Better Math, Better Medicine: Mohan Giridharadas Unlocks Capacity Across Hospital Hubs with LeanTaas

(upbeat music) Welcome back to the future of Health Care AI podcast, I'm Paul Mosquitz, and I'm here with my co-host, Devon. Thanks Paul, that was a great conversation with Mohan Gary Harness, the founder and CEO of LeanToss, which is one of our portfolio companies we invested in in 2022. LeanToss is a really fascinating and high growth company that provides predictive analytics and operations optimization software to over 1200 hospitals and 200 health systems, really innovative, exciting company. It's always great to talk to Mohan. Totally great. Mohan just has such a fascinating background. He spent nearly 20 years in McKinsey developing their Lean Six Sigma practice, working in super complex logistics sectors, I think air traffic control systems, yield optimization for airlines, really complex scheduling systems where capacity optimization was key to profit. He had the realization that it would be interesting to try to extend some of those learnings into other areas and other sectors that hadn't benefited from the same automation curve. And so he saw a huge opportunity in healthcare where capacity optimization and league workflows were not yet part of the common parlance. But he was able to translate all those insights that he developed from a consulting perspective into vertical software and build an AI native company that's really transforming the operations and the revenue capacity of health systems all over the country. Yeah, it's really awesome to see how that AI native approach was useful in family company in this space in really revolutionizing a new area within healthcare IT. And it's also really compelling how the whole company has really embraced this AI first mentality, not only just in the products that they build for their customers, but in the whole company. AI first mentality around how do we do finance? How do we do marketing? How do we do HR? A lot to learn that we've taken away from Mohan, the team that is relevant for any company really. So for those interested in how platforms like MENTAs are untangling the operational complexity of the health system, helping to drive revenue throughput, increased quality and access. This is gonna be a great episode for you. Mohan, thanks for joining us today. We're looking forward to talking about LENTAs. It's great to be here, Devon. Thanks for having me. It's great to have you on the show. Before we get into some of the content, you just have such an interesting background to be great to share with some of the listeners your career journey, both as an engineer and then the leader of the Lean Six Sigma practice and McKinsey. What is it that ultimately led you to healthcare and to founding MENTAs in the first place? Over the 18 years I spent at McKinsey, I had the privilege of leading LEN operations work at excellent asset intensive service businesses, airlines, logistics, banking, just operated a staggering scale. And my aha moment that led me to leave McKinsey and start LENTAs at the end of '09, was that I realized that even the most sophisticated clients were driving operational improvements on the backs of Excel spreadsheets. And my thought was Excel is middle school, Matt. Middle school, Matt will give you middle school answers. So what I wanted to do was drive operational excellence and sophisticated map, optimization, simulation, machine learning, data science, and AI even before it was cool to call it AI. I also felt that the consulting model created dependency on the consulting team. And that after the project was done, things tended to go back to the way they were before. And so my thought was, if Amazon can tell you what to buy and Netflix can tell you what to watch, surely we could build software that use sophisticated map and then made the recommendations to the front line, which should then be more enduring and permanent than being dependent on a consulting team telling you that. And Stanford held Ketonda to be our very first customer. We were doing patient satisfaction work for them in 2010. And then we were also in multiple industries, Google, Flexronics, Home Depot, et cetera. And then the breakthrough was in 2013, our executive sponsor at Stanford, who was the head of the cancer program, suddenly said, why does my infusion center look like a train station in the middle of the day? But it goes down in the mornings and nights. And we ended up solving the infusion optimization problem and realized that we were on to something, that mathematical capacity unlock was a thing. And from there we went on and built an over product with UC Health and then an inpatient product. And today, 1200 hospitals below me into 200 health systems use one on waterfall products. Super impressive how you built the company and the vision that you've had, Mohan. Maybe following up a little bit on the airline reference you made during your time at McKinsey, I think you've often referred to Lintoss as the air traffic control center for a health system. Can you explain to the audience, what does Lintoss do and how you evolve from that initial start in infusion centers to other parts of the hospital health system as well? So one of the ahas from me was thinking about capacity in manufacturing you can fake it if you don't have capacity. If you get an order to ship a thousand widgets and you can only build a hundred widgets, you can pull 900 from the shelf and ship it. But in a service you can't pull it from the shelf because the supply