Interview with Lean TaaS founder Mohan Giridharadas

In this episode I welcome back Mohan Giridharadas, CEO and founder of LeanTaaS, who shares his  journey in revolutionizing healthcare capacity through AI and prescriptive analytics. Mohan provides insights into LeanTaas's evolution since pivoting to healthcare in 2015, tackling pressing challenges such as increased demand, financial pressures and staffing shortages. With innovative solutions now optimizing a third of the U.S. chemotherapy capacity and enhancing 15% of the nation's surgical capacity, LeanTaaS is transforming the landscape of healthcare efficiency. Our discussion highlights the  strides made in optimizing infusion centers, operating rooms, and inpatient beds.

We explore the intricacies of improving hospital flow using predictive AI. Mohan and I discuss the similarities between hospital bed management and hotel room turnover, offering insightful comparisons. By employing algorithms to predict bed shortages and prioritize discharges, hospitals can improve patient flow and efficiency. The conversation extends to interconnected issues like ambulance capacity and post-acute care constraints, emphasizing the potential of predictive analytics to revolutionize hospital operations and patient care.

Finally, discover how LeanTaaS is harnessing the power of generative AI with their innovative iQueue products. Mohan discusses the challenges of preventing AI hallucinations and the importance of human oversight in decision-making processes. We also touch on LeanTaas's cybersecurity measures and the critical need for robust data protection in healthcare. Additionally, Mohan shares his excitement about a newly released book and reflects on the company's growth, including reaching a billion-dollar valuation. 

As of March 2025 HealthBiz is part of CareTalk. Healthcare. Unfiltered and can be found at the following links:

• Spotify https://open.spotify.com/show/2GTYhbNnvDHriDp7Xo9s6Z
• Apple https://podcasts.apple.com/us/podcast/caretalk-healthcare-unfiltered/id1532402352
• YouTube https://www.youtube.com/@CareTalkPodcast
• CareTalk website https://www.caretalkpodcast.com/

Host David E. Williams is president of healthcare strategy consulting firm Health Business Group.

Episodes through March 2025 were produced by Dafna Williams.

Transcript

0:00:01 - David Williams
Hospitals always seem to run at full capacity, making it hard to get a bed or schedule a surgery or infusion. But are they really maxing out, or are they just not so well optimized? Today's guest first appeared on the podcast three years ago, when he discussed how his company was bringing the optimization to healthcare to match the capabilities of a major airline or logistics company. He's back today to share the progress they've made since then. Hi everyone, I'm David Williams, president of Strategy Consulting Firm Health Business Group and host of the Health Biz Podcast, where I interview top healthcare leaders about their lives and careers. My guest today is Mohan Giridharadas, ceo and founder of LeanTaaS, which seeks to maximize healthcare capacity with AI and prescriptive analytics. Do you like this show? I hope so, and if you do, please subscribe and leave a review. Mohan, welcome to the Health Biz Podcast. 

0:00:50 - Mohan Giridharadas
David, good to see you again. 

0:00:52 - David Williams
Yeah, welcome back, as I should say. Now there's a little mess up with my format because I ask people about their upbringing and their childhood and all that, and it doesn't do as well for a sequel. So what I'm just going to guess is not everyone remembers exactly what you said three and a half years ago on the podcast. So if you wouldn't just mind refreshing us a little bit on your upbringing, your education and any childhood influences perhaps that have stuck with you throughout your career? 

0:01:15 - Mohan Giridharadas
Wow, that's a tall ask. So, david, as we talked last time, born and raised in India, my undergrad was in double E at the IIT in Bombay. Came to the US, to Georgia Tech, to get a master's in computer science. Spent five years as a software engineer at Intel and mentor graphics. Went to Stanford Business School. Then spent a long time at McKinsey 18 years at McKinsey in many offices. Ended up running McKinsey's lean manufacturing and lean service operations practices. Left McKinsey at the Manufacturing and Lean Service Operations practices. Left McKinsey at the end of 2009 to start Lean TaaSs because the idea was imagine delivering a lean transformation, but on a software as a service platform and putting in sophisticated math, simulation, optimization, machine learning etc. To deliver process improvement, capacity optimization etc. Initially we were across many industries, but in 2015, we pivoted entirely to healthcare and today we have three products one to optimize infusion centers, one to optimize operating rooms and one to optimize inpatient flow. So that's the quick recap. 

