Ritu: Hi Sameer. Welcome to our podcast, The Big Unlock, season six. My name is Tum Roy, and I’m the Managing Partner at Damo and BigRio, based out of Gurugram, India. Always a pleasure to chat with CMIOs, CIOs, CDOs, and Chief Information and AI Officers about all the latest developments in healthcare systems. I hope this will be a really exciting and engaging conversation. Welcome to the podcast once again.

Rohit: Thank you, Sameer. Welcome to the podcast. I’m Rohit Mahajan, Managing Partner and CEO at BigRio. Over to you.

Sameer: Thanks for having me, folks. Excited about this conversation.

I’m Sameer Sethi. I function as the AI and Insights Officer for Hackensack Meridian Health. We are a health system based out of New Jersey, a combination of 18 hospitals and close to 500 care sites. We do this with the help of around 40,000 team members, both clinicians and non-clinicians. Our care is mostly across New Jersey, and we also have a very big digital presence.

My focus within the organization is primarily on four areas. One is data and analytics. This is where we produce and ingest data from various source systems, with our electronic medical record being the biggest one. My team helps normalize that data and make it available not just as raw data, but also as insights.

The second is AI. This involves moving beyond explaining what has happened to understanding why it happened and predicting what may happen next. Those capabilities are now more easily available to us through third-party models. Building capabilities that leverage large language models and generative AI is a big piece of our portfolio as well.

The third is robotic process automation. This is where we write code to emulate human functions where appropriate. We have a large program that focuses on automating mundane tasks or tasks that require accuracy that machines can deliver better, or speed that humans cannot match. We now have over 200 automations in production today—a really large portfolio we’ve been building for quite some time.

And the fourth is software development. At times, when purchasing software or capabilities is not viable for Hackensack—either because it’s not what we need, it’s too expensive, or it’s not available—we will write software and deliver it to our clinicians and operators.

So that’s what I do for Hackensack. I’ve been here for three and a half years, and it’s been a really rewarding experience.

Ritu: Thank you for that introduction, Sameer. Lots to talk about. We were very interested to hear how you think AI is not just for the sake of AI, but about putting it to the right use. We would love to hear about some applications at Hackensack that you’ve built where you really feel AI has demonstrated transformational change and made the patient experience better.

Sameer: Absolutely. We’ve been excited about this. We’ve been at it for almost three years now, with a very heavy focus on AI in the last two years. Three years ago, the conversation on AI was more about machine learning. Now it’s more about generative AI, and the talk track behind why this is so important has changed as well.

At Hackensack, because we are so serious about this, we’ve really thought about how to put AI to use. This goes beyond just talking about how AI is good and the governance around it. While we do that, we are also focused on building the right solutions. There’s one thing to look at tools like ChatGPT and say you can ask a question and get an answer back. But the real value is in embedding these technologies into workflows. That’s what we are most focused on.

Giving someone access to a conversational AI—whether by voice or typing—gets you to a certain point. But the real impact is in customizing it for a clinician at the time of care. That’s where Hackensack has put its efforts.

A good example: as a hospital system, our clinicians see patients, and when they do, they face a sea of valuable data. Reviewing that data takes time—time they often don’t have, or time taken away from the patient. One of our first use cases when large language models became available was to use generative AI to summarize that data. We essentially wanted to improve the experience.

We’re all consumers of healthcare. I’m sure you’ve been in situations where you go to a doctor’s office and the doctor is staring at the screen while trying to listen to you. What that clinician is doing is trying to understand your clinical record, which takes time. We made that easier. We provided a button for clinicians to summarize a patient’s information instantly—in seconds.

That use case has evolved. Initially, the summaries were general. But through feedback we realized summaries had to be by specialty. An oncologist, for example, needs different details compared to a primary care physician. So, we started with eight specialties and today it’s almost 15, giving physicians the option to summarize by specialty. That has been very valuable.

That’s one way we’ve used generative AI technology and put it to good use. This continues to grow. For example, we’ve also applied it to patient survey data. We collect survey information from patients and families post-discharge. The structured data was already in dashboards, but we realized patients were typing a lot of comments.

So, we created sentiment analysis on top of that unstructured data and gave it back to the business at a unit level. This wasn’t happening before. Operators can only read so many comments. By synthesizing comments in volume, we can surface trends across unstructured data. Again, we used large language models to deliver those insights back to the business.

This is an important part of the conversation. When we talk about AI and how to put it to use, there’s the aspect of finding the right use cases, but there’s also a technical side. There’s a lot of discussion around tools like ChatGPT not giving the right answers. At Hackensack, we put together a center of excellence that focuses on indexing. Why? Because indexing makes the answers better.

So, at Hackensack we’ve not only been exercising the muscle of finding the right use cases and delivering them, but also learning how to harness the technology and use it better. This technology is here to stay. It’s really good—you just have to know how to put it to the right use.

Ritu: Thank you, Sameer, for sharing those use cases. It’s really incredible that you were able to do clinical note summarization, especially across so many specialties, and provide customized information to physicians. Did you feel there was any pushback? How did you handle the literacy or acceptance part of it? Because some people we talk to feel innovation is difficult for healthcare professionals to accept, and it takes much longer. I’m sure they saw the benefits, but it might have been a steep curve. What was your experience?

Sameer: At Hackensack, this has been a bit of us learning from our mistakes over time. How you build these things is really important. And when I say “how,” I mean the non-technical side.

What we did well was first understanding the problem that needed to be solved, and then partnering with the clinicians who wanted to solve it. When there’s mutual participation, there’s better acceptance. We also set clear expectations from the start—what this technology is, what it isn’t, how accurate it is, and how accurate it isn’t.

For example, with clinical note summarization, our intent was never to replace anyone’s work. Our intent is to help people work at the top of their license. So, we tell clinicians: this isn’t 100% accurate. You still have the whole patient record if you want to review it. If something in the summary doesn’t look right, it’s your obligation to check the full record. But what we’ve given you is a strong head start. That’s what reduces pajama time. That level of transparency has helped us get much better acceptance.

And by the way, my team and I welcome pushback. It teaches us a lot. As technologists, we’re keen to build solutions, but they have to be used by clinicians. When we get pushback, we consider it feedback. Maybe we’re not explaining the solution well, maybe we haven’t designed the right user interface, or maybe we haven’t embedded the capability properly into the workflow. We look at the positives in pushback, and it has worked to our benefit.

Ritu: That reminds me—many years ago I attended a lecture by Marshall Goldsmith where he said, “I don’t know why people call it feedback. It should be called feed-forward.” Because anytime you get feedback, it’s actually helping you move forward. I really love that line and share it with so many people.

Sameer: We actually make a very strong attempt to build feedback—or feed-forward—mechanisms into almost everything we deliver. Not capturing that feedback over time results in lack of adoption.

Ritu: Yeah, because you start drifting away from what people actually need.

Sameer: Exactly. Even a basic thumbs up or thumbs down is valuable. A thumbs down signals us to go back to the user and ask, “What happened here?” We try not to complicate feedback collection, but we always make sure to build feedback loops into our solutions.

Ritu: So, you’re saying one of the keystones of AI adoption at Hackensack is bringing everybody along for the ride and ensuring buy-in from all stakeholders.

Sameer: You know, I also want to add something valuable to this conversation—we’ve thought a lot about governance. I think all of us have, but it’s still such a word out there and people don’t always know what it really means. So I’ll tell you our journey.

We call it our governance pyramid. At the bottom is the technical tier, which is our team evaluating and building AI. On top of that, and more important in this context, is what we call our AI and Automation Governance Group. This is made up of 13 domains across the health system coming together. When we built our governance process, we invited representation from every domain—legal, HR, clinical documentation, finance, and others—so that every voice had a seat at the table. That inclusion really helped us.

Then we have the next level up, which is our governance committee, primarily the executive leadership team. These are the folks that help us prioritize, because everything can look good, but we can’t do everything. So the committee decides which problems to solve first and which areas deserve focus. That model has been very helpful in keeping us on track.

Our board has also played a very large role, and sitting on top of all this we had our annual summit last year. As you can imagine, when we started doing AI—and especially when generative AI came in—there was a rush of demands from the business. Everyone was saying, “I hear this is available, do this,” or “This looks really cool, let’s try it.” At a certain point, I honestly didn’t know what direction to look in.

So we went to the board and established six areas of focus for Hackensack. The reason we did this was because we knew if we didn’t, we’d end up chasing shiny objects. With the help of the board and our executive leadership team, we identified six strategic areas, and the principle was simple: if a use case didn’t fall into these areas, we wouldn’t do it.

