. What did you learn about scaling? I mean, that just sounds like a phenomenal experience pretty early on in your career. speaker-1: No, it was great. I think, you know, Dell I like to think that Dell is where I grew up. You know, it's like you you I started as a you know engineer, a lead engineer, product manager, you know, product marketing to becoming a sales to being a GM for this business. And it was interesting. you know, lots of you know, and and Dell is one of those, you know, amazing organizations where they figured out operational discipline. How do you build organizations that can operate at a high degree of conviction, velocity, and can make change happen. Like look at the way they captured the the cloud market to now the AI market. If you own Dell stock, I mean, you know, you'll be re I mean, it's probably grown up like two, three hundred percent over the last year, right? So it's been insane. and it's it's it's a very founder led organization. So Michael is an amazing, you know, founder to to watch and learn. And I had a lot of that opportunities to So at Dell, like you a lot of the it builds a lot of the operational muscle that I now carry into my entrepreneurship life, which is like, you know, how do you build predictable business models? How do you really look at data objectively and not make emotional decisions? And how do you go and you know, there's like there's a the the there is the there's a there's a lot of work that Dell does on building conviction, right, in terms of like, you know, from data, from conversations, the customer centricity. to really le you know lean in and listen in keeping you know what you think is right versus what the customer's actually telling you because most people put that filter, they're just reconfirming what I know. But if you really listen, you'll know that they're telling you probably something else and something different. So all of that was a skill set on the personal basis. And then like you know building teams, right? Like, you know, when you're a you know four or four and a half years ago when we started, we were two, you know, me and Chris, the two founders, And, you know, it was just us to, you know, twenty people or ten people that we initially hired in India to now having a hundred and thirty, hundred and forty people, very different scale. And you need to bring and exercise different muscles as a leader to go make this happen. And I so I built a lot of muscle memory to bring that back, you know, during my days at Dell. And, you know, it's also I mean, but you know, it's the fundamentals. I mean the the the funny thing, the interesting thing is like Business building is all about the fundamentals. Like you bring great people together for a great mission solving the hardest problems that intrinsically motivates them and give them the right structure and the right operating rhythm. And the rest is history. Right. So it's just the, you know, super compressed in what I learned from Dell. speaker-0: So as you're building that muscle, did you find yourself leaning on mentors or how how did you start to learn some of these new skills and capabilities? speaker-1: You know, you know, it's it's a you know, mentors were a huge part of my entire career at Dell. Like at every stage I had you know, you always had somebody who, you know, who took a chance on me and who gave me, you know, direct feedback. And I s you know, some of the bosses that I worked with and I wasn't very happy working for them at that time, looking back and it's like, Man, that was the best thing that happened because they grew me, they gave me the muscle around it. But, you know, like as you as you Think about as a founder, and you know this too, Amari, right? In a as a in a business, you you have to think through like, you know, look, how do I scale as an individual, as a CEO, right? How do I have the skills? Have a real good assessment. What are the constraints for your scale? How does your leadership team scale? And what does that mean? Does it mean do you need get different talent? Do you need to actually coach them and have them, you know, get the right assistance and mentors and stuff to do that? And then you need to actually have the organization, how can the organizational scale? The organization scale, when you have, when you move from a you know, I like one of my coaches stuff says this, you move from a react and repair to a predict and, you know, plan kind of a mode, which is like, you know, as you start up as an early stage company, you're constantly just reacting to what's happening in the market. And in markets like this, when AI, it's a very, very important skill. But then you have to start thinking about like how do you actually then go build a business, then just react to a new sort of product or an update or a technology to say, look, and the fundamentals of the business are still the same. You know, you be in a business to make money, to return value to shareholders, and to, you know, like to to have fun doing it. And it's a journey that everybody needs to be on. So it's a it's a it's it's a mix of different things. So mentors, I lean in on a lot of my peer groups, like I a lot of the founders I, you know, I call, hey, I'm struggling with this, you know, tell me more about it. My co founder, Chris, and you know, we we like there are there's only there are certain things only founders feel. Right. So it's a lot of those different things that I don't know, we get we just go through it. There's no I don't think there's a silver bullet for it. speaker-0: No, no silver bullet, but I I love the idea of trying to transition your leadership approach from react and repair to kind of predict and enable. And that also seems like an approach that you've taken now with autonomize. So we fast forward, we get to 2022, we're on the heels of COVID, and you decide to launch a healthcare AI company with Autonomize. So what what drove that decision making process? speaker-1: boy. I think, you know, so f first of all, I mean in the middle of twenty twenty, I kinda I just left my last startup and, you know, was really in the throes of like I it was it was it was really tough times, right? Like COVID during that time, a lot of unknowns, a lot of lack of visibility in how the world is gonna evolve. We all I mean, a lot of people thought you this is it and you know, the world is gonna look like this forever and all of that stuff. And You know, I got a lot of very interesting perspectives during that time. One, I was part of Governor Abbott's task force for COVID nineteen, which gave me a very front seat view of the state of the healthcare infrastructure, the national infrastructure. One of the biggest problems they were trying to solve is how do you get care to the right people at the right time? You know, who gets the, you know, ventilators and who gets oxygen tanks and how do you manage hospital beds? And a lot of lot of good things came out of that. But, you know. Boy, I mean the infrastructure for healthcare is really, really old, and we all know that, right? That's that was one perspective. I spent some time also thinking through, like, you know, I was my my wife was, you know, gainfully employed and I had the luxury of actually just taking care of the kids at that time and you know, looking around and seeing, and I talked to a lot of people, the folks who really show showed I would say like a lot of stellar Courage at the time was the frontline