0:02:22 - David Williams
Got it Well. You know you specialize in healthcare. Of course it is 20% of the economy and growing at least as fast as the economy, so you haven't put yourself in too tiny of a niche by doing that, thankfully. 

0:02:34 - Mohan Giridharadas
That is for sure. And it is a pressing problem because there has been a seismic shift in all of the forces at work. Demand is growing because more and more people are covered. The population is aging, there are more chronic diseases, so the need for healthcare services has just gone up and up and up. Meanwhile the financial pressures facing health systems have also increased and therefore the old playbook of build more hospitals, build more operating rooms, build more facilities has sort of run its course. And meanwhile there's pressure on reimbursement with the whole value-based care and the cutting back on reimbursement levels. So hospitals are squeezed from all sides and therefore need to find more innovative ways to get more out of the existing assets they have, since they can't just add more assets. And an overlay on all of that is the staffing crunch, which is always kind of there in the background, has become much, much harder and is likely not going to get better anytime soon. 

0:03:38 - David Williams
Now, when we spoke back in 2021, I do remember speaking about the Infusion Center product, and that's one that I have some personal familiarity with trying to schedule an infusion for one of my relatives and seeing the challenges of it Sounds like you've got three products now. Maybe you had some of those at that time as well, but what have been some of the main accomplishments or progress of the company since we last caught up? 

0:04:02 - Mohan Giridharadas
Yes, I don't know exactly where we were in 2021, but the infusion product was invented at Stanford in 2013. Back then it was one infusion center, 50 chairs. Today we control 700 infusion centers, 14,000 chairs across the country, which is roughly a third of the chemotherapy capacity in the US is taking place in an infusion center optimized by LeanTest's IQ product. That's an enormous responsibility, just because in a given month, more than a half a million patients are getting their chemo optimized by our product, which is great. The operating rooms product was invented in 2016 at UC Health in Colorado. Back then, one hospital, 50 ORs. Today 600 hospitals, nearly 6,000 ORs. So 15% of the surgical capacity is optimized on our platform. And then inpatient beds is our newest product. We've worked on it off and on for the last six or seven years, but we really pushed it forward in 2023. So it's only been about two years now end of 22, early 23. And today 25,000 beds at about 100 hospitals are optimized on our inpatient flow product. 

0:05:18 - David Williams
I'm interested in this concept of the flow, the inpatient. Why do you use the term flow and what does that mean? 

0:05:28 - Mohan Giridharadas
So when you think about an inpatient bed, we've all experienced this when we go into the hospital and are waiting for a bed either post-surgery or just to be admitted for an illness that invariably the waiting time is several hours. And so if you peel that back and say why does that happen? The nearest analog to think about is a hotel room. And guess what hotels do? They kick you out at noon and they don't let you check in till four, which means they have noon to four to get the room cleaned and ready for you. So what they have done, essentially through policy, is to make sure that departures happen before the arrivals. Now think what happens in a hospital. The arrivals start early morning. Surgery happens at seven in the morning, so people are looking for beds from eight in the morning. The emergency departments get clogged overnight and so when the hospitals open in the morning, there are all these patients in the ED waiting to end up getting admitted. So the arrival pattern starts early in the morning. Think about when you ever got discharged from a hospital it was invariably around dinner time. So the arrival pattern is before the departure pattern. That's great if you have plenty of excess beds, but now we're running out of capacity so we don't have excess beds. So every day the day starts out at the. They call the borders borders in the ED, meaning they're waiting for a bed. The PACU, the post-anesthesia care unit, has got patients waiting for a bed and every day it starts out behind the eight ball and, like Groundhog Day, again it repeats. So our angle on this is to algorithmically unlock the right discharges to allow the flow to improve. So how do you think about that? So take a 500 bed hospital which has an average length of stay of five days. So typically on a given day a hundred patients are getting discharged. Well, hospitals know they've got a hundred discharges to do today and therefore chase all hundred at the same level of rigor. What we are doing is saying, mathematically, let's predict which units are going to be short, and we predict with mathematical precision that six hours from now, unit two is going to be three beds short. So, of the 100 beds, if you chase those three with a little more energy and got them out the door, you would fix the pothole before you had to drive over it, and so unit three would not even notice that. Oh wow, you guys magically solved this problem for me, right. So that's the power of math, because, again, I think an analogy, so a simple analogy think about the difference of the LA freeways on bank holidays versus regular days. 