The six areas are creating personalized and equitable patient experiences, streamlining administrative and clinical efficiencies, capacity management and reducing burnout, disease detection, precision treatment by customizing therapies for patients, and research and innovation. These are broad enough that most initiatives can align, but focused enough to give us discipline.

Every use case that comes to the governance committee now goes through a scoring process, which looks at ROI and how many of these buckets it fits into. The more it aligns with our focus areas, the higher its score, and that’s what determines where we put our efforts. That process has been very valuable for us and has helped us stay focused on the right problems.

Ritu: So, most of the CIOs we talk to say that they’re concentrating more on the digital front door and the patient experience. They want to stay away from disease detection and anything that requires like FDA approval. I’m curious, when you talk about disease detection, what are some of the use cases there and how do you get around that?

Because. Exactly the problems that you mentioned earlier with LLMs still having hallucinations and other things, and a human in the loop having to review all the results. So, for disease detection, how do solve that.

Sameer: That’s a great question, and I think you’re asking it with generative AI in mind, which makes sense because that’s where most of the buzz is right now. But the reality is that disease detection is much more a machine learning use case than a GenAI one. With machine learning, you don’t really have hallucinations, or they’re almost nonexistent. And that’s an important distinction. Too often people hear “AI” today and immediately think only of GenAI, when in fact, there’s so much more to it.

What’s interesting is that when we had our last board summit, we actually ran a voting exercise across the six areas of focus we had defined. And surprisingly, disease detection came out on top. Right after that was streamlining administrative efficiencies and addressing burnout—clinicians, as you can imagine, pushed for burnout relief, and operators pushed for efficiency. But overall, the board said disease detection should be front and center.

And honestly, it makes sense. We’re a hospital system. If we can detect disease earlier, we can keep patients healthier. That’s the whole point of what we do.

Let me share a couple of examples. The first is mortality detection. This one is personal for me. A few years ago, I lost my mother-in-law, and I’ll never forget the palliative care team telling us, “We should have been here six months ago.” I remember sitting with the physician and asking, “What happened? Why didn’t anyone flag this earlier?” And the honest answer was, they’re not wired that way. Physicians and nurses are trained to keep fighting, to keep trying to fix the patient. Letting go and saying, “It may be time for palliative care,” doesn’t come naturally.

So what we did was build a model that produces a mortality score. And then we embedded it directly into the clinical workflow. Here’s how it works: a clinician might go into the patient record to prescribe something simple, like Tylenol. At that moment, a best practice alert pops up and says, “The model predicts a high likelihood of mortality within six months. Consider palliative care.” The clinician can review the labs, the patient history, and if they agree, they click “yes.” That automatically creates an order for the palliative care team to engage. If they don’t agree, they hit “no” and provide feedback, which we use to improve the model.

So what we’ve created is a chain: a mortality score that becomes a nudge, that then becomes an order—all with the clinician still in the driver’s seat. And the result is that patients are now being referred to palliative care much sooner. They likely would have ended up there eventually, but now families are getting the support earlier, which makes a huge difference.

Another example is chronic kidney disease. We’ve built capabilities that allow us to detect CKD much earlier in the progression of the disease, and that early identification can dramatically improve patient outcomes. And right now, we’re also working on chronic asthma detection.

So those are just a few of the use cases, but the bigger point is this: disease detection is core to us as a health system. It’s not about replacing clinical judgment. It’s about surfacing signals—signals that help our clinicians make the right calls, earlier, and give patients better care.

Ritu: That’s really interesting because we haven’t heard that from other hospitals, so, that’s really incredible to hear.

Sameer: You know, one thing I always want to emphasize is that what we’re building are not decision tools—they’re decision support tools. There’s a big difference. The final call always rests with the clinician. Take the mortality model I mentioned earlier: yes, it provides a nudge, but alongside the “yes” button, there’s always a “no” button.

Ritu: You’re alerting them, but the decision is still theirs.

Sameer: Exactly. And when they press “no,” they’re asked to provide a reason. That feedback is incredibly valuable to us. It helps us refine the model, learn where it might not align with clinical judgment, and continue improving it. But at the end of the day, the conversation with the patient, the clinical decisions—that all stays firmly in the hands of the physician. What we’re doing is simply surfacing insights at the right moment, in a way that makes patient care better.

Ritu: I think this has been a really insightful part of the conversation.

Thank you, Sameer, for sharing those use cases for disease detection. Looking forward, you know, if you looked at the next six months or a year or maybe two years, what do you think the top three trends in AI would be? You know, I know we talked a little bit about agents in our prep call yesterday, and uh, you had like a very good story about agents as well, where you said that agents are, you know, more, mostly an orchestration platform.

So, if you would like to talk a little bit about that, share what you think the next three big trends are going to be. Or not trends, but like things which are going to, you know, come to the front line and be a focus area of focus for Hackensack.

Sameer: I’ll tell you my top three are going to be agents, agents, and agents. Agents. This is an opportunity for us to use technology that is here now to orchestrate. For me, when I think about agents, it’s a concept. The concept is: how do you orchestrate different technologies to work together and provide an output that a human can interact with better? That’s an agent.

In the past, folks like me have tried to build these ourselves, similar to how we built machine learning models. Today, those models are available—now they are machine learning models versus large language models, but just using that as a concept. What I’m trying to portray is that we’ve been trying to build these agents for a long while, but the tech had to be developed from scratch. Orchestrating RPA software and a database to all work together in a certain sequence wasn’t easy to do. We did it, and today, technology companies like UiPath, Google, and others have built platforms to orchestrate that.

Now the orchestration has become a lot easier. Why is orchestration important, especially in healthcare? Because we are now not just relying on humans to do certain things. In the past, we built AI and gave it to a human. Rightfully so, in some situations, a human has to interact with it, like the mortality example. But there are instances where we’re building AI signals that do not require a human. The next step can still be taken by a machine.

In the past, when building these signals, we would rely on a human to interact with it and do something, even when the human wasn’t needed.

That doesn’t hold true for all use cases and applications. In certain places you still need humans, but that’s what I’m most excited about. In places where I don’t need a human, how do I build and take it to the next step? Because then those insights will be used, rather than relying on a human who may not have time, or the program lacks buy-in. What I’m most excited about is how do I take those instances and use all these insights more effectively. Now you can move to the next step without relying on a human. I think that is the evolution of AI, and that’s why I choose this as my top three focus areas.

Now look at the portfolio I’m privileged to own: RPA and AI and software. Now the world has given me a platform where I can plug all this together. I’ll use an example: as a health system, we receive a lot of denials from insurance companies.

Rohit: Mm-hmm.

Sameer: We rely on humans to take that denial, synthesize it, get information, then generate a new appeals letter and send it. That’s a process—requires expertise, humans, and time. The impact is that getting this together delays the cash flow until the insurance pays. What we’re building is an agent that will do most of that work: receiving a denial letter, extracting information about the denial (maybe a missing modifier or pre-authorization), then using a large language model to synthesize that information.

Then it taps into a rule-based engine we built: if the denial reason and modifier meet a certain rule, do this. I’m oversimplifying, but it relies on the rule-based engine to identify the opportunity and the deficiency in the previous claim or submission. Then it creates a new claim or appeal, using generative AI. RPA pushes it to a human who reviews, approves, and then it’s processed again with the insurance company. This was all done by humans previously. Now, orchestration platforms can switch between a large language model, a rule-based engine, generative AI, and RPA—moving all these pieces together. That saves time and a lot of money. Humans are still involved toward the end, reviewing, but the things we didn’t need humans for are now done by machines.

Ritu: Amazing. Great to hear that, Sameer. Rohit, do you have any last questions for Sameer?

Rohit: Y Sameer has given a lot of food for thought. I’ll be taking exhaustive notes. I have enough questions for another complete podcast, but what I’d like to ask you—and I’m very interested in the center of excellence on the data indexing part, so I’ll come back to you on that as well—but could you share with listeners what motivated you to get into healthcare, how you started, and your story? That would be great to end the podcast with.

Sameer: That’s a very good question. I love sharing this story. This goes back almost 15, 16 years. I worked in financial services. I married Simon, she’s an occupational therapist. She works as hard as I do, if not more. I still remember her coming home after we got married in 2006. I was at Deloitte, not in healthcare then, and she said, “I see you on the computer all day. You do great work and provide for the family, but I want to share some feedback: I made someone walk today. What did you do?” That triggered something in me. I wasn’t much of a charitable person, but I thought—maybe I can do something that saves a life or saves money and care. I looked into different things and eventually, I landed in analytics.