workers in healthcare. I mean, when everybody is trying to stay away and masking and going out of this thing, they would go and you know suit up, you know, put hazmat suits, whatever it takes to go and sell patients over and over again. And the state of infrastructure that they had to deal with was just it was almost a sin to let them all work like that. So that was one of the, you know, key motivations. Then I spend a time about another six, seven months or so consulting with several of the leaders who were leaders in healthcare companies, but they were looking at AI. So I had the luxury of being having the background of in AI and machine learning and NLP at that time. And I just like, look, I mean, I want to learn from you. And so what problems you're doing. And they're like, why don't you help me with thinking through AI strategy and stuff too? So took the opportunity to go spend a lot of time with them and I learned a few things. One, healthcare didn't have a dearth of problems. There's so many problems to solve in healthe. But the way we solve problems in healthcare was actually very unique to healthcare. We would take the symptom of a larger problem, build a vertically integrated solution, get into you know, put the ship that, solve that, and everybody will say sing kumbaya and say it's working. Let's go do it. And if you look at the last 10 years, last decade of healthcare, a decade and a half, if you ask any investor, has health tech done well, health IT done well, health services done well, you'll see, it's great. It's been returns have been great. When you go and talk to an average patient, has healthcare gotten better or worse? We know what we're going to hear, right? So there was this big disconnect between all the effort and the dollars being spent in fixing healthcare. And then it doesn't make a difference to the foundational fundamental problems in healthcare. And why is that? So one of the reasons was that look, you were solving peripheral problems, then solving the gut of the problem. The gut of the problems were like access to healthcare was, you know, r i it was, it was a was a roll of a dice, getting the right care for the right person at the right time. If it remote, it makes it all difference. It's sixty, seventy percent of the time of those healthcare workers that they're highly educated, licensed, trained, all of that stuff, spend their time doing administrative work. It was right in front of your faces, right? For one point five trillion of the four or five trillion spent goes on wasteful administrative processing they have to do. I mean you take any process for that. You go to a doctor, they spend the first You know, twenty minutes before they see you or fifteen minutes before they see you just reviewing charts. They talk to you and that time they're not making eye contact because they're looking at the screen because they have to take notes. And then they all go home end of the day to pajama time after the kids are in bed to write up the notes and capture and hopefully I mean they they do they do a lot, right? I mean so but sixty seventy percent of their time is administrative processing and stuff like that. So and then The constraints that this is where like, you know, you mentioned the unstructured data aspect, right? Which is like, it's been hard to solve this problem because most things in healthcare, and it's not so much the unstructured data, it is knowledge in healthcare is very fragmented in notes, in not in a structured format, in people's heads, in the fact that Sally has done something for 30 years, only she knows how to do it. All of that stuff added to this challenge that. Solving for reducing administrative burden was almost an impossibility. Right. Right. And so what we resorted to as an industry was like you find a problem, you'll create a cottage industry to solve that problem. You know, like you can take any example, claim editing for that matter. Like if two people don't agree on a claim, there's another industry trying to actually say that they're gonna come and fix that. And that becomes an industry in itself. Nothing wrong with it, but then it just you know, you're you're not doing, you know First principles fixing of problems in healthcare. Right. Skills like that. That's what we want to do. Right. speaker-0: Kind of like whack them though. Yeah. Yeah. speaker-1: It was just like, and it just, it was this, it was, this has been going on for a long time. And then I'm not saying we've figured it out, but what we wanted to do was like instead of trying to solve more problems silo, and it's hard to explain to investors early on when I would tell them saying, Look, okay, so are you a care management company? Are you a utilization prior company? Are you a claims company? I'm like, we're all of the above. And they're like, that's just you're just so defocused. And I'm like, no, the idea is like you don't need another care management solution in the market. In fact, nobody's asking for it. You go to talk to any CIO, they're not waking up saying, I'm gonna change my systems of record that is for care management for the last thirty years that all my nurses are trained on to with a new system. They don't want that. What they want is, man, my nurses, my MDs are spending fifty, sixty percent of their time doing mundane BS. Right. Can you take that off? I don't care how you do it, but you know, but then for you to actually drive that effectiveness, you can't just go and give them a new mouth strap and say, Go try this, right? So you what they what we wanted to focus on was like how do you build an infrastructure layer for healthcare that is AI native that allows you to not just solve one problem but lace the infrastructure to solve any problem. Right. Now obviously as a you know, as a company, you have to go and win that, you know, trust with one or two areas, one or two area you know, places saying, This is your burning platform, let me go solve that for you. And and we did that, and that's how we got the initial scale. And then once they see the value of saying, I'm gonna go into a customer or if it's a help provider, it will just go and say, look, let me solve your pride dot submission problem. Well, when you do that, they actually see we see that like, do you know that your care gaps you're not crossing because it's actually on the charts right there? that's a different team. We got to hire some nurses to review it. I'm like, don't do that. That's an adjacency problem that we can actually solve. You can deploy agents for, I have to do appointments. I'm like, Well, why don't we just, you know, create a an army of agents to do text messages and appointment setting. So you don't have to take out nursing time to go do that or front test time to do that, right? So there's all of these things, all these problems are interconnected, but we never look at it as an integrated problem. Because end of the day, it's not about the function, it's not about the workflow. It's about the patient, it's about the member. And you know, it just somehow we kind of lost a little, we got a little lost in healthcare doing that. speaker-0: And and so it's interesting. So you're there at the front line or seeing the frontline activity during twenty twenty and COVID, starting to really get a strong appreciation for how broken some of these systems are. And coming from your background, obviously from with artificial intelligence and big data, kind of really