Now, not everybody stay at home. Only 10% of the people work at banks. But you know what? It's the right 10%? It's the kind that work from 8.30 to 5. If you got them off the freeway at 8.30 and at 5, life is better for everybody. So the challenge of mathematics is find the bankers. How can you find the right 10%? So, instead of focusing on all 100 discharges, how can I find the right 10%? So, instead of focusing on all 100 discharges, how can I find the right 10? If I unlock those and push them through quicker, it'll make life better for everybody else, and that's the power of our AI and our math. 

0:08:36 - David Williams
Got it Well? This is a question, but maybe it's also a product suggestion for you. I was familiarized with. One of the challenges of discharge that hospitals are having is that the post-acute capacity was jammed up, and then we were actually exposed to it because we were working on a project related to EMS and ambulances. Ambulances have trouble getting people to ride on the ambulance the same sort of story, but it's even worse, than the ambulances are sitting in the bays waiting to be able to take somebody into the emergency room, which is first of all taking their capacity, and also EMTs like to be on the road doing exciting things, not standing there. So how much do you have to solve the problem on both ends, or one of the other ends too, to make it really work? 

0:09:19 - Mohan Giridharadas
The ambulance problem is a real problem. We haven't actually immersed in it too much because there are private ambulances of services and the hospitals have their own sometimes, or a combination of the two. It is a real issue. We have actually not dug into it so I don't think I've got a particularly informed point of view on that. The post-acute care does matter a lot because invariably when we think about discharge what we are doing mathematically at the point of admission is thinking about two things and trying to get it mathematically right. When is this patient likely to get discharged? Two days from now, three days from now, four days from now? It's called expected date of discharge. 

Today most hospitals try and enter it manually. They get it wrong half the time. The field is missing half the time. We do cohort analysis and classification analysis, look at similar patients with a similar diagnostic, with a similar physician, et cetera, and predict when they're going to leave. And not only predict when they're going to leave, but how are they going to leave. Are they going to leave and go home, are they going to go to a long-term care unit, are they going to go to a skilled nursing facility, et cetera. 

And based on that you can start planning and preparing well in advance, as opposed to waiting for the last minute. So what happens with these skilled nursing facilities is they are capacity constrained and it's a hyper local problem because you can't send someone to a skilled nursing facility that's 100 miles away. It needs to be in the neighborhood and they tend to have a very tight capacity problem because their patients stay forever. Sometimes they stay weeks and months, and so if that gets jammed up, the bottleneck goes upstream. Now they can't get out of the hospital bed because there's no place to send them to, so it is a downstream problem. We haven't worked on it directly because we focus mostly on the hospital, and so this is a real issue, I agree. 

0:11:13 - David Williams
Sounds good. So last time we talked about math, I remember that we talked maybe a bit about machine learning and maybe we touched on AI a little bit, but AI has been a big topic of discussion since we last spoke, and there's AI and then, of course, generative AI is one of the reasons why there's been so much discussion, and when I was preparing for the podcast, I was first assuming well, you're probably not doing much in generative AI, but it seems that you are. So what's the role of AI broadly different types and how does generative AI fit into what you do? 