That’s what motivated me—I wanted a job and career where I could make an impact without going back to school to become a doctor or clinician. This is my calling, where my team’s hard work hopefully helps someone live better, or maybe, as in the example, die better. For me, it’s a very passionate thing. I tell people I came into healthcare and am never leaving, just for those reasons. Whenever I think about an opportunity beyond healthcare, I hear my wife saying, “I made someone walk—what did you do?”

Rohit: That’s an amazing story, Sameer. There is a saying at Harvard Business School about the benefit of a physical footprint, but what about an intellectual footprint that can span the world? You have so many great ideas that can be propagated. I’ll definitely be back in touch with you on that. Thank you so much.

Sameer: What’s most exciting for folks like me is that, first of all, the data is there. There’s so much data being generated now—from our watches, phones, doctors—it’s all available. Second, we’re not alone anymore. I remember the days when the only model you had was one you built yourself. Now, Google and OpenAI—everyone has a model. There’s so much opportunity between available data and capabilities. The third thing is willingness—consumer willingness. I have a relative I help with doctor visits. Pre-COVID, she always wanted to visit her doctor in person. Post-COVID, she asks, “Can I take a telemedicine appointment?” It’s a big change.

Consumers are open to digital means of delivery now. Technology will do well, though healthcare will always be a very people-centric business. But I’m excited—the data is there, capabilities are there, and consumers are ready.

————

Subscribe to our podcast series at www.thebigunlock.com and write us at [email protected]  

Disclaimer: This Q&A has been derived from the podcast transcript and has been edited for readability and clarity.

Q: Hi Regina. Welcome to The Big Unlock Podcast. I am Rohit Mahajan, Managing Partner and CEO at BigRio and Damo Consulting. It’s so nice to have you on the podcast, especially since you are a practicing physician. Would you like to introduce yourself to the audience? 

Regina: Well, Rohit, thank you so much for having me, especially since I am a practicing physician. I’m not a technical founder or someone in the data world full-time.

For you listeners, I am Dr. Regina Druz. I’m a board-certified cardiologist since 2001. I’ve gone through all the traditional cardiology training and worked in hospital settings, academic settings, and private practice. Along the way, I went down the entrepreneurial route, which led me to personalized precision medicine, longevity medicine, and rethinking what cardiology should be. One of my colleagues calls it “comprehensive cardiology.” It’s about redefining what I can bring to patients and how they can benefit beyond traditional, expected care.

Q: That’s very cool, Regina, to know that you are now on the path of your entrepreneurial journey as well. So tell us, what attracted you to healthcare in the first place? How did it all start?

Regina: Good question. My parents are not physicians, but I grew up with amazing role models. My aunt and her daughter, my cousin, were the original power women, though the term didn’t exist then. Both were physicians—obstetrician-gynecologists—and were excellent at what they did. My aunt, after her education, was asked to go to an underdeveloped country to help build healthcare and improve maternal mortality. Many women at that time were dying in childbirth, and she brought in new practices that made a huge impact.

They were inspiring role models. Since I was young, I said, “I’m going to be like them. I’ll be a doctor, I’ll be strong, I’ll cure the world.” My father, an electronics engineer, pushed me toward technical fields, and I enjoyed science. But biology really grabbed my attention. The rest was conventional: medical school, residency, fellowship.

I didn’t end up a gynecologist, which disappointed my family a bit. They said, “Oh my goodness, you’re choosing to be a cardiologist?” But I said, “Women can do cardiology too.” Luckily, more women have entered cardiology over the past 20–30 years, and specialties once dominated by men are now more welcoming. Women are making their mark. That’s me in a nutshell.

Q: That’s a very cool journey, Regina. So you talked about something in your introduction which caught my attention around personalized medicine and longevity. So please tell us like what does it all mean and how’s it coming together?

Regina: That’s a great question, Rohit. Maybe I’ll take you back to some of my more traditional cardiology days. When I was working in a hospital setting, I was responsible for an area involved with cardiac imaging. We have various procedures to image the heart, understand how it functions, if the blood vessels are open, if the valves are doing okay, and if we can foresee the future and understand what’s going on. My particular area was nuclear stress tests. Patients typically had to exercise on a treadmill as part of a stress test. For those who couldn’t exercise, we used a chemical stress test, but the majority exercised on a treadmill. We would inject a little bit of radioactive material to see how blood flows into their heart and determine if their heart is healthy.

What came out of it is that as doctors, we were a bit biased. We would expect younger people to do better, and that older people would not do as well. But 20–30% of the time, we would get the opposite—an older person who did much better than expected, and a younger person who did very poorly. This started my journey into personalized medicine. I started to ask if there was a pattern—what are the older people doing to preserve this ability into older age, and what are the younger people doing, or not doing, that actually takes away from what we think they should be able to do, cardiac-wise and physically?

I started asking patients questions that weren’t part of our standard intake. Do you exercise at home? Do you garden? Do you travel? What do you like to do? What gives you purpose? I looked at other physicians who introduced this idea of “n of one,” of personalized medicine. There are reasons, systems biology, root causes driving health and longevity, and also driving disease. It’s not standardized; it’s not a population-based metric. It’s literally the N of one.

That piqued my curiosity, so I decided that the best way to address it was to take what I know in cardiology and bring it into a personalized medicine approach. The best fit for it was integrative and functional medicine, where we could expand our lens and look beyond just hard numbers—like cholesterol, blood pressure, and medications—and ask why levels are elevated, and what we can do lifestyle-wise to optimize health. Optimized health is a major contributor to longevity. Physical health, cardiovascular health, brain health, and systemic health are really the foundation of any longevity equation. You can’t talk about longevity if health isn’t optimal.

That was the beginning of going off the beaten path. My colleagues thought I was crazy because they could not understand it. They said, you have great credentials and roles. At that time, about ten years ago, there was no room in the institutional setting for this. There could have been some departments inclined toward integrative or complementary medicine, but nothing institutional. That meant going on my own, right as medicine was beginning to consolidate. People said I was insane and would do very poorly. I opened my first tiny practice with about three patients in an area saturated with cardiologists. Even back then. Today, it’s exponentially so. I’m here to tell you this works, and this was not only a huge benefit to my patients, but it was a benefit to me.

Q: That’s awesome, Regina. And I know you are very interested in all the new technologies coming our way. In any podcast these days, we have to touch upon AI. How do you look at AI and digital health from the lens of cardiology? What does it mean to you, and how do you see some use cases from a physician, patient, and caregiver perspective? How does it all play out in your world?

Regina: It’s a great question, Rohit. I was an early adopter, so back then there weren’t as many things available to us. Right now we have so much more. In any digital health domains, in cardiology or any specialty, you have to ask how they benefit a patient, and how they help doctors, patients, and nurses to do the right thing without burning out, saving on cost, and staying compliant with regulations.

I started with telemedicine. I was an early adopter, and since my practice was personalized, telemedicine is a great tool. It gives patients a lot of access, but it has a major deficiency: you’re not getting a physical exam. You can’t listen to a patient or do those things. I needed their blood pressure, electrocardiogram, and a couple of other measures to assess if they are okay. This led me to explore digital tools available—there are quite a few. For example, a company called AliveCor created a device you could clip onto an iPhone, and by putting two fingers on metallic plates, it would record a single-lead electrocardiogram. It wasn’t a full EKG, but it was the beginning.

Other companies like Omron and Withings came out with Bluetooth-enabled blood pressure cuffs, letting patients check their blood pressure and transmit the information to me. For the first time, with telemedicine adoption, I wasn’t blind anymore. When COVID came and in-person visits were canceled, I wasn’t stuck. I already had the telemedicine platform and some of these tools.

Now it’s even more diverse. I routinely use the tools from these companies. Now we can do a Bluetooth-enabled 12-lead electrocardiogram with AliveCor—no traditional machine needed, just a small remote controller with a button. Cardiac and vascular ultrasounds now have devices like Butterfly IQ, which has ultrasound on a chip that connects to a phone or iPad, providing diagnostic images affordably.

Digital devices first went to practitioners and are now available to patients. Patients use Apple Watch, Withings, Fitbits, and other trackers. Suddenly we have overwhelming streams of data. Now, AI has come into the picture. Patients are becoming “citizen patients”—they investigate themselves, upload their data into large language models, and get outputs. They have agency to do this, even if they can’t always tell whether the output is correct.