understanding that there's gotta be a way to to make this more more functional. What's interesting to me is when you think about physicians and nurses, clinicians, we refer to them as practitioners, but The reality is is the vast proportion of their time is not being re used towards practicing is sort of the administrative burden behind the scenes. And it sounds like for you, one of the critical things you needed to do was you understood that an approach, this approach could be applied to so many different areas behind the scenes, but you really needed to find something that you could first kind of put a stake down in, you know, as you mentioned, a burning platform. So what was the first sort of burning platform that you really kind of turned autonomize loose on that really opened a lot of people's eyes. speaker-1: Yeah, I think, you know, it it evolved from the initial days to later. So initially it was like we went after, we said, look, healthcare is broad, but even the way we think about healthcare, there's providers, there's payers, there's life sciences, there's med devices, biopharma. We think it's all, you know, everybody's in the service of the patient, right? I mean, like yes, they'll have economic incentives to do things a certain way, but it's all in the service of the patient. So we we we and we have we're ambitious, we go across as what. So we started with life sciences. Because one of the largest problems we heard when we were initial conversations with customer discovery was that look, a drug launch takes about ten to twelve years and two billion dollars to do. Right. Eighty percent of that or seventy or eighty percent of that goes towards the postclinical, like all doing your clinical research, getting patients on the trial, managing them, keeping through this process and getting a drug approved by FDA. So we tried to we our first burning platform we thought was the Patient matching for clinical trials. So that's what we did. So we actually ran well for one of our first early clients, a CRO, we ran a a large oncology clinical trial for them. We when they say we ran it, we basically helped their PIs and nurses, you know, go through 10,000 potential you know, patients to narrow down the 120 that makes the difference. Right. And then that was basically Matching all the unstructured data content about their charts, five, six thousand pages of medical information, lab reports, and so forth, to the five to ten inclusion criteria and 10 to 15 exclusion criteria. And that process was very manual. You're comparing notes and it's very both of them are unstructured. So we built a system to go do that. That was the we thought was going to be the the burning platform and everybody's going to buy it and stuff too. And we did get some success. We had some early customers who actually, you know, bought it. They're seeing the value. But then there was no path beyond that because the life sciences industry was very fragmented, right? The pe the folks, the company, the organization that focuses on making the trial happen didn't care about the safety, didn't care about the patient access, didn't care about the things around it. So we wanted to be in a space more provider and payer in the core healthcare, in the delivery system side, and the delivery side, where you have concentration of multiple problems to pay out our thesis that you can lay an infrastructure to solve it. Because otherwise you just need another product. or another siloed product to solve the patient matching problem. So we moved from, you know, we still have life sciences, we do some work with them to but we moved into the business of care, as we call it. So we we found health plans, pharmacy benefit managers, value based care organizations and life and and and providers and we specific the burning platform everybody told us was prioth. And, you know was a regulated span, regulated process. It's got a lot of bad drop in the industry for the right reasons. You know, there is different incentives on different parties and a particular prior art request process and stuff. But end of the day, I think, you know, that was a burning platform. It still is a burning platform in a lot of ways. So that was one of the first initial focus areas that we actually went in on and you know, several parts of it, right? It doesn't priorat doesn't add we want a big thing, but that whole process includes the intake and case setup process, the medical necessity review process, the letter and communication process. So it's multiple sub workflows within that whole utilization management workflow. So that's where we focused on initially. And then you know speaker-0: Point in time was there a point in time when you first looked at the prior authorization process and just said to yourself, Wow, this is really, really broken. speaker-1: yeah. I mean I I think there's st that point in time it's you know every day. You know, it's like there's so many th there's so many things, you know, wrong with it and how it's done. But just think for a moment, right? I mean, th there's some interesting things. Like an average health insurance company, health plan, you know, for your national payer, you receive anywhere from a hundred and twenty to a hundred and fifty thousand requests a day. Right? If you're doing Medicaid and so it's it's different depends on the the business and stuff. You would probably you have an army, sometimes hundreds and sometimes thousands of people who will process these requests. And everything from it goes through a life of you get a request that could be in a hand scribbled format, that could be written in a form, that could be an electronic submission, multiple ways. That could be a phone call and you have notes scribbled down. And somebody is actually turning that into as much structured data as possible. This is a data entry and update and case setup process. And then you attach a bunch of medical records, usually formal PDFs that's sent to it. It goes to a nurse who then have to spend the time first ensuring that the eligibility and the benef verification of benefits happens. You know, you understand copay, all of the payment things and stuff like that. And then it passes on to a nurse. And the nurse actually then has to review that against policies, medical necessity criteria, you know, guidelines, all of that stuff to make a recommendation saying yes, it's an approval or a denial. Or you know, need more information, whatever they d choose to do. And then it goes to a group that has to communicate that in form of letters, in form of, you know, doc updates to the doctor. The whole process takes like, you know, I would say anywhere from thirty to forty minutes a request. So imagine the amount of human capital needed to actually go process this, right? Now, on top of that, you know, the patient wants to get care faster, right? The provider is just waiting to serve their patient. I'm requesting approval. I need the approval. If it is an approval that happens, you just spend all this time, a patient waits for two or three weeks to go through this manual process because they got to the case, they didn't get to the case, they looked at it, reviewed it, and they finally approved it, or they saw found that there is some information missing here. Can you furnish this? You know, you're asking for a continuous glucose monitor and you know, you you know, your A1C is under control. I don't see a medical necessity here unless you give me