and demand have to match. And so the ahas from us was thinking about how airlines manage the supply demand balance of something as complicated as an airline seat. And they do it dynamically and at a staggering scale. So I got to watch, they call it yield management and there were a thousand people are dealt are doing that. So think about the complexity of yield management. Every day delta flies 5,000 flights with 200 seats of plane. So there's a million seats up in the air and you can buy tickets 300 days into the future. So there's 300 million seats up for grabs. Every minute of every day delta knows the supply demand balance for the demand of every one of those 300 million seats. They can tell today that the March 15th, 2026 6 a.m. Jackson, little to New York is running ahead of plan. And therefore we should not offer free tickets. Meanwhile, health system struggle to balance the 50 appointments that are coming over the next two days. There's clearly a gap in the sophistication of supply demand balancing that we started to use. And as we got into it, we realized that there are many other learnings where you think about how hubs and spokes operate, main assets versus secondary assets. And we've incorporated all of those. So we're easily the first company in the health and analytic space to bring the sophistication of yield management and airline operations into health care. And it's incredible, you know, what we've seen over the last several years is that that matching of supply and demand across these scarce assets with really fixed cost structures can drive incredible revenue on lock for health systems because they're able to flow more volume through the same business and same fixed cost that they had before. Oftentimes we're seeing customers with staggering ROI is 510X plus. You know, as you think about the health system of the future, we are probably going to have to operate in a much more streamlined way. Reimbursement is going to continue to be challenged. How are you thinking about helping those health systems do more with less? And what do you see as the most exciting opportunities to help them either bend the cost curve or just become more productive over time? So we focus entirely on this concept of capacity and flow because in a bottleneck asset, they're reinforcing loops. If you unlock capacity, patients flow faster, just like when they're fewer cars on the freeway, the average speed is higher. And so the old alternative was to just build more. That's equal to build more freeway, build more lanes. But when you start running out of capital to do that, you've got to make the lanes you have work better. So we are using our algorithmic insight to unlock the capacity which then unlocks flow. And you can do exactly as you said, see more patients, et cetera. Now, when you take the whole complexity of the health system, how do you do that? Let me again, go back to the metaphor. Delta serves 200 airports in the US. Do you know how many hubs they have? 9. That's it. 9. With 9 hubs, they're able to move velocity across 200 airports. So now, if you take 10 years ago, Delta moved 160 million passengers a year. Today, they moved 200 million passengers a year. 40 million more per year means three or four million more per month. You'd say, how are they doing that? Well, it's not at more flights. They had 5,000 flights a day back then. They've got 5,000 flights a day right now. It's not at more speed. The same 737s are flying at the same speed. It's not bigger planes because they took the super small regional jets and made them slightly less, super small. But the average load factor of a plane hasn't changed than the number of seats that much. So what are they doing? They're doing two things which are remarkable. They're filling their planes better, which is the yield management I talked about managing the inventory of 300 million seats. And second, they're cranking their hubs faster. The turnaround velocity in a hub is much more. A second-tier airport doesn't matter. They move the hubs faster. With the hub velocity comes the velocity through the whole system. Now, let's apply that to what we're doing in health care. The core hubs, just like Delta has nine across the country, there are five or six hubs that matter in a health system. The ORs, inpatient, imaging, emergency departments, oncology services. The others are satellites. the individual oncology clinics, the individual surgical clinics. are like second-year airports that are a little further away. So they don't contribute to the flow. If you then get utilization to work better, you can do the same thing as in terms of velocity. And that's how we are applying it one separate time. That's been really impressive how you've taken some of these concepts and applied them to healthcare. And what's really awesome about it is it's making healthcare more efficient, but also increasing patient access. And yeah, that's obviously something that's great for patients and it's great for the overall health system. So it's really great to see what you and the team are doing. You also reference the complexity of the health system and getting all the clinicians, all the administrators to align on how things are going to change when they're really harnessing all the power that's embedded in lean toss. What's your approach to this change management opportunity that is really presented by your software? It is a very interesting journey because we created the category. In the beginning, people just accept it as a way of life that things are crowded