0:11:45 - Mohan Giridharadas
So if AI spans the gamut right anything from intelligent algorithms or using neural net or you know gamut right Anything from intelligent algorithms or using neural net or you know other machine learning algorithms to make better predictions, better optimizations we are using AI. So we've been using AI in our products long before. It was cool to be using AI, so it's been with us for a while Now. What generative AI is doing for us is the ability to we've started to launch it, by the way, so our products already got it. The inpatient product will have it in the next few weeks or months, and then the infusion product is third, so we are rolling it out to our existing customer base. 

The challenge of generative AI is to make sure it doesn't elucidate, and so if it generates recommendations that become nonsensical, then you essentially cast doubt on everything you have done. So we're being very careful about it and we're constraining it to just the knowledge from that hospital's data set, and so we are using it, arguably in a slightly more limited format, but a very useful format. We're using generative AI to be like the resident expert of our product for the customer. So imagine a customer who's just learning how to use the product cannot ask a really complicated question like Dr Smith is asking for more block time. Is it a good idea or not If I give him more block time? Should I give it to him on a Tuesday or a Wednesday? Well, our product can solve all of those, but you need to know how to navigate and slice and dice and drill, et cetera. 

With generative AI, we allow them to just verbally say Dr Smith's asking for more time, should I say yes, and from that we can interpret what the user wanted, structure the queries and tools so that it navigates our tool in the right way and surfaces up the right analytic along with a specific recommendation yes, you should give Dr Smith more time. We would suggest you give it to him on Thursdays, and the way you'll find time is to take it from Dr Jones, who is better served getting it on Wednesday and not Thursday. So that's kind of how we're using generative AI, and you can do it verbally, you can through voice or you can type it into a chat box. We call it AskIQ and it's IQ Autopilot, which is our generative AI solution. We see that happening more and more as we move forward. 

0:14:08 - David Williams
If you constrain the LLM, or generative AI, to looking at specifically your own data, your own product, does that 100% solve the hallucination problem? 

0:14:19 - Mohan Giridharadas
In our case it does because it doesn't make up anything. It's got no basis to make up anything right. Where it tends to make up is because you can't vet all the sources it's going to and so on. That's when you lose control of the plot. But we are tightening it. We can't afford to have it hallucinate, because we'll inadvertently cast aspersions on our analytic robustness if we allow that to happen. 

0:14:44 - David Williams
I spoke with another company that was doing something similar with hospital websites and was using AI and is just trained with sort of what's in the website, and my thought about where that goes eventually is that it's great because it makes it easy to navigate and you solve the hallucination problem. But if you think about people using that as their interface in the future, you might actually put other kinds of information into that website database. That's not literally on the website. I don't know if there's something similar with how the product your product works. 

0:15:13 - Mohan Giridharadas
You could. Right now we're constraining it to analytics. We've got internal use cases, for instance, where we're thinking about it as a amplifier of the customer success function. So each of our products obviously has got a customer success team. Customers will ask them hey, how do I do this, how do I do that? And we've obviously got thousands and thousands of internal documentation pages. Imagine feeding all of that into an LLM and now suddenly even a relative newbie on our team can punch above their weight because they could ask the LLM and get a very sophisticated response to that answer which maybe a more seasoned customer success person would have been able to solve. So we think of that as amplifying the capabilities of our team and but we're constraining those to internal use cases and not eventually, you know, putting them out in the hands of all of our customers. One could, if we push that further, you could imagine a self-service button where any customer can ask for any help they want and the LLM can answer it for them. Got it? 

0:16:22 - David Williams
All right. Now I preface this next question because you told me before that a guy with a dozen letters in his last name should not be trusted to name something. But I am interested in this term, autopilot. I hear a lot of these generative AI applications being referred to as copilot and I actually don't like that name. But why autopilot? 