evidence of, hey, you've tried these different things and this is why it's actually needed and so forth. So I'm just I your your payer is actually contacting the provider for that information. So there's a lot of back and forth and stuff. Imagine you can just completely make that process a lot more efficient. Where you know, if it's a submission, all the documents are there, the medic patient meets the medical necessity. An approval shouldn't even go through a process of human touch. You can have agents that'll actually just understand the information, match it against policies, and auto-approve that particular If it's you know, you don't meet all criteria, yes, you need human intervention, you need probably an MD to look at it. If there's like missing in most most cases are missing info, right? Because you also on the doctor side, they're having to pull information from medical history, different nodes. EMRs are not the most user-friendly, you know, pieces of software in the industry, as you all know. So they have to look for their EMR, whether it's Epic or AllScripts or whatever they use, and then get all that information. So it's just very mundane and manual for either sides. So what we do is for the provider, we make the submission a lot more easier. One click, you go and get all the information depending on the request, and you avoid the burden for that. For the payer, when the requests come in, we go through all the processing that they can do that is non-clinical and gathering information for clinical folks to actually make a quick determination. Right. And then you automate all the stuff that is in there, which is an interesting way to think about it, right? So And then on top of that, right, now so we we talk about just the transaction of prior authorization. But take a step back. we work with a Metrisk provider and we did this experiment with them on a small scale. They were look, we were looking at PET scan requests for confirmed cancer diagnosis. Well, 99% of the time it's auto, it's approved, right? But then you just if you treat it as a transaction, you get a request, it's a PET scan for an MRI. I'm just gonna approve it and move on. But what you just have is that information, that first contact from a patient that says, Hey, I have cancer, I've been diagnosed, and I need a scan to, you know, understand the the locality, all of that stuff. That should be a trigger to go get the care manager and the care navigator involved to say, you know, what does the, you know, because they you can pretty much predict the pathway for that patient for the next six to seven months. It will involve chemotherapy, it'll involve radiation therapy, it'll involve Nutritional counseling, it'll have, you know, all of that kind of stuff. What if do you give a call to the patient this afternoon as a health plan to say, hey, I know you're going through a tough time, but we've actually, you know, mapped out how this treatment plan needs to look like. You have a choice in selecting where you're going to get that treatment. But if you choose our list of providers here who we've actually verified and made sure based on my provider contracts, you can have a very seamless experience and a predictable, you know, out of pocket payment because most people worry about. Hey, am I gonna live? And the next thing they worry about, am I gonna get bankrupt? Right. So there's a lot of things around this. And I'm taking an example of of one example where if you don't treat prior art as just a prior, then it's a bad thing, but it's the request coming in for a start of a care journey that you can now go and predict what the future journey is gonna be and then turn that experience around for the member and the patient. There's a huge opportunity that's left to, you know, to see even like, you we You know, you you come from the med device background. And prioroth is one of the biggest challenges of patient access, right? And I think, you know, so there is this is there is a common ground here that you can actually go and solve for within the confines or whatever the regulatory reasons that we have this and stuff like that. Medicare just announced that they, you know, like what's it one point five billion dollar savings or something that they did with all the fraudulent thing that is happening on Medicare. So it's like it there is something here that you need some kind of you know, policing, watchover processes and stuff to go do. but I think, you know, it's it's not done efficiently at all. speaker-0: Right. So you start with the the one burning platform with the life sciences organization, you add to it with the pri authorization, which is in my opinion is a huge, huge challenge and issue. And so I'm I'm really grateful you're tackling it. Was there some point where you started to look at the data in that specific space and say, Wow, this is really working? speaker-1: Yeah, no, I think look, we we were able to, you know, like for one of our largest national health plans, you know, they were able to compress their clinical time spent on reviews by about forty percent, translating to about thirty-three thousand hours a month of clinical FTE hours. It's it's insane. I mean, like I would even think, like, imagine redeploying that all on the bedside to help patients. I mean, you call any hospital CEOs, they'll tell you. their largest backlog of they don't have enough nurses, they don't have enough physicians, you know, because they're all trying to hire. And then on the other hand, you know, you're putting them to administrative work, right? So there's a there's a there's a disconnect here. That was one example. We just you know launched a case study of the work we did with Cigna Help Plan. And y it was it was a beautiful story. It was like it was the what is it the there was a benefit request so if you have you know, an expensive drug, then the the health plan has to make a determination based on how they have configured the policies, what is the right course of care and all of that stuff, whether it's going to be paid through the medical benefit or through the pharmacy benefit. Something sounds pretty mundane. Well, patients are waiting for care because it takes days or sometimes weeks to go make this determination. Because you have to go s check multiple policies, multiple systems, multiple, you know, benefit packages and how the patient is the what the member and the patient is actually part of, which group all of that stuff. And, you know, we just used AI to go take that entire process a lot more seamless. Very, I mean, no zero human touch to make a determination for the human to act on. 50 times faster to actually go do that with the use of AI. Right. So we're having, we're seeing some really strong, you know, results right now. And you know, w one of the things about AI is like, look, I there's a lot of voice agents, right? you know, in the market. It's one of the most hardest things. I mean, tell me one really high large scale success story out there, right? I mean you can hear a lot of and because it's sometimes it feels like, you know, you gotta choose the right problems. I'll give you another example on care management that we did. You know, we we this is h for a health plan, you know, a care manager or case manager spends all their days, you know, they take about fifteen to twenty calls a day to patients and they they reminding them for their, you know well checks or screenings and coaching them