and that things take time. We all accept for instance, we're going to drive downtown and rush hour, it's going to take you twice as long. And we just accept that that's the cost for big business that's life. And so in the beginning, it was quite a push to get people to understand that there is a different way of doing it. And so the early pilots proved it. And when they suddenly started to see dramatic results like a 50% reduction in wake time, not 2%, not 5%, 50% of 15% increase in capacity, they started to get it. What we realized then is this cannot be software thrown over the wall. So the task in lean test stands for transformation as a service and we stand by it. We take full responsibility for the training. This is not a train the trainer, throw the software over the user guide and hope for the best. We take responsibility for the training. We train as often as we need to. We have teams that monitor usage. We push many help guides in the form of tool tips in the app or video veneers through the app. We structure our commercial agreements in a manner that doesn't inhibit engagement. When I call our corporate lawyer, I can almost hear the $600 an hour meter running on the other side of the phone line. We deliberately don't price for professional services because if we did, the health system will say actually let's do it on our own. We don't need these guys. We'll get an invoice. We don't charge for travel, which is contrary to growing up in a world where you automatically send your travel bills. We don't charge for travel because we don't want the health system to tell us actually let's do it on zoom. If we believe we need to engage, we would rather eat the cost of doing that. So the way we frame it is, this is an all-in solution. You're paying for it on a per asset per month basis. And all the support services you get are included in the price of that. And I describe it to executives like it's like hiding NASA to do your physics homework for you. You will get a very good answer that's mathematically sophisticated and more than you could have done by yourself using Excel. I would have liked to have NASA help me with my math homework when I was in high school. Wouldn't we all. Yeah, one question that's really interesting is you think about scaling a business. I mean, Mohan, you're a rare executive that's taken a company from founding 0 to 1 to 100 million of ARR. Now close to 200 million of ARR that's very rare in healthcare IT. But along the way, we've actually made some pretty interesting strategic acquisitions as well as some very interesting strategic partnerships. In particular, we acquired Hospital IQ, which brought us a very interesting inpatient asset. We've acquired some surgical assets for scheduling. And now starting to enter into some very interesting partnerships with vendors in the AI ecosystem like Hippocratic AI. I'd be curious if you can tell a listener's a little bit about how you think about when to build versus buy and where M&A fits into the product roadmap of a growing company in healthcare IT, as well as like how you think about the role of partnerships in the world of AI. So we are very clear about what our unique expertise is. We can mathematically optimize capacity and flow. That's our wheelhouse. They're very strong on that. So within that path, if it's tightly in that wheelhouse, we'd rather build. And so as we are going from adding to a node or expanding the interconnections between the nodes, we'd rather build. But at the same time, we can build everything. It's a resource constraint and there's a speed constraint. So we're very thoughtful. We look at many different factors. One is, doesn't adjacency enhance our product. So some of the things we bought in the surgical suite area are things that yes, if we had two years and if it put resources against it, could we have built it ourselves? Sure, we could have, but someone had built it successfully and it amplified our product and created the adjacencies. So we will acquire for adjacencies very quickly. We will potentially acquire for new nodes where if somebody has a node that we haven't touched, we know it takes a couple years to build a node. If we find something where they've done an operational capacity improvement, we may acquire that. We look at all of those. Now on the partner side, we have to be thoughtful because there are many things where there are many people playing in the game. And in that case, we need to be Switzerland. We don't need to pick a particular winner. And so in those cases, we will partner with potentially one, but do it in a way that it could be a plug-and-play. So you could imagine a voice agent or an identification of email interactions, etc. where we don't necessarily want to pick a winner. We will deliberately be Switzerland. Another example could be the staff communication platform. We don't want to pick a winner because some hospitals use teams, some use secure chat, some use title connect, and some just use SMS on the phone. And so if we picked one vendor, then we can't be effective in another health system. And so in those cases, we decide to be very flexible around how we do it. So with each of these things, we've got examples of all of the above building it, acquiring it and partnering. And that's the level of total fullness we constantly bring to it. Yeah, you've been really thoughtful about that approach that I think has worked really well. It's a framework that you know, I've looked to kind of replicate in some of our other investments. So I think it's been