0:16:43 - Mohan Giridharadas
Well, I think, if I think about autopilot on a Tesla, yes, it can do a lot of the driving on its own, but it's better you're there to make the final decision on it. So in some sense, we don't want to make the final decision, because there are lots of things, even when you think of it as a operationally simple decision, that we should be able to execute fully. You know, give David the 10 am slot. Well, there's clinical nuance on that. What if there was a clinical condition that warranted you coming in earlier or coming in later or running something else by? So we want to have our AI make the recommendation, but the final mile, the very last step, we want a human being in the health system to make it, because, one, it leads to a better outcome and, two, it leads to better buy-in, because nobody wants a black box making a decision for them, least of all in healthcare. 

0:17:37 - David Williams
I did turn on the full self-driving when it was offered for free back, I think, in April, and it's fun and it's impressive, but you don't necessarily want to leave the car to just do that. It is interesting. 

I'll tell you my concern about the word co-pilot, which I know is not the term you're using which is that I think about a co-pilot on a plane, and they may not have equal training to a pilot, but they are like a pilot and they could take over as a pilot and they become a pilot, and I don't like the idea that that's what the AI is doing. That doesn't completely make sense to me. The way you described autopilot is not a pure substitute for the pilot. It's something the pilot can use to extend their capabilities and so on, but I do like that term better now that you explained it. 

0:18:21 - Mohan Giridharadas
Right, by the way, I spend a lot of time thinking about airlines as well. And you're exactly right, the left seat is the captain and the right seat is the flight officer, and they can seamlessly swap one to the other and, essentially, you've got the controls, I've got the controls, and they're both driving it or flying the plane as equivalent, but just a shared understanding of who happens to be driving at any point in time. I don't think of what we're doing as equivalent. I think of it as making a very intelligent recommendation, but the final blessing is offered by the human being, and that's why autopilot resonates more for us. 

0:18:56 - David Williams
Now I saw that you are announcing some financial results and focusing on annual contract value, so I'm interested to hear first of all how it's gone and then also why that's a metric that you focus on. 

0:19:08 - Mohan Giridharadas
Yeah, we focus on the three metrics and we've hit them in recent times, over the last call it year, and we're thinking of it as a trifecta because we've hit all three, just because they happen to be nice numbers. So we hit the billion dollar valuation January of last year, which is obviously a meaningful valuation because it's the classic unicorn valuation. So that was January of last year, so it's been a while on that. We hit the $100 million in ARR annual recurring revenue also early last year. But then we hit the $150 million in annual contract value only in Q3 of this year. So basically, the quarter we are in right now and all three of those is a fairly significant indication of scale. 

It's very difficult to scale a company in healthcare. Lots of healthcare analytics companies just get off the ground. They get three or four pilot customers. They can prove it. But building something that's operationally sophisticated at scale is quite hard, and so we're very pleased with what we've done. 

To us, annual contract value is a leading indicator of annual recurring revenue because we want every number we put forward to be validated right. And so let's say, we sign for delivering ORs across 10 hospitals right the day they sign it. It's the annual contract value of all 10 hospitals, all ORs across the 10 hospitals. But they may start one hospital at a time, and so the annual recurring revenue is after we get the first hospital up and running that, and then the second and then the third. So the annual contract value hangs out there and then the annual recurring revenue catches up to it. It's a good leading indicator because it sort of shows backlog. We've got contracts that we have signed that we've not yet implemented. You obviously don't want the backlog to be too long, because that means you have to wait a long time to get there. But you also don't want the backlog to be zero, because then you don't have the next thing for your teams to do so. We like looking at both and we like having them six to nine months apart in time. 

0:21:19 - David Williams
That makes sense. Yeah, I'm jealous of you. I was at a company yesterday that I'm on the board of and they had a little toy unicorn and the horn was snapped off. I'm like what the heck? And I said, did you snap it? Or they claimed it just was in an accident or something, but it was certainly not a good omen as far as I'm concerned. So I'm glad that you've got more than the toy, but actually the valuation so congratulations on that. 

Let me ask you about cybersecurity, which is another topic that's come. It was already probably on people's minds when we last spoke, but more so since it's a big weakness of hospitals. You look across different sectors. Many sectors are affected, but let's say, if we compare with financial services, they're a lot more robust than healthcare is in terms of cybersecurity. How do you think about cybersecurity within the customer base and then maybe for your own business? 