and mentoring them on their course of care, coordinating care for them and all that stuff. But they spend thirty minutes before the call and fifteen, twenty minutes after the call doing all the paperwork. So paperwork is like collecting information from claims, from medical charts, from doctors notes, all of that stuff, lab reports, to really form a hypothesis on what they have going to direct the care journey to towards and, you know, and may getting sign out from MDs to do that. And then they after that, they have to go and make notes on it and make so it's useful for the next care manager, next case manager, all of that stuff. By just cutting that pre-call workflow to completely agentified, where you have agents doing the work and giving them a summary they can review and edit and check before they make a call. And by doing the other side, one of the large health plans, they saved almost $50 million a year on doing that. Because think about it, it's like if it's 10 minutes of time per case manager and they have hundreds of case managers. It just adds up, you know, well, you know, per day, right? It's just a ridiculous amount of savings. So there is a lot of places in healthcare. And it's not, I mean, not not to say these are not easy problems. These are hard problems. But to me, these are problems that are worth solving because, you know, there's a lot of other problems too. I mean, like you can improve the communication and all of that stuff too. But you really have to ask yourself, like for us at least, it's been we've been very focused on How do you go solve the problem that we are true north is reduce the admin burden for pay for for everybody involved in the p in the in the ecosystem so that you know more of that can go towards patient care, more of that can be given to practice rather than just you know paperwork. speaker-0: Yeah, and you've managed to to develop a lot of momentum, which is great. You're growing quickly, you're operating in some of the largest health systems in the country. I'm curious if you could share a little bit because selling into these organizations is very complex, takes a lot of time, and and as we've kind of hinted at, healthcare generally isn't on the front edge of technology and artificial intelligence. So what does it take to sell into these types of organizations and really deploy inside of these organizations? Sometimes the decision makers who make the decision to bring you in are not the users. The users might be different. Those who implement are different than the actual end users. So it's very complex. What are some of the the takeaways you've gotten from selling into some of these large organizations across the country? speaker-1: boy, it's i it's long, it's tough, and it's very multi persona, right? I mean I would say there is the selling in general is hard. Like especially in this market where it's noisy, there's a lot of you know, noise in the system, people don't know and buyers don't know how to buy AI, for example. I think the the the the fundamentals of this is like one, you know, you gotta get ready for some of these large healthcare d conversations are it's gonna be a multi party enterprise sale. And it's not just a technology sale. You gotta get multiple actors a part of that particular process. You gotta get the clinical team, the CMO and their team aligned. You gotta get the CO, CFO aligned in terms of they're looking for ROI for investments. You know, there is the the the the the department lead, like a VP of clinical services or or claims or payment integrity, who is your actual customer, they're looking to get outcomes. So there's like multiple folks that are actually involved in this. And so it's it's there's no again ma no magic bullet, but you have to treat it as a multiparty long sale. That's one. Second, I always say this to to to the teams, right? Like, you know, you have to solve a problem that is a hair on fire problem. If you don't solve a problem that is Like y y it's it's a very simple ask, right? Like if we don't solve this problem, what's it mean to you? Right? And the organizations usually leaders will tell you, well, I mean, it's we have it on our book. We haven't gotten to it for two years. We wanna do it. Well, like, why didn't you get to it for two years? What wasn't important? Is it why is it important now? Well, I don't think it's that important, but we haven't gotten to it yet. So I'm like, you know what, it's not gonna happen because starting from the fact that the the opportunity is gonna take about twelve to eighteen months to close. You start off with a place where it's not an important problem. And you're hoping that over time it'll become more important is actually just, you know, you're you're you're you're not investing your time well, right? I like to say to to to my enterprise selling teams, like there are two winners in a deal. This I stole it from one of my old sales leaders. And there are two winners in any deal, right? The one who gets the deal and the one who gets out first. Right. Because I think the one that gets the deal gets the deal. The other one didn't waste time actually figuring out they didn't get the deal. Right. So it's very quick. And and it's and and buyers and c organizations will really appreciate it when you're very upfront with that. When you go in and say, look, is this a is this the most important problem that we can solve? We've solved all these things. Now we have the luxury of showing like here's our showcase of all the problems we've solved. It's a wide variety of stuff across organizations that look like you, smaller than you, bigger than you. Let's pick a problem. So we have the ability to solve this. Now, let's pick a problem that really will move the needles. And sometimes it's, you know, it's it's it you know, you would ask, like, hey, why can't you just pick an easy problem, solve it to get in the door? You can do that too. A lot of people do that. And you know, what we've learned over time, especially with AI, is AI is very unlike SaaS software, right? Where in you build something, you deliver it. And then you're you're timestamping the value you're delivering at the at the go live date. And it kind of stays in there. Like things may change over time and stuff, but it's in there. AI is different. You deploy an AI agent or an AI system in production, it evolves, it changes, it degrades. It is a living entity that you need to actually nurture and you need to really bring together, like you need to bring the operational design, all of that stuff. So it is a long journey and a long partnership. And if you treat it speaker-0: So I'm t I'm curious to unlock that have you implemented a way for there to be kind of a discovery process? 