These data streams are becoming routine in medical practice. Places like Mayo Clinic use large data repositories and AI tools to analyze electrocardiograms and predict heart failure years before symptoms appear. This is a huge opportunity. We could work with high-risk individuals preventatively, sparing them clinical heart failure. Once someone has their first heart failure episode, five-year mortality is 50%. Imagine if we could be ahead of this curve.

We are evolving from curiosity to clinical implementation and, hopefully, to clinical competency—making this standard care, not just a nice-to-have. Regulatory bodies recognize the value of these devices. Medicare, for example, pays for remote care monitoring (RCM), usually reserved for higher-risk patients and accomplished through digital devices and AI interfaces. They allow us to identify what’s happening—a sort of third eye.

Q: That’s pretty deep usage of AI and digital devices in different ways, Regina. One other thing—I know you mentioned large language models. Another thing that’s become very popular is AI coding tools. We were just chatting before the podcast, and I have several physician friends experimenting with such tools. So tell us your experience. You mentioned an article too; tell us more about what you’re seeing in this space. 

Regina: This is very interesting because when ChatGPT became a household name, I said, okay, let me give it a try. At the time, I was studying for my boards. Cardiology boards have different ways to maintain certification. One newer way is that every year you do a small cardiology knowledge exam. It was time for my exam, and I was studying using traditional stuff from the American College of Cardiology, but some explanations to questions left me wanting more. So I started putting those questions into ChatGPT, and it gave me deep, nice, long explanations that covered some of my blind spots. That was interesting. I think large language models, when they don’t hallucinate, are great educational tools. You can constrict them to look at specific journals or guidelines.

But the real breakthrough for me as a practicing clinician came from ambient AI scribes. They became indispensable in my practice. There are several companies; I personally use one called Heidi Health, a startup from Australia. Most people think ambient AI scribes just catch conversations and save time on notes, which is true, but there was something else—a positive externality. The structure of the AI scribe allowed me to generate versions of notes, including one I could send as a message to patients via patient portals. This summarized next steps clearly—what they need to do and how. Instead of reinventing this, after some tinkering, I generated a template I liked. This enhanced my patient care meaningfully by adding another touchpoint, another opportunity for engagement. It also helped me not have to remember everything but to clearly outline the next step.

In personalized medicine and longevity health optimization, this requires a lot of participation. It’s very different from traditional medicine where doctors just tell patients what to do and wait for them to come back. Here, it’s the opposite; you need to reach out regularly. The AI scribes gave me that ability to be efficient and effective and to “excite and delight” patients with better communication.

As a clinician entrepreneur, I am also experimenting with Agentic AI. Some experiences are helpful; some are questionable. One unsettling issue, raised by AI experts like the CEO of Tropic, is that AI is a black box. We lack full visibility into how it generates answers, especially with agent AI. I asked it for a simulation on a complex patient with cardiac and immune issues, including all parameters we know. The AI produced a great output with references and did not hallucinate. My initial hunch was correct, but I wanted to know how it built the case. In medicine, we use frameworks for diagnosis, workup, and prognosis—check and balances learned from real outcomes. That’s where AI’s black box presents challenges.

Whether it’s LLMs or chatbots or more complex agent functions, we are in uncharted waters. We will use them going forward, but we need more visibility into their processes for trust and safety.

Q: Of course. Regina, I think we hit on something during our pre-podcast discussion about the digital divide. You said people with high agency will propel ahead using such tools. Could you explain your thought process behind that? 

Regina: Great question, Rahi. The digital divide often comes up with digital tools like wearables or telemedicine platforms. We have to ask whether certain patient groups will face barriers. Older patients may not be technologically savvy enough to navigate telemedicine or connect Bluetooth-enabled devices like blood pressure cuffs. They might struggle with mobile apps or synchronizing devices. Plus, healthcare privacy is heavily regulated.

Traditionally, the digital divide affects older patients, those with cognitive issues, non-English speakers, and those who can’t afford or access these tools. But I believe another digital divide is forming. Recently, I read a Wall Street Journal piece about AI tools that almost anyone can access at a free level. To truly benefit, users must be “high-agency” individuals—able to instruct the tool, understand its limitations, refine outputs, and keep moving forward.

I see this with my patients. Those with high agency get more information, more touchpoints, and a very different experience managing their health compared to those less inclined to use these tools. Physicians sometimes reinforce this divide. One colleague, Dr. Kim Williams, former ACC president and lifestyle medicine champion, uses ChatGPT as part of his team. He inputs patient scenarios—not for medical advice but for cultural context—to help patients transition to healthier diets without overwhelming them.

Q: That’s awesome. As we close, Regina, looking into your crystal ball over the next 3–5 years, what exciting things do you see coming our way? 

Regina: It’s an interesting one because I ask it to myself very often. In the physician community, there are a lot of voices who say AI is going to replace doctors. And then, from the tech world, there are predictions like no more radiologists.

Of course, there are other voices that are more introspective and say no, AI is not going to replace doctors. To me, it’s very clear that we are going to see in the next five years a rapid emergence of agentic AI in all verticals of healthcare. It’s going to take care of mundane tasks like pre-authorizations, appointment scheduling, and supply chain management.

Then it’s going to really make its mark in staffing decision making, especially when calibration is needed between demand and supply. And it’s already doing it in clinical decision support. For example, there is a company called Aidoc that allows radiologists to identify life-threatening pathology on brain scans that potentially they may have missed. There are AI-enabled mammography tools that can predict occurrence of breast cancer before it even happens.

There is also an investigational AI tool that allows clinicians to predict the onset of Alzheimer’s dementia before any cognitive decline. As a personalized medicine and longevity doctor, this is where I see the biggest impact. Transitioning into clinical decision support, I predict it will make the “N of one” personalized medicine a standard type of medicine. Eventually, we will abandon our focus on large populations because AI will enable us to address population health differently. We’ll be able to leverage small data, that N of one, to build much more enriched phenotypes. Populations as we know them for investigations and healthcare will be redefined through AI, and that will be a huge impact.

Q: That’s awesome. Thank you, Regina. This has been a very interesting conversation. Really appreciate it. Any other last comments? 

Regina: Maybe just a quick note to anyone listening who might be a physician, nurse, nurse practitioner, or anyone in healthcare with a medical affiliation. If you’re thinking, should I be doing it, should I not, should I jump into this entrepreneurial route and see where it takes me? There are no guarantees, but if you are one of those people, the best time was 30 years ago. The second-best time is now.

————-

Subscribe to our podcast series at www.thebigunlock.com and write us at [email protected]  

Disclaimer: This Q&A has been derived from the podcast transcript and has been edited for readability and clarity.

Q: Hi Inder, it’s great to have you on the podcast. I am Rohit Mahajan. I’m the managing partner and CEO at BigRio and Damo Consulting. This Big Unlock podcast was actually started by Paddy Padmanabhan, who founded and started Damo Consulting. I’m very happy to carry on his legacy. He’s pretty well known in the healthcare circles. I’m sure you’ve probably met him at one of the conferences as well.

So, super excited to have you here Inder, and over to you for your introduction.

Inderpal: Thank you, Rohit. And yes, you bring up Paddy, and who doesn’t know him and what a great individual he was. I’m glad I had the opportunity to work with him. Thank you for this opportunity. I’ve been in healthcare technology space for over two decades now, consistently working on innovation and digital transformation across complex organizations. I have very good experience working with some of the best organizations in the area. I landed in healthcare at Columbia University Medical Center, had a long stint at HSS in New York, and then Englewood Health in New Jersey. I’ve gained experience by rising up the ladder over these years and have a good grasp of the entire continuum of healthcare from a provider and patient delivery perspective. I’m really excited to talk about what my experiences have been and what your viewers are interested in. 

Q: Absolutely. Inder, like you said, you got started in healthcare over 20 years ago, and you worked with some of the finest institutions. You are well regarded as a healthcare leader in the space. Tell us more: what attracted you or how did you gravitate into healthcare? 

Inderpal: The first few years of my career, I worked in India, Southeast Asia, and the Middle East, mostly around banking and financial sector. Even in the US, I had a stint working with CitiMortgage for almost two years, and then the project ended and we were looking for a new project. I landed a new project at Columbia University Medical Center. I did not know at that time how lucky of a break that was because I landed at the biomedical informatics department, a pinnacle of informatics research back then and even today. That’s where I first entered healthcare by accident, but then decided to stay in. The clinical research part, clinical trials management—all that was new to me but interesting. We also developed solutions which we deployed at New York Presbyterian Hospital for patient care. That was my first exposure to what bench to bedside looks like, and that really solidified my interest to stay in healthcare.