0:22:08 - Mohan Giridharadas
It is incredibly important to us because PHI is just such a big red flag if there's ever a leak. So we go to extraordinary lengths because you can imagine health systems obviously jumpy about dealing with anyone. And so before we sign any contract, their IT teams give us 100 page questionnaires. And we've been through the vetting process literally hundreds and hundreds of times because we've got a thousand hospitals that use our products. So we've been through the vetting process literally hundreds and hundreds of times because we've got a thousand hospitals that use our products. So we've been through it a lot. 

What steps do we take? We do a lot. First off, we don't co-mingle data across health systems. So while it's one body of code that runs all our algorithms, every customer's data is uniquely separated. So we don't mingle data right so that one contains it. Two, we encrypt the data. At rest and in motion the data is encrypted, so that way, even if there were a leak, first of all it's contained to one and second of all it's encrypted. 

Third is we've got very tight rule-based permissioning. So I can't access half of our data because I don't have a need to know, and so we will deliberately just contain it very tightly. We have robust logs so we know who's touched anything, which then protects us from. You know, obviously you've got to even protect against ex-employees and so on. So we have tight access permissioning on all of that ex-employees and so on. So we have tight access permissioning on all of that. That's the full set. And then we minimize access. So, for instance, we don't allow access to our sites from outside the United States. So we, you know, just based on IP addresses and stuff, we protect it. So we've got layer upon layer of security. And then we've got an infrastructure and security team that has modern things, modern tools deployed that are constantly monitoring and making sure that all of our systems are secure. So this is one we pay a lot of attention to, invest a lot of time in, because if we had security issues then we won't be in business very long. 

0:24:15 - David Williams
Yeah, no, it makes good sense. I get a little chuckled because when the hospitals give you those huge questionnaires, if you turn it back on them, and even more so if you actually use technical means to assess the answers, you would find something that's not as pretty as you might like or as you might need to be at standard. You're held dear. As pretty as you might like or as you might need to be at standard, you're held dear. 

0:24:33 - Mohan Giridharadas
We're lucky because we served in the cloud and we deploy cloud-based services, and so our ability to keep things more secure is also very high. 

0:24:44 - David Williams
Let's talk to the last question, which is a repeat question but I think is a fair question, which is about if you've had any chance to do any reading lately, anything that you like that you might recommend to our listeners, and also if you've been reading something that's terrible and you recommend that we avoid it. 

0:25:01 - Mohan Giridharadas
I'd like to know about that, as well, I got put onto this book called Extreme Ownership. It's a Navy SEALs team, it's Jocko Willink and apparently he's got a whole following with podcasts and all kinds of things which I didn't know about. But I just bought the book Extreme Ownership and I've started reading it. I'm about a third of the way through it. But then just last week I saw a new book on leadership. It's called the Journey of Leadership. The two of my McKinsey colleagues Hans Werner Kass from the Detroit office and Ramesh Srinivasan from the India office have written, along with a couple other co-authors, wrote a book on the journey of leadership and I think this is based on interviews and learnings from hundreds of Fortune 100 CEOs. I had to pre-order it and Amazon just delivered it to me yesterday, so I just unwrapped it. I haven't even cracked the book open, but since I know those two guys, I'm excited about getting it. 

0:25:58 - David Williams
I was going to say I would expect a signed copy would have been sent your direction. 

0:26:02 - Mohan Giridharadas
Yeah, no, no, I bought it on my own and the next time I see those guys I'm going to make them give me a signed copy, but to pony up on Amazon for myself. 

0:26:09 - David Williams
Excellent. Well, mohan Giridharadas, ceo and founder of LeanTaaS, thanks for joining me today on the Health Biz Podcast. You've been listening to the Health Biz Podcast with me, david Williams, president of Health Business Group. I conduct in-depth interviews with leaders in healthcare, business and policy. If you like what you hear, go ahead and subscribe on your favorite service. While you're at it, go ahead and subscribe on your second and third favorite. Thank you,