'Cause essentially what you're you're articulating is the fact is that these the agents and the deployment can go in so many different directions. I'm curious whether you've implemented an easy way to kind of go through the discovery. So there's multiple aha moments as the user goes through the speaker-1: No, we do that. So we we just recruited. I mean, like I think about early this year we hi hired a head of transformation and for the company and it comes from a very consulting background. It's a very consultative sale. So we have we do discovery workshops with our customer, and sometimes it's very, you know, as customers say, Look, this is my hair on fire problem. I let's not, you know, waste time. Let's just go solve it first. We're like great. But in the process, we actually go give them. For example, we in our software capabilities We have a platform. We deploy the platform everywhere, right? And then we go in and say, we're going to go turn on the first hair on fire problem, but you have all these other things that's available for you and for us to actually work with. So and then there's a lot of consultative work workshops after that. We help them identify, prioritize. Sometimes we tell them like that's you think it's a very important problem, but your juice is not going to be worth the squeeze that you're going to put on it, right? You're going to put all these efforts and nobody's going to touch it. You know, it's so many stories like that. So and we you know, and then sometimes we still do it because the customer says, No, no, no, we wanna do it. And it's for other reasons. It's not just for ROI. It's like saying, I want to give users something. Like classic example is chatbots. Everybody wants a chatbot, nobody uses a chatbot in the enterprise. Right? Everybody wants it, nobody uses it. It just and then you and then you you get it. I mean, there are times it works. You know, f we we did this a long time ago. payment integrity. We said, Look, I mean, people are looking for these answers. So We shouldn't implement a chat bot because they tried it on Glean or something because of data restrictions, they can't use it. So we're like, okay, we'll do that for you. But we told them like nobody's gonna use it. We're like, no, no, everybody says they want it. And then they use it. And then we learn from user studies that after that is like they like it, it's beautiful, but it changes their workflow. They don't want to do anything different. They want the easy button. This just made it harder because I have to go type something that would remember copy paste from different things and put it in there and stuff. So the same capability instead of just delivered as a chatbot, they were trying to do some work by filling up some kind of a claim system after that and stuff after reviewing this information. We just auto-populated that and helped them go click go. In the same idea, instead of them having to ask, we prefetch information based on what they're dealing with, we do the thing the agents do the thinking for them and then you populate that information. So the same capability delivered differently, slightly differently. makes a huge difference, right? So my point on like this thing. So a lot of consultative discovery, a lot of work that we do with them. We really deeply try to understand what they're trying to solve. I have another thesis on this, right? Like, you know, the era of SAS was like you could solve a problem and then rinse repeat that across everybody. That's how we got here even in healthcare. Multitude of digital health solutions that are just point solutions solving one little thing. What we're trying to solve is the class of problems wherein this is where like, you know, Everybody can get you to that, you know, sixty, seventy percent with a out of the box SaaS solution. Right. But the opportunity with AI to get really narrow on that thing to highly customize it to your need and your particular thing, you know, workflow and then make a difference to it. And that's what we are actually enabling with this thing. And that involves a lot more consultative work and a lot more high touch and so speaker-0: Well and I think that's the critical part. I mean, having been in hundreds, if not thousands, of hospitals, in ORs, in electrophysiology labs, in interventional cardiology labs, one of the most consistent themes is that every single one of them is different slightly in terms of its workflow, in terms of its approach. And I think that was probably a little bit of the shortcoming of the SAS models is it's kind of a one size fits all. And really healthcare isn't that way when you really get down to the workflow level. So the ability to get that extra thirty to forty percent, as you say, through AI to get a hyper customized solution, I think has been the piece that's been missing for so long. speaker-1: Totally agree. And I think, you know, that adds complexity in how you manage these solutions, right? So that's where again you can use AI internally to make sure you can deliver this at scale, you can do it efficiently, you can still get the SaaS economics as a company instead of being a professional services company to that. So there's a lot of that, you know, the the layers that we need to unpack to go do that. speaker-0: And you do such a great job of articulate articulating how autonomized really is focused on taking away the administrative burden. We want to get the right procedure, the right product, the right patient at the right time, making things easier to access, faster access, more accessible, accelerating things. All of those, I think, are beautiful articulations of the direction in which healthcare needs to go. I think a lot of times people also question sort of the potential nefarious aspects of AI. And we've talking about prior authorization, for example. And there's been some discussion around concerns that maybe AI could actually help payers deny care faster. And and certainly that's something that you're working against. I'm curious on your your commentary around, you know, how do you really make sure that you're providing greater access to care and that these tools can't be manipulated in sort of a more nefarious direction. speaker-1: No, look, I mean it's it's a very important question. I think there's a few things that you know we have to remember. And this is gonna be a much broader discussion than just healthcare, right? Just because you have agents doesn't mean the agency you're absolving responsibility as an organization who's using that agent, right? Very, very important, right? It's not that, the agents made the decision. Nobody can I no court will actually allow you to get over that thing. And it maybe eventually we'll figure out models to do that. Right. So One is like I I I actually think there is, you know, there's a lot of bad rub on things like denying care and stuff like that, using AI and stuff too. I think mo the the most a lot of the states mandate that it has to be any denial of any care should be signed off even by an MD. You cannot use AI to do it. I think there's it's there in all the you know, so you have r legal recourse to actually correct that and you see a lot of that cases that happened in the past and stuff too. And even payers are very, very careful about that these days, right? Well, I think the big opportunity to me is like like look at the side of like if you're requesting care and a medication, you shouldn't have to wait for two weeks just because there's not enough human capital to review it and there's a lot of volume going through it. So to me, that's a very simplified example of how you can improve access to care. If it's an approval, no human should ever touch it. Just go through the whole process, right? things like, you know, how do you move from the reactive episodic care to more proactive, you know, planned, you know, coordination of care. Right. The example I gave you, like if you get a, you know, request coming in for a, you know, on call confirmed can cancer diagnosis, you can now instead of having humans, you know, get together and plan for it and do the coordination and not drop the ball over the course of months and