really a smart of you. One other thing that I've been really impressed with, I mean, the company that you founded was based on machine learning and AI from the start. But with respect to generative AI, you went from conceptualizing what that would mean to lean tossed, actually having a product extremely quickly. Can you talk about how you absorbed that concept in the organization and got it launched so quickly? It was super impressive the way that you and the team did that. We looked at it and tried to figure out what's the right use case for us as a starting use case for us to get smarter about it. And the biggest risk for us was the notion of hallucinations where if you turn a generative AI agent loose on the internet, it comes up with potentially random answers. And for us, if we pride ourselves on mathematical precision, if one of our customers asks a simple question like how many hours of block time did Dr. Smith use last quarter? And our tool hallucinates and comes up with a random answer, then suddenly we are casting doubt on all of our analytics. So we had to make it very, very tight. That actually helped us because if it was such a constrained generative use case that we could not let it wander outside of the specific data set of the specific customer. So the initial use case became very simple. We said, let's pretend that the genie tool is as if you had a lean task expert navigating the tool in front of you. So any question that a customer could ask how many hours of block time did Dr. Smith use? Well, if one of our product experts was there, they would know the sequence of screens to click to get there. So we then structured it. So it generated the query that generated the right result and confined its field of view to exactly the data for that customer, which then gave us a genie AI use case, a chatbot like assistant that was always correct, never, never steered off course. And we were able to deploy that very, very quickly and show that we could put a conversational UI front end onto our platform relatively quickly. Now, of course, we're expanding and doing other things like synthesis and so on, but that's kind of how we got to it very quickly. Now, I think that the learning around having a narrow enough use case that you could get after quickly made a lot of sense. And I think that again, that's a, that's I think a good, a good lesson to be applied elsewhere. And now how you're taking that learning and expanding it into into new areas. One of the other just follow up. I've been impressed the way you and the team at LeanToss have what I call an AI first mentality. You're launching generative AI products for your customers, but you're also applying those principles to how you run your own business. Can you talk a little bit about that philosophy? Yeah, for us, the interesting part is when we look at all these debates of will AI take your job, not take your job, etc. We find that that just creates needless panic. It's more productive for us to look at it and say, today we are 400 person company with this level of footprint. Let's fast forward four years and see the size of company we would be. And then what does it take for us to not have to grow the head count in exactly the same way? Had none of this genie I stuff happened, we know what our default would have been. And so can we bend that growth curve? And in order to bend it, function by function, what are the areas where having a powerful geniac internal toolkit would help us. So for instance, our teams help our customers month after month, do special analytics and review their performance and find the errors that went well and didn't go well. This is a perfect Genie I internal toolkit. It's again, bit like putting NASA in their at their shoulder, helping them solve their problems. When our sales team prepares for a meeting, we're going to meet or 10 people. What have they said publicly? What do they know? What do they order their biases? How do they think about what's the background? We're now unleashing internal Genie I toolkits to help our sales teams be more prepared for meetings. So we're going function by function and building selective internal tools to amplify the productivity and effectiveness of our people and to do it in a way that brings folks along because we don't want our people to be nervous about is this signaling or headcount reduction? Is this Genie I is coming for your job kind of a thing? For us, it's more about make us all more effective at our jobs and help us grow the company in a more productive way than we otherwise might have. That really resonates and I think is we have conversations with a lot of our companies. That is really the modal theme that people are thinking about Genie I from an internal perspective is how to do more with the same resources that they have today to make them more productive and to drive more equity value for employee and ultimately to make the company more valuable, free and more operating resources to accelerate our G-Cycles launch products faster and ultimately potentially be able to deliver products at better value for customers which is really exciting as well because price sometimes is the key hindrance to adoption. So I think it's great that you guys are harnessing that mindset as well. One fall on question to Devon is I guess as you look forward into your own product roadmap and you think beyond this initial co-pilot use case what are the other applications that you see for Genie I across your product portfolio and two or three years from now how do you