I was part of the Epic transformation journey for many years, as HSS grew and made collaborations with other hospitals like Stanford in Connecticut, Florida Tenant Hospital, and New York Presbyterian. I was part of clinical integration and external integration workflows. It was a great stint at HSS, and then an opportunity at Englewood to manage all aspects of IT as CIO. This broadened my horizons further; application was my home base, but I also took over cybersecurity, infrastructure, and networking, led modernization efforts around infrastructure, storage, servers, data centers, disaster recovery, and built the cybersecurity program. This gave me a holistic experience, and that has been my journey—23, 24 years in strictly healthcare organizations.

Q: That’s awesome. So could you please share with us in the aspects of this journey with us in terms of what are the digital health programs that you kind of like, were very close with and you, you kind of have oversee, and what have you been your current priorities? Any recent deployments and any outcomes that you have?

Inderpal: Yeah. So my digital health or digital transformation journey really started at HSS. HSS was a two-part, right? There was a huge emphasis on operational efficiency because we ran 39 ORs. The more efficiently we ran ORs, the better it was for patient care, for business, for the organization.

So we started with this whole concept of a command center, a digital clinical command center and digital twin, which was a very new concept back then, to really look at and run models on operational efficiency, clinical efficiency, clinical care, and outcomes. That journey started there for me.

In parallel, HSS was also doing what was typical digital transformation of patient outreach, patient engagement, and front doors. Then COVID came and kind of brought a screeching halt to that journey. But lucky enough for me, I was also working in parallel on a very innovative solution for digital pathology.

As you know, radiology has been digital for ages now. We don’t even know what film x-ray looks like, but pathology is still on glass slides and microscopes. That is one thing I wanted to change, and I found the right champion in my chief of pathology. Together, we put through a program, almost a year and a half of development with Epic, our PACS vendor, the scanner vendor, and my team, to develop a digital pathology solution, which wasn’t FDA approved then, so we couldn’t use it for primary diagnosis. But then they applied for FDA approval and once FDA approval came, we were ready to hit the ground running. You talk about outcomes—this is where the outcomes were significantly faster because of the situation we were in.

During COVID, within a year, 70% of all pathology cases were being diagnosed digitally. Huge change for an organization or for a group of physicians who have always been tied to the microscope. It also offers better patient care because now the radiologists, the pathologists, the surgeons, the infectious disease doctors all can correlate images—both radiology images and other ologies—which was not the case before; they had to sit across a double-headed microscope. And of course, for teaching, for research, for second opinions. And then, back in 2021, we did not have it, but now all of the tools are layering on top of it, because digital pathology lends itself to a lot of AI tools. It’s about cell counting, pattern recognition, quality of diagnosis—all of that could be improved and normalized across the institution with the tools. So a really great program and I’m very proud of it, and that happened to be the first in the country back in 2021.

And then I carried on the same digital journey when I joined Englewood. We initiated our whole digital transformation strategy. It was a three-pronged approach: patient engagement and patient self-service; clinically integrated care—because Englewood is a large physician practice organization, 700 physicians in practice across multiple specialties—so we feel it’s much better for us to take care of a patient and we can take care of the patient much better compared to sending the patient outside because of the availability of all data and historic diagnostics for the patient within our system; and then the third piece of that was when the patient is not with us—how do we continue to manage the patient remotely?

Since you talk about outcomes, I think that’s most exciting for the team. As part of our digital front door strategy—the typical patient self-scheduling, rescheduling, request for services—we saw about 18 to 20% increase in online scheduling of appointments, and that number was consistently rising.

The big challenge with that is not technology. It’s aligning the physician organization to open up schedules for patients. Then we ran campaigns for patients. This was post-COVID. A lot of preventive care was put on a back burner by patients, and those campaigns were delivered digitally through Epic’s CRM product. So really no manual assembly and actionable text messages. For first-time annual overdue screening mammograms, we saw a 21% success rate—huge in this case. For first-time screening, we saw 6%. So you’re reaching out to your eligible patient population, and once you set the campaign, it’s really in auto mode. From that point, the patient gets a text message which is actionable: schedule an appointment, move on. Then we extended this to colorectal screening and lung cancer screening. All of them saw at least 7 to 10% success rate. It looks like a small number, but we were doing these campaigns via paper and were not seeing anywhere close to this number. A 10% success rate is huge versus not getting anything.

Another aspect I talked about was when the patient is not within our four walls. So how can we take care of them? We initiated a remote patient monitoring solution integrated with Epic. That’s key—so the physician has the data back in Epic. That program saw success. It was for blood pressure monitoring initially, and 84% of the patients who were part of the program within the first six months saw better numbers—not on any new meds or anything, but with active monitoring, care coaching, and the new technology. We saw a trending line which was better than before—positive trending for those patients. So early success factors there.

Then, this was all around patient care, and we also started working towards physician burden because they’ve been overburdened with all the technology we have laid out. So ambient charting was one of the products we rolled out in our physician offices, and ambient charting has a three-pronged benefit: for physicians, it’s timely documentation, real-time documentation, and being able to spend more time with the patient. It saves a few minutes in an appointment. And what does it mean? It means you’re not staying up late and using your pajama time to do documentation. We saw a 40% reduction of that time for the physicians who were using ambient charting.

The other benefit for ambient charting is better coding. Englewood was at the top quartile to begin with, so we didn’t see better results there, but I read statistics the other day at a hospital in Iowa that was saving an average of $13,000 per physician per year because of the ambient charting product, and most of the recovery was with better coding and charging. HCC coding resulted in $10,000; E&M coding resulted in $3,000. Huge number, and a huge physician satisfaction, even patient satisfaction, because the physician is talking to them. We saw those benefits.

Then we worked on a lot of other back-office functions. We’ve all been very fast at reaching out to physicians via MyChart messages. That burden is huge on physicians with everyone sending those messages. So we put in an AI tool—augmented response technology in Epic—to draft and generate a response to the patient based on the question, chart history, and what the patient has asked in the past. That is again an efficiency function. A quick chart summarization—you’re seeing a patient after a long time; you don’t need to click through the chart— a quick chart summary will come up with all the pertinent things. Then we also moved back-office functions like denials and appeals letters and prior authorization. This is a combination of RPA and some generative AI, but the outcome is important for all of these.

Q: That is cool to know. So obviously, during all these initiatives, there would be challenges in building and delivering the digital capabilities—whether technology, people, resources, or process. What are some of your learnings there, and what could you share with the audience on how you surmount these challenges?

Inderpal: Challenges are many. I will classify them as technology challenges, process challenges, and resource challenges. Technology challenges, for the most part, we are able to overcome, but what we call technical debt sometimes gets in the way. Technical debt is not necessarily obsolete operating systems and obsolete technology. Sometimes technical debt could also be in the form of how a modern system was built and deployed when we didn’t plan for digital transformation a few years later—how appointments were established, appointment types, or departments were established in a solution. All of those build decisions, if they were not planned with a digital lens, will require us to go back. And that happened with us. We had to go back and fix it. It wasn’t about a technology solution.

Sometimes the technology debt comes in if your data architecture is not where it needs to be—you don’t have your right data definitions. We want to make data-driven decisions, and for that, we need to be aligned and have confidence around the organization’s data. That’s one challenge for most organizations.

Then the other is moving your organization—executive and operational leadership—onto an agile process. The first thing I prepared my organization for was: be prepared to fail occasionally. Everything will not be a success because we are not dealing with an ecosystem like big vendors like Epic and Cerner, where it’s a proven thing and it’s a 12–18 month project and we’ll have success at the end of it. We are dealing with an ecosystem where there are a lot of new players in the market. They have promise; some will pan out, some may not. But the good thing is that we will not wait 18 months for the outcome. We will be agile. We’ll try for three months, measure the KPIs, and make a decision. That’s a mindset change for an organization. Once a project is authorized, it’s difficult to walk away from it. That’s one challenge.

Then resources. We are all short-staffed and our resources have a full plate, but are very dedicated and specialized. Now, suddenly in digital, we’ve been used to working in a controlled environment until about five or six years ago. We know the clinicians, how they will use systems; we map their workflows; we build a system according to their requirements. But now you’re dealing with an uncontrolled environment of the patient. You don’t know how they will use it. Suddenly the UX design, the patient engagement piece—your teams have to learn about it, think about it, and the environment they cannot map. They can only predict as a consumer how they will use it. That is a big challenge for resources—not necessarily just a resource gap, but also the mindset change for your teams.