weeks. you can now use agentic workflows to go automate a lot of that stuff to get the right people involved to go make this happen. So the patient's actually now actively engaged instead of actually just reacting to what's happening to them. Right. So that's another way to actually do that. there's also like, you know, the the the broader question on nefarious use of AI, it's a very general and broad question, right? One is I think we still haven't figured out as an industry. I mean, we do the basics, right? Like for example, like everybody is now doing the basics, that we do a lot of things like, hey, how do you make sure the AI doesn't hallucinate? How do you make sure the AI doesn't misinform? Like, you know, what are the guardrails? What are the things, you know, it's it's very, you know, everybody uses large language models so much. I mean, you we both know you ask the same question twice, you get slightly different answer. In the context of clinical or some of these coverage kind of issues. That could be really bad, right? So we built technology, you know, like we're gonna announce it as a public release that can be available for other platforms outside of just autonomize, to ground that always in a you know, a medical parlance and ontology. You know, we have a context graph that does that today. So our customers actually will never see hallucination or I would say never is a very distinct, but we'll catch hallucination before it does any harm. There is evals. Like a lot of the things in healthcare and AI is going to be built around evals. So how do you evaluate? How do you use AI to validate and evaluate things like that, you know, a lot more deeper? you know, to understand all the different impacts of what could happen and so forth. There's a lot of work being done on that front. So I think the industry will get better, I mean is getting better at that part. The larger, bigger question is like, what does it mean from like we do a lot of things like red teaming, right? Like we do a lot of, you know. security work on like how do you go have a team of people who know this just beat up AI and see what happens. The frontier labs do a lot of that work and stuff too. But even the larger questions are more like like who you know, who owns a responsibility when something bad happens, right? And I don't think, you know, we're we're in a place to answer that and stuff. That's like the industry needs to figure that out. What is the legal structures for actually doing, you know, AI driven care or AI driven thing? I mean, there's all this, there's a Jama paper that came out. last week or the couple of weeks ago that argues for, you know, I think it was Vinod Kosla was one of the authors for it. And he talks about like how, you know, AI doctors can outperform human doctors. And it's a very, you know, out there, controversial paper even. But then what do you need to do? And then there is value in it, right? Like for example, if you're Like in remote India or remote Africa and there is no doctors available, I would at least like to have an AI doctor. If I we don't have real doctors, right? But then what kind of, you know, what kind of legal structures, what kind of accountability structures can you put in place? These are big questions that I think, you know, if we put our minds to it, the industry we can answer. But it it's gonna be it's gonna be a long road trying to get there, to be honest, right? But I think I do think that those are bigger questions are still being thinged Part about like, hey, when we run an AI that actually is recommending something, yeah, there is another AI looking at it to make sure it doesn't, you know, make up some BS, right? So those kind of things are being solved today. The larger accountability questions are still open. speaker-0: Yeah. Yeah. And it it strikes me that they're they're big questions. Some of them are not resolved, but part of it people throw in, you know, the term human in the loop as kind of a reassurance for for these types of things. How how are you thinking about human in the loop as it applies to autonomizes solution? Yeah. speaker-1: I think there is like two parts of it. There's human in the lead, wherein like, you know, you can have this power of agentique AI as agents that'll just be an extension of our human agency, right? And we're seeing that. A lot of the cases where care managers need to understand what's the next best cost of care. They unleash a bunch of agents that'll understand, you know, different specialty and looks for different reports and stuff like that. Come back with the recommendation so they can act on kinda like your deep research agent, you know, for for medical cases, right? And then there's human in the loop wherein you have workflows like which are highly transactional, high volume workflows, but you know you need to have guardrails and points within that workflow. You don't want a agent to make up a decision. You want because of the accountability problem, you want to go pull the right human being into that particular loop and make a decision. So that's the human in the loop side. And I think we we we we support both of that. What we're seeing happen is, you know, and it's an evolving thing, right? Like there's initially there's a lot of lack of trust for the humans with the experts in the workflow with agents doing it. So they almost like human in the lead initially, they're actually driving all of these stuff and you know agents are reporting back. They have to click go, yes, confirm and stuff like that. But as when you do that, you're saying, look, every time I've said yes, you make a recommendation of this kind, I'm saying yes. Why why don't you just do it? And then I will do I'll put in another audit process. To make sure that you know we can actually look at this thing. And that could be an AI audit. We've done a lot of that things like that, like you know, different use cases. But the whole idea is like, look, I have high confidence because the users are saying, I you get it right, you got it right, you got it right, you got it right. It's 99.9% right. I think that's an area that I can just say I can take off the entire workflow because even if I'm saving four minutes of per case on that particular thing, and I multiply that by hundreds of you know folks times every day. it's it's gonna pay out itself. It's gonna be a huge ROI, right? So there's things like that. I think the the there's humans will I mean I I also do think that all of these models and so they're all getting smarter and stuff too. But that doesn't mean that it hell healthcare is a lot about care and about the human experience than it is about data gar you know, munging data and making recommendations, right? So with that in mind, I think I do think that At least in healthcare, it is gonna be human in the loop and human in the lead for a foreseeable future, right? In other than in places like what you said, like where you have zero access to care. There's nobody's even there's no doctors in that particular zip code or like ten zip codes around it. Can we actually enable a symptom checker that helps them navigate whether they have to take the eight hour bus journey to the next, you know, hospital or they can just, you know, use some over the counter medications to take care of? Those kind of use cases where you'll start seeing fully autonomous care and stuff happen. I do think though, like you know, that that said, I'll also