see people using Genie I inside of Lintoss? So when we go back to the roots of how we think about optimization we think of it as nodes and edges so we are optimizing the individual nodes of the OR and the impatient the edges of the connection points to optimize the whole system we have to start swimming upstream. So imagine going let's take OR as a use case going all the way to the clinic where the patient had their first encounter with the surgeon and they jointly decided surgery needed to be done. There are then 50 things that happen between that conversation and the patient actually on the table in the OR getting their surgery done. Those 50 steps lend themselves to a lot of agentic and Genie I work. For instance we could generate voice agents that call the patient and start sorting our prioritization whatever else is required. We could spool agentic AI to have workflow like email back and forth to get the documentation lined up. We could start spooling Genie I agents to research the entire encounter set of that patient across other health systems pull it from the health information exchanges the Q&N network that's out there summarizing synthesize and surface risk issues that the surgeon might want to consider so you could take that entire journey and optimize it. Part of the problem with healthcare is they encounter bottlenecks by getting stuck in it. Think about how we all drove before we had navigation systems in the car on navigation systems on our phone. The first we knew about our traffic snaffos when we got stuck in it until then we didn't know about it and then when we got stuck in it the only way out was if we happened to know the neighborhood otherwise we had to just suck it up and work our way through it. That's what healthcare does with its processes. Imagine now putting an navigation system in there where today I can say I'm going to land in JFK at 6 o'clock tonight and then drive to Manhattan along will it take me and it knows the three routes and how long it'll take. Imagine we could start doing that for all the routes that a patient has to go through and dissipate the bottleneck, work your way around it, use Genie I to facilitate it. Then all the pathways start to work as if you've got navigation systems activated to help them there. So everything we do is around unlock capacity and dissipate the bottleneck, use technology to relieve the bottleneck, increase velocity which then improves capacity. So we're creating a self reinforcing mechanism which will help the health system as a whole do what I described in the notion of C30, 40, 50% more patients with the same resource footprint. It's really exciting vision that would certainly be a lot for patient access and efficiency of healthcare. When you step back, Mohan, even just thinking beyond what you're doing at Lean Toss, what do you think are the most interesting areas where AI is really going to disrupt healthcare more broadly? I think we haven't spent any time on clinical AI but the notion of clinical AI getting to be reliably good is just a game changer because of the shortage of skill clinical staff. The reality is we hold clinical AI to a higher standard than we hold humans. So when we get a image reading AI algorithm that's wrong 8% of the time, we get very upset but humans reading scans are wrong 20% of the time and that's okay. So we still got a bit of a disparity on that front but if you could imagine that happening and then start to amplify it with the ability to provide global clinical help to places like rural America or developing countries around the world that don't have the infrastructure to have the resources, suddenly you can start to do that. I look at the Optimus robot that's coming out that can do highly dexterous physical things and if you imagine 10 years out having humanoid robots transporting patients, part of the problem is patients get stuck because they can't be transported from the inpatient setting to their long term care facilities etc. If we started to have that sort of help then the cost curve goes to zero when you suddenly have a machine capable of doing these things. So the expansion into clinical and the expansion into intelligent dexterous humanoids just opens up runs of possibility that I don't think we can fully comprehend this yet. And what do you think when you think about health systems really harnessing AI? What do you think are the capabilities or things that the health systems really need to do to to really realize the potential of AI and generative AI? I think health systems are built to be conservative. That's probably a good thing. If I'm undergoing surgery I don't want to surge in using a scalpel ephanodamazon last night. I'm happy that he's cautious and uses only equipment and processes that I know. So I think it will be slower than the technologists would like. But probably that's not a bad thing. What I see happening which is going to be the biggest unlock from a clinician's perspective is the need to have your hands on the keyboard and your face on the screen continuously which is what happens today. So what the EHRs do is they almost require the physician to be engaged with the EHR rather than with the patient. When we start going into an ambient listening mode then suddenly the physicians and all the providers can talk to the patients without worrying about keeping up with the notes. And that just changes the dynamic entirely. That will be the first place where I see the broads acceptance of AI as a tool. And then if we can deliver a non-black box algorithm