So most organizations are building separate digital teams which have more ear to the ground in terms of what consumers are looking at and are also looking at other industries and how they’re utilizing. I’ll end with one thing: every patient understands—me and you—that seeking healthcare is more complicated than ordering a meal or hailing a ride. It doesn’t mean we don’t want that with healthcare. Deep inside, we all want that kind of convenience, and I think that’s the big challenge on the other side of providing that level of convenience, which is unfair to say because it’s a lot more complex—but still.

Q: That’s true. That’s a good vision, Inder. Moving on to another topic, I wanted to touch upon the role of EHR versus non-EHR platform choices and what that means in the healthcare setting. You make a lot of decisions; you’re grappling with a complex system with lots of enterprise software. Any thoughts there?

Inderpal: I would say this—I’ll answer on both sides. First, my responsibility and everyone else’s is to maximize our investment. If I’ve already invested in a solution—and it’s not just dollar investment, but team, resources, and knowledge investment—and there’s a function offered out of an EHR solution, it’s better because it’s already integrated in the workflow. It will not require relearning a new system and may deliver the same value. That’s the good part. I’ve increasingly found it often turns out economical because the base is all there. The downside is being bound by an EMR/EHR vendor’s product life cycle. To be fair, they have a much bigger responsibility managing the entire hospital: clinical documentation, compliance, patient safety, efficiency. They consistently improve their product as that’s their focus. Customers ask why their million-dollar product doesn’t do certain things. They add modules. They are torn where to focus. Their primary focus has to be patient safety and core system functions. That tussle always exists and may or may not work for me, depending on my timeline.

If I go third-party, we all have been there. Best-of-breed solutions focus on one to three matching functions—often done better—but then integration with EMR is essential, or there’s no utility. There’s added cost, team training, etc. It depends where you are on your journey. Every organization is different. If you absolutely need, say, a patient campaign function with clear ROI, it makes sense to invest. That investment can be interim until your EMR adds a similar solution or maybe never. I think the value is timely implementation and deriving clinical and business value from the solution. If your EMR vendor’s timeline aligns, it’s a good choice. More often than not, the functions you need are in third-party solutions and need evaluation for longevity and business sense. I think both have significant roles and will continue to.

Q: That’s a good point. I like how you said that if a third-party solution integrates with the EMR/EHR, it has more value. 

Inderpal: Oh yeah, that’s the sign nowadays. Otherwise, no technology leader would entertain a standalone solution. 

Q: Yes, integration is key. Now, coming to the fun part, no podcast is complete without AI, GenAI, LLMs, and all these agentic AI things. Where do you see the future of digital and AI transformation? What are some things you are currently or looking to focus on in the future?

Inderpal: Sure. I was listening to someone the other day say that, like everything else, AI had a hype—blockchain had a hype, everyone had a hype. It won’t do everything for everyone, and I agree at a high level. But it’s here to stay and scale. Healthcare did not wait on the sidelines. We all wanted to jump in, even if limited, and most healthcare systems use at least some tools in the space. I use a few; everyone does. Right now, we’re doing what I call low-risk patient engagement, self-services, scheduling—all of those things. That’s how you want to dip your toe in the water. For my GenAI-based virtual agent, initial use cases are appointment scheduling, rescheduling, info requests, payments—no clinical data. That will continue to evolve; this is normal in other industries. I was trying to get services from Verizon yesterday; their chatbot and virtual agent were great—I didn’t need to talk to a person. It’s everywhere; we will get better. Integration and alignment are key to offering those services to patients, so that’s here to stay.

Ambient documentation is here to stay and will be all-pervasive. It’s already beyond physician offices—to inpatient rooms, nursing. Eventually, ambient and dictation systems will merge, become redundant as separate systems. Ambient will continue evolving over 12 to 18 months. Workflow alignment will come. Technology is already aligned and giving benefits; workflow alignment is required to best use it and get outcomes. My long-term guess is that in 4 to 6 years, clinical use of AI will be a differentiator.

The pinnacle will be personalized medicine. We talk about it now—dosage based on population sectors. Personalized medicine is probably the pinnacle, but before that, point-of-care recommendations and decision-making will be reality in 4 to 7 years. Agentic and agent-based AI will take root in many back-office functions, which have many manual processes that can be automated. RPA did some, but now agent-based AI will ramp up.

Digital twin technology is cost-prohibitive today but could be a game changer for personalized care and letting health systems try things rather than run long pilots. If widely adopted in healthcare solutions, digital twin could speed design changes and implementations faster than parallel testing and pilots.

Q: True. We’re seeing a lot in the agentic AI space, personally and at BigRio, with clients including voice agents. 

Inderpal: You’re right. We did omnichannel virtual agents—voice, text, and web—with Zoom, Epic, and Amelia. I saw your agentic AI webinar recently; very cool. I personally haven’t used agentic AI but would like to try it. Sometimes it looks too good to be true, but it’s happening in real time and performing roles. I wish to try it in the next months or years.

Q: Awesome. Thank you for sharing your thoughts and vision on this podcast. Any parting thoughts to wrap up? 

Inderpal: I read a quote recently that stayed with me: “AI won’t replace clinicians, but clinicians who use AI will outperform those who don’t.” I think it’s for everyone. Real transformation isn’t about tools; it’s how we imagine using them, how care is delivered, and how patients experience interactions. An AI agent can’t do that alone—it requires humans: clinicians and operators. That quote stayed with me.

Q: That’s a fabulous quote. I’ll end by saying we have an exciting partnership with a startup to bring AI at scale to healthcare systems. You know what I mean. So yeah, very happy to, uh, collaborate with you in there on that as well later on. So thank you once again and have a great day and we’ll catch up soon.

Inderpal: Thank you. Thank you for the opportunity and thank you to your viewers.

Subscribe to our podcast series at www.thebigunlock.com and write us at [email protected]  

 Disclaimer: This Q&A has been derived from the podcast transcript and has been edited for readability and clarity.

Rohit: Hi Ashis. It’s great to have you back on the Big Unlock podcast. Thank you for joining us, 

Ashis: Rohit, Ritu, it’s wonderful being back. It’s been a few years and I’m excited to join again. Thank you for having me.

Rohit: Thank you, Ashis. As we discussed before the podcast, we are carrying on Paddy’s legacy and are very fortunate to be doing so. Our podcast is now more than 170 episodes, so we have a very excited base of listeners for this interaction with you. I’m Rohit Mahajan, Managing Partner and CEO at BigRio and Damo Consulting, based in Boston and host of the Big Unlock podcast.

Ritu: Hi Ashish, really nice to have you on the podcast. Welcome. My name is Ritu Roy, Managing Partner at Big Rio and Damo Consulting, currently based out of Gurugram, India. Looking forward to a very engaging discussion. Thank you for being here. 

Ashis: So good to be here. Thank you again for having me. I am the Chief Digital Technology Officer at the Hospital for Special Surgery, which is a hospital on the Upper East Side of New York and the tri-state region on the East coast. We specifically focus on musculoskeletal care. You may know us as the orthopedic surgical hospital, but we actually have the largest rheumatology practice in the world, as well as psychiatry and other aspects of skeletal care, so we provide the full spectrum.

I started my career in Texas and I am a pediatric gastroenterologist. I’ve never stopped practicing and I practiced for a long time at Baylor Scott and White, which was the position I held when I was fortunate enough to meet Paddy long ago, where he became what I consider a friend.

I told him long ago, the Big Unlock was absolutely the main podcast I listened to—one of the first I started listening to when I wanted to learn about digital health. I was a frontline doctor and felt that digital health was going to be something significant for healthcare. Before people really knew, I think Paddy was onto something, and the fact that he made it his mission to talk about it, to educate, and to really move the field, is inspiring still. I think that has really created this forum of where it is now.

When I was at Baylor, I was fortunate to be at a system that was very far ahead in the virtual and digital space. I then moved over to Highmark and Allegheny Health Network as the Chief Digital Information Officer. I did that for a few years—a fantastic time. I learned about a whole new world as a doctor: insurance and value-based care specifically, and was able to make a lot of great programs there and do a lot of good work.

Then, when HSS came calling—which we can talk about more in the podcast as to why I made the move—I jumped at it about 10 months ago, and it’s been the best decision of my career.

Rohit: That’s great to know. Ashis, so very curious. You said that you never stopped practicing, right? You are looking at digital transformation initiatives at one of the largest hospitals in orthopedics and MSK in the New York area, which is one of the largest states. So how do you balance it, Ashish? That is one of my questions. And second, what motivated you in the first place? I know that you started your work as a physician, but what motivated you to become a physician? Please share that story with us as well.