cla clarify, I do think that healthcare as an industry is very much set to go into a future where operations are going to be autonomous. And to me, that whole take out the admin layer, we foresee a health enterprise in the future where we'll have doctors, nurses and almost They will own all they will do is actually engage with patients, give them care, practice, as the practice says, and then almost everything else, from mailing the note out to putting in a prescription to getting you know, follow-up visits rescheduled to, you know, engaging and doing follow-up calls and making sure they have appointments or lab reports being lab orders being put in, everything is going to be done by an agencique layer. And we wanna be that layer. We wanna be that layer that actually We're we're well on our way to actually ha ha having that happen across different parts of the industry. But I do think that that is the big opportunity that we will see in the future. speaker-0: Well, and to that point, if you're to fast forward then five to ten years from now, and what do you think we'll look back on and we'll say, we really got that right with healthcare AI? Is is that the piece that you think we'll really have gotten right? And vice versa, what do you think we'll still need to go try to fix? speaker-1: I think, you know, five five, you know, ten years is too long. So let's say two three to five years. I I wish, you know, the the big ideas to me is like we move the conversation at healthcare from breaking fix and, you know, getting treatments and that becoming the he becoming the headlines to, you know, or not having treatments and stuff to really elevating the conversation around like, hey How do you increase health span, not just lifespan, but health span? How do you live longer more healthily? Talk about longevity. It's a very taboo topic in several societies and communities. But I think we because we are still stuck in the today and we're trying to solve these problems that are practically in front of us. My hope is that in the next three to five years, we'll get to a place where a lot of the mundane infrastructure problems that we talked about will be taken care of. Because AI also brings in humans. You know, today LLMs probably are more variant on their responses than humans, but even practice and people of the same from the same college practice differently, right? So the lot of the consistency will be brought in by technology, right? And then the human touch can be layered on top of that. So my goal is, my hope is like in the next three to five years, we'll move the conversation from healthcare being episodic sick care problems to more long term, you know, preventative care, to talk about longevity, to talk about health span extension, all of that kind of stuff, right? Now what does that also mean from a health enterprise? It's like, you know, you know, we it'll be a profitable exercise to actually have a health enterprise because it's not today, right? Use like single digit margins if you're and if you're a hospital chain and stuff like that and all of that stuff. So I think it'll be a lot more profitable to do that. I also think that patients are gonna play a much more active role in healthcare. I mean one of the things around healthcare around the world is patients are not as engaged in healthcare as they should be. Because they didn't have the tool because they, you know, they had to go through a medical schooling process to actually even understand what is going on, right? Well not anymore, because the information, the knowledge and even the access to that kind of information becomes available through an API, through a know, a chat bar or an app or so forth. So that is gonna shape healthcare more than we all give it credit to. It's not technology forces, it's the patient force because they are gonna be this is the age of a more informed patient. And you know, like if you go to any doctor, like most doctors, my doctor friends tell me, like, hey, this 15 minute appointments that we used to have won't work anymore. Because the patient comes in because they've already put their question, you know, their medical reports in Chat GPD or Cloud. They come back with 20 questions. They're very informed. They ask me questions. I have to show up, you know, with reading the research, what's available, or I have to do follow up. Or I to put my PA to actually do all of that stuff. It's changing. I mean, cardiologists, if you ask like their whole diabetics management, it used to be a big problem, right? People say you have diabetic. I'm like, huh, okay. And then they'll send them educational materials to just watch their diet and stuff like that. It's changing. Slowly people are getting more healthier. They're actually more informed about this. So I think it's gonna shape the biggest force that's gonna shape healthcare in the next three to five years is gonna be the patient and how they're gonna be more engaged because of it. speaker-0: What do you think we'll still need to go fix five years from now? speaker-1: in five years from now, we'll get into a lot more deeper moral questions around healthcare, right? What is the right level of care? You know, should everybody have access to I don't know, ge you know, gene-based therapy? You know, should we do, you know, DNA cloning? Do we I mean all of those will become the bigger questions? And and again, this is my assumption based on the fact that well, we will solve the easy problems. Then the question is gonna be like, this is great that I can do. you know, like even today, like transplants. We talk about kidney transplants and now you have on a list and stuff like that. Well, in five years you'll have synthetic kidneys that people will just have to start attaching to, right? It'll be grown in a lab and things like that. There's already work being done. So then the question becomes, how does this change into, you know, how how does access change for different, you know, parts of the society? And so we'll hopefully we'll I mean th it's a good problem to have because that'll just force the real dialogue around How we live as a society and a civilization, right? You know, there might be some interesting emergence of like, you know, like I there's a lot of bets on home health growing, right? In terms of like home as a site of care. it's it's like, you know, people having personalized health coaches and you know, things like there's a lot of things that's gonna come in, but something I need to think about, like, you know, in terms of how the next five years is gonna shape up. speaker-0: Yeah. Well it is cer will certainly be fascinating to follow and to track and see how things evolve. But in in the meantime, really excited about what you're doing and the advances you're making and really reducing the administrative burden for for the clinicians and the practitioners that are out there. So thank you so much for coming on today and discussing your background, your journey, your perspectives, and all the things you're doing at Autonomize. speaker-1: Of course, Mari, thank you. It was so much fun. Thanks for your very thoughtful questions and enjoy being here. speaker-0: Absolutely. Ple pleasure to spend some time with you, Ganesh. And thank you all of you for joining us today on Powering the Future Within. If this conversation inspired you, follow the show and share it with your med tech community. I'm Omari Baugnight. Stay inspired and keep innovating.