so they can see what happened and they can question it. Operational AI will be easier to accept than clinical AI for a long time. And because people are used to accepting your navigation system telling you better routes and not saying I know better than you have lived in this town for 10 years we all accept and listen to our nervous systems. So they will listen to systems that guide them operationally. So that's the sequence I would see. I would see clinical coming out slowly as a copilot, second gas on the more routine things before it swims upstream to the most sophisticated. It's super interesting that you mentioned machine vision and ambient listening as this massive unlock. One of the things we've also been hearing about in this series and from other entrepreneurs is this idea that the advent of machine vision will create a much more liquid market for the supply and demand of healthcare and the ability to beam in resources whether those are agents, avatars, humans in another place such that you don't necessarily have to have all those clinicians on staff all the time and really break down the fixed cost problem in healthcare. I'm curious over how many years do you think that evolution to a much more virtual dynamic workforce will occur? And what do you think the mix of the resources between clinicians as we see them today versus more virtual clinicians will be in the future? It can never go as extreme because at the end of the day it is a contact sport. At some point some physician or some caregiver has to have their hands on a patient. And so what the virtual sitting and so on does is just increases your span of control. In a physical world you can one nurse can watch four beds, maybe they can watch eight, maybe they can watch 16, maybe they can watch 32, but you cannot get it so far removed that if something were to happen it takes too long to get a person there. So if it goes fully remote and fully liquidized as a marketplace and you start to see an issue pop up and it takes you an hour to get a physical person into that room. It's too late in a clinical setting. So it's got to be like 911 response times in seconds, not hours, right, which then automatically puts a bit of a threshold on what you can do. I think where you get the amplification of the machine vision is where you can implicitly train. And the problem you've got potentially nowadays is the tenure of nurses are skewing towards the more junior. You've got many retired nurses and not as many filling it up. And so you've got a more junior workforce. Now the easiest way to train a junior workforce could be with one senior machine vision driven AI nurse guiding and providing positive reinforcement where for instance, a junior nurse turns a patient over to avoid the pressure injuries and the AI documents the notes and gets that nurse to click that yeah, by turn the patient left to right and years what I did. And that's the reinforcement learning loop. So it can start to amplify the training and the skills. So we'll definitely get enormous leverage out of it. But I don't think it's, it's a completely virtual world just because somebody's got to close the loop with a physical contact at some point. Mohan, when you found it, lean tasks, you know, you took some twist and turns to get to where you are. And it's really amazing to see this this company that you built. If you could give some advice to your former self now, having been through and created this really awesome company, like what would you tell yourself when you were first, first starting, starting the business? I think back to that one of my favorite quotes is the Wayne Gretzky code. You never scored on a short, you don't take. You've got to take a lot of shots. A lot of them will flame out, but he just got to take a lot of shots. My favorite one of taking a fly was Liz Concordia, the CEO of UC Health, gave me 30 minutes on the phone. And we didn't have money for travel, but I said a phone call wouldn't do it. An in person meeting would matter. And so I flew to Denver and saw her in person. And in those 30 minutes, she got excited about what we're building. Agreed to do infusion with us. Ended up building the OR product with us. Ended up bringing one of our seed investors and truly transform the company. Little things so easy to look at the 30-minute meeting and say, yeah, it's just one more meeting. Nothing's going to come off it. But there's something in there that said, maybe something will come from it. So there's a little bit of trust your gut on taking shots that matter, but then don't hesitate to take the shots. Because if you run the cost benefit on everything, it paid no sense to fly to Denver for a 30-minute meeting with someone I had never met before. But it probably is one of the pivotal things that happened. I also think that in a startup, you can't be half in. You can't hedge your bets. Either all in or you're not in at all. Once you're committing to a path, you've just got to get everybody lined up behind it and you've just got to go flat out. Play the the ORG and buy 17 lottery tickets and hope one of them hits that just doesn't work. Well, Mohan, this has been an awesome conversation. Thanks for taking time to join us on the podcast. Great, no always great to chat with you guys. Take care. Fantastic conversation. I think we all learned a lot about how you can really unlock operational productivity inside the health system, fascinating and essential problem to solve for society. More to come next time on the Future of Healthcare AI podcast.