Ashis: Oh gosh. Okay, so I have trouble with brevity at times, so I’m going to do my best. It’s a big question. Let me start with the latter and then you can remind me the former. As far as being a physician—and you said balance—I don’t believe in balance anymore.

One of my favorite books is Conscious Business. One quick premise of the book is that work-life balance is kind of like eat-life balance. You eat as part of life and you don’t really think of it as separate. Sometimes you have to stop to eat, sometimes you really enjoy the meal, and that meal can be at 11 o’clock in the morning, or it could be at nine o’clock at night, or anything in between. The point is, if your passion is your work, then there doesn’t have to be a line between the two. There are times in the day when I can take my son to a game at 3:00 PM and that’s okay, but then I may be working at nine o’clock at night, and that’s okay too. The balance aspect of things is always a struggle, but I think that if you strive for balance, you’re maybe focusing on the wrong thing because you probably don’t have balance with what you’re doing at work. Does it feed you? Does it give you passion? Does it give you energy? I don’t have a problem with it. My wife may argue at that point, but I don’t have a problem with it in the sense that my work is my passion. It’s what gives me energy, it feeds me, and I enjoy it.

I think before the podcast, you told me that when you met me the first time, you didn’t get too many words in because I was speaking with such passion and energy and fervor about what we were doing. That’s just me. I apologize for that, but that goes to show how much I really enjoy the space. My life is my work, meaning my kids, my wife, the dog, and everything else comes with it—just as passionate, just as energetic—and they melt together.

All of that being said, I’m a doctor first. I’m absolutely a doctor first, a technologist second, and I am to this day. I think that’s a differentiator. I think it’s a bit different than most people with my titles and roles, and I hope that brings a different perspective and value to the equation and discussions. I grew up during the High Tech Act and turned on a lot of Epic systems and what have you.

To answer your question directly about why I became a doctor: I am the first doctor in my family. So that’s not a typical Indian answer—many doctors and engineers in Indian families—but my parents immigrated right before I was born. I was born in Chicago, we moved to Texas when I was little. I’m an ABCD kind, if anybody knows that term. I was born in the States and grew up in Houston, and it was a passion from the get-go. I always knew I was going to be a pediatrician. I really enjoy families; I enjoy working with kids. That was something I just gravitated to, and there was never a question of anything else—pediatrics was something I wanted.

The other side of it is, we took a family trip to India when I was a kid, around eight years old. I actually contracted typhoid fever when I was in India—salmonella typhi. It’s very significant, and I was in rural India at the time. It was misdiagnosed as malaria and I was being mistreated. As an eight-year-old skinny Indian kid that maybe weighed 60 or 70 pounds at best, I lost 25 pounds at that time, and I was actually very close to passing. By luck, my cousin was marrying a physician in India—Manish, who lives in Ohio now. So, Manish, if you’re listening, you saved my life. Manish knows this. At every wedding we see, he says, “You’re alive.” To this day. Manish Bai came by because he heard that I was ill and he quickly figured out that I had typhoid fever. He brought me to his home, treated me, and I was able to get the right treatment and get back to the States. Obviously, I’m here and well.

That played a significant role in my life; I wanted to be a doctor like Manish—he is a family practice doctor in Ohio now. Pediatrics was something that resonated with me because I was a kid at the time. It plays a big role in who I am today. I received the wrong care, I didn’t have access to care—I received the wrong care until I got the right care. So the questions are: how do we distribute, how do we democratize, how do we get the right care to all people? Again, being at that age, I think it was a profound effect on me.

Rohit: That’s an amazing journey, Ashish. Thank you for sharing that background and insight. And Ashish, that brings us right to the persistent challenges in healthcare. We all kind of know what they are. I have a laundry list here: limited access, rising cost, clinician burnout, admin burden, and systemic waste. There are many persistent problems that we are trying to solve. I think we are making great progress on several fronts. How do you think about your digital transformation efforts at the current organization and what you have done before? Can you share with us some thoughts and ideas on how you approach this and what solutions you’ve been able to put into place to address these?

Ritu: Yeah, just before you answer, at Human X, there was a panel and a very interesting perspective on this—how slow things are in medicine, and how it takes years for doctors to adopt new ideas. Even the stethoscope took many years before it became acceptable. Because you bridge that gap between being a doctor coming into technology—when you address these problems, do you feel you have to address the issue between how technology moves, especially now with AI moving at light speed, and how things in healthcare move slowly? Is that frustrating for you? I just wanted to bring that in.

Ashis: Now you know that they have long-form podcasts that go on for 48 hours. There’s so much there. But knowing this is not one of those, let’s keep it succinct but try to create value in my answer.

A short answer—frustration—or it’s one of those things: how do you look at it, and do you see it as an opportunity? Do you see it as something we can really go do? The short answer is the time—even best practices. The amount of literature is doubling at a rate that is unreal. As a pediatric gastroenterologist, I can’t read enough journal articles to know what is published, let alone keep up with MSK, orthopedics, and everything else. There just aren’t enough human hours in the day, with all the other burdens of healthcare. There’s much more data and information constantly coming up. The average time for a physician to put a journal article or best practice into practice is something like 16–17 years.

Then, the innovation timeline—from something being invented to being used at HSS—is very long. The first knee implant was invented at HSS; it’s a culture of innovation. But by the time things get invented and then widely used, that timeline is massively long. How do we shorten that? I believe AI can significantly shorten that journey, especially with cognitive knowledge—getting best practices to pop up in the care journey, nudging clinicians when new evidence emerges. That is 100% doable, and you’re already seeing some vendors partnering to bring that insight into the workflow. That excites me.

For the broader question of digital transformation, there are many ways to frame it. The very broad way: how do we improve clinical outcomes and reduce administrative waste? Generally, let me zoom out and say—I was talking to an academic professor recently, and she said, “The only two things we should be talking about in healthcare right now are Agentic AI and change management.” To some degree, you can debate that, but they impact all aspects of healthcare.

I’m a techno-optimist, but I don’t just want tech for the sake of tech. I want tech that works, that actually solves problems. When I was a frontline doc, there was so much tech thrown at me, supposedly logical and great, but people didn’t realize it added burden to my workflow. So, part of my journey is representing the front lines—knowing what problems actually need to be solved. That takes a lot of listening, learning, and observation.

What excites me most now is Agentic AI. Healthcare is about workflows—it’s not about a moment in time. No doctor, nurse, patient, or consumer spends their healthcare journey in just a moment. Point solutions and things like generative AI chatbots still solve only moments in time. If I have a rev cycle authorization tool, or an ambient scribe, or an OR dashboard, that’s only solving a moment. But in healthcare, problems have downstream and upstream effects through workflows.

When I think about Agentic, it’s a workflow orchestrator—something that hasn’t existed outside the EMR. Of course, EMRs have workflows, but with lots of friction and clicks. Agentic has the potential to be a workstream orchestrator for everyone—consumers, operators, clinicians, administrators. This orchestration is where the magic and value will be, because every handoff in healthcare—from scheduling to aftercare—is where friction and waste happen due to lack of coordination across silos.

I believe as we solve across, and make the “white space” between verticals go away, we improve experience, coordination, and outcomes—and reduce cost. But that world has to be purposely built—it won’t magically appear from a platform or from thousands of point solutions that somehow orchestrate together.

My teams are working on that, with the first part being our data. We are building Lakehouse architecture—making sure our source of truth is in one place, linking all the data within context. Whether it’s consumer data, finance, operational, HR, clinical, wearable data, etc.—with all that linked, an agent system on top can orchestrate to take real action, which can then inform humans.

Lastly, it’s important I say all this is not to imply that there will be fewer humans in healthcare. If you’re touching a human, if you’re in front of a patient, we need to double down on that. We shouldn’t orchestrate, automate, or agent that away. Rather, we want to take away backend processes so we can double down on human engagement at the front end. The question is: what should be done by humans, and what by AI? Change management—upskilling, reskilling, onboarding—really means asking what needs to be human and what is best done by automation. That shouldn’t create fear that there’s loss; I think there’s actually gain—more time with patients, more humanized healthcare. I genuinely believe, if done right, AI has the potential to actually rehumanize healthcare.

Rohit: That’s true. 

Ritu: Great answer. Ashish really hit the nail on the head there. 

Rohit: I think it’s a different way of thinking, like you said, Ashish, about workflows, change management, Agentic AI, and all the white space that can be filled in. Would you think of any possible example? This data preparation that you’re building in the lakehouse is not a trivial effort. It’s going to be, possibly, a multi-year journey. And now you’re layering agent AI on top of it to fill this white space and orchestrate everything in the workflow. So is this too far in the future or near, and what are you seeing in terms of timelines? Also, in change management, what are some of the challenges you’re looking at, and how are you possibly overcoming those?

Ashis: It’s a great question and I understand the essence of it. This comes back to why I am at Hospital for Special Surgery. Let me frame that first, because I think it’s important. I was at much bigger companies, in terms of revenue and size. There’s a very purposeful reason for my move, based on my learnings. One of the struggles—and it’s understandable, especially since I’m very mission-based (as you may recall, I once said we’re missionaries, not mercenaries, and I still hold true to that)—is that healthcare is so complex. We all know that. Healthcare is complex due to the human factor, payment complexity, and more. In big systems, what’s harder is orchestrating between cardiac, cancer, peds, and different geographies and cities. What you described could take five, ten, even twenty years—it’s like boiling the ocean; it’s massively complex. You end up being one inch deep on everything and don’t know where to start. If I go to another system, do I start in cardiac, cancer, or orthopedics? No one wants to be second or eighth on the list; there’s politics and competition among divisions, with everyone wanting priority.

Healthcare is now hyperspecialized—doctors think inside their narrow specialty, which sometimes holds us back from thinking more broadly. The “focus factory” that is HSS—just doing one thing and doing it at a superb, world-class level—is very important. It lets us go deep, which is extremely difficult for broad systems. Even though orthopedic and musculoskeletal care isn’t super narrow (30–40% of people have mobility problems, so the impact is great), we get to go deep in one specialty. A lot of what we do is elective and algorithmic, which aligns well with Agentic AI.

So for us, there’s a bit of a perfect storm of positive aspects: we are the best at what we do, focusing on one thing, and it’s highly algorithmic (while still having plenty of human factors). I believe what I’ve described has a timeline of two to five years, rather than ten to twenty. At HSS, my and my team’s hope and vision is to create a very talented team to do this. We’re building a lighthouse for what is possible in healthcare, to show the world it’s possible to automate backend processes and orchestrate workflows. Our mission is to show that this role is possible—because of this focus factory aspect and being the best at what we do—and then distribute that globally.

If we codify the knowledge base that’s here at HSS, the best at what we do, why can’t AgTech orchestration distribute that, not only across rural America and the US, but also the globe? It can be codified. We really see ourselves as having an obligation to take what’s incredibly special at HSS and extend it. When you walk into HSS, you feel the experience, passion, and outcomes—it’s incredible. The question is, why is it only those living in the tri-state region who get access to what healthcare should be?

It’s really incredible. How do we take what’s so special, these care pathways, and this innovation, and broaden that knowledge, education, and capability globally? If we’re at the tip of the spear—using AI navigation, robotics, and other aspects of musculoskeletal care—then we want to broaden that knowledge and ability for everyone. 

And you know it, and the surgeons and the rheumatologists and the nurses at every level, an exceptional and it’s, you know, for me. It’s, why is it only if you live in the tri-state region, do you get access to the, what healthcare should be? 

And you know it—the surgeons, the rheumatologists, and the nurses at every level, they’re exceptional. For me, it’s: why is it only if you live in the tri-state region that you get access to what healthcare should be?

It’s really incredible. How do we take what’s so special? How do we take these care pathways? How do we create this innovation?

If we have something we’re able to use at HSS—and we are at the tip of the spear, using AI navigation, robotics, and other aspects of musculoskeletal care—how do we broaden that knowledge, education, and ability across the globe?

Rohit: That’s amazing, Ashish, very admirable vision. For those people on the podcast and for myself, I’m curious to learn a little deeper—it’s a more tactical than strategic question at this time, but you mentioned algorithmic. Could you throw some light on what you mean by that?

Ashis: Let’s talk to that quickly because I don’t want this to sound like just medical futurism—like, “oh, in 30 years we’ll do it.” How do we get there? As I said, data architecture is super critical, and one thing I’ve learned, and for which I’ve hired significant talent, is really focusing on the data. From my position, growing up as a doctor first, then becoming a technologist, everyone talks about the wonderful things, but what’s step one? Where do I put my first foot down? It’s very difficult. What I implore anyone on the podcast: really start with the data and make sure you have a data strategy that allows you to put an agentic system on top of it. The capabilities are there, the tools are there, but the context may not be. You have capabilities like OpenAI—now they’ve launched ChatGPT agents, it’s incredible. That’s the capability, but it has no context, no data to inform or make an intelligent decision. We need to get the context right, which is getting the data right. First step is making sure we get the context and the data right, and we’re centralizing that.

The second aspect is that we’ve made key partnership decisions. This is not something we’re going to do alone. We’ve made some key partnership decisions to bring big tech and startup innovation into our ecosystem. I’m a big ecosystem fan. I think it’s going to take an ecosystem. So, this week and others, we’ve launched ambient listening—it’s not just listening. We don’t see it as only ambient listening. Describing is only the first step. Having the scribing inform coding, CDI, authorizations, scheduling, even CRM tools—this is how we see the platform going forward. We’re looking at that.

Secondly, we’ve partnered with Palantir. What is Palantir doing for us? Palantir is thinking through the journey from end-to-end, from a consumer lens, from the beginning of care to orchestration. From operational flow, from a business intelligence standpoint—though that term minimizes it because it’s much more than that. It’s creating a kind of GPS—I want to define the main road of care, and when you get off the path, the GPS system activates to get you back on the main road. I think of Palantir building that with us.

Third, I’m a strong believer in low-code, no-code for the masses. Palantir is pro-code; it’s heavy and not something I can just democratize for any employee. So I need another agent layer of low-code, no-code. I think the future of work in healthcare is operator and engineer, unlike when I was a pediatrician and had a great idea but had to put a ticket into central IT, then wait six months only to be told it’s lower priority. I knew it would help my patients and outcomes, but I didn’t have the tools to build it in a safe, secure, PHI/HIPAA-compliant, easy workflow.

I think that’s changing. If I have an agent platform, from a coding perspective, I can say, “Here’s what I want,” and that’s what’s changed. OpenAI gives coding ability to everyone—it’s in plain English, written or verbal. So now, why can’t my revenue cycle folks build their own workflows? Why can’t my surgeons? That’s real capability. Why doesn’t it work today? Because the data is siloed. If they build an agent for calling post-op patients, another for collecting prompts, another for getting access, that’s like five different phone numbers—they’re not orchestrating. It goes back to vertical agents that aren’t orchestrating care, creating more friction and cost.

So, the answer is orchestration—having one agent platform for the enterprise and making sure the data is structured so it allows orchestration of those agents. So, it’s ambient, our Palantir work for high-code, and then an enterprise orchestration platform as well. Those are the three partnerships we’re building.

Rohit: That’s great. So Ashish, I think we’re coming towards the end of the podcast. As you said, there is so much to talk about—it went by so quickly and was a very different perspective. Would you like to offer any parting thoughts? Hopefully we’ll have you back soon to build on what we discussed.

Ashis: It’s important to me that we have a responsibility to transform healthcare and make it better. While I think about AI and agents—Dr. Michael O’Hara, our chief data analytics officer, always says there are two E’s in AI and everyone forgets the second. The first E is efficiency—everyone talks about that. We need more efficiency in healthcare, but I think we miss the ball if we only focus on one E. The second E is effectiveness. We can do better than we do today. It’s not just about doing things efficiently; the goal is to actually do better. People have unmet needs, lack access, or get the wrong care more than the right care. We need to do better—not just efficiently, but more effectively.

Hospital for Special Surgery does the best at what they do, and that’s true. What we’d love to do is ask: what do we do better, how do we codify and share that, and help make healthcare more effective in musculoskeletal care? If we can package up what’s special at HSS and make it a global brand, we should do that. Lastly, even though “surgery” is in our name, we actually do more non-surgical than surgical care at HSS—not many people know that. We live in the world of movement, and people care a lot about movement—wearables, watches, step counts. When you lose movement, you lose much. A lot of healthcare is about loss—loss of function, cancer, chronic disease—but what excites me is that musculoskeletal care is about gain. People want to move better, play with grandkids at 80, run marathons at 60, live better and longer. The number one component of longevity is movement. You don’t want to be 120 years old and wheelchair-bound—you want to be active. So we’re thinking more broadly—movement, not just orthopedics. I’ll leave you with that. Thank you.

Rohit: That’s awesome. Thank you so much, Ashish. It was pleasure having you on the podcast. 

Subscribe to our podcast series at www.thebigunlock.com and write us at [email protected]   

Disclaimer: This Q&A has been derived from the podcast transcript and has been edited for readability and clarity.