Episode 182

Introduction and Guest Background 0:00

I take it in a futuristic state, and I can't tell you when that future is, perhaps. But none of the modern AI tools is anywhere close enough. Now, think about this in the non-health care, for example, autonomous, full self-driving cars. And the billions and billions of dollars that have been spent on that effort, we still need a human in that loop. And so the way I think modern AI works best, and I can't predict 15, 20 years from now, but suddenly I over the next decade, is that there are some tasks that AI excels at.

And then there's some task that humans still continue to excel at, And that's what we mean by what is foundational to our platform is this human in the loop design. Welcome to Medical Money Matters, the podcast where you can find experts, answers, and resources so that you achieve mastery over the financial and business aspects of your practice. Hello everyone and welcome. Today's guest is Sarub Gupta, founder and CEO of Coremetrics, a company at the forefront of transforming how healthcare organizations and legal teams interact with complex medical data.

With decades of experience spanning healthcare technology, data strategy and enterprise innovation, Sarab has built a career around solving some of the industry's toughest information challenges. Through CoreMetrics and its Verix AI platform, he's helping organizations harness AI in a way that is practical, trustworthy and deeply grounded in clinical reality. We're going to talk about the future of healthcare intelligence, the promise and pitfalls of AI and how better data can ultimately lead to better decisions, better operations and better outcomes.

Saurabh, welcome to the podcast. Jill, it's so good to see you again and thank you so much for inviting me onto your platform. Yes, absolutely. I'm excited to jump right in. So for our listeners who may not be familiar with Core Metrics, can you give us a little bit of the origin story and talk a bit about the problem in healthcare and data management that you were trying to solve when you founded the company? Of course, with a long background in healthcare, Jill, as you know, we are now in an era, or we have been since the advent of the modern electronic health record system, where data capacity is actually not a problem, it's data abundance.

And for everybody who has seen electronic records, who's interacted with them, the amount of repetition that exists, and the data that exist, is not always useful. And that is the genesis of what we're trying to solve. And healthcare data, as you know, is personal to all of us, right?

CoreMetrics Origin and Healthcare Data Problems 2:56

I mean, healthcare is a major part of the economy, but more than that, it's just such a personal endeavor. So whatever we can do to increase the quality and how we interact with that data and learn from that. It is good for us in general as a society. Yeah. Coming through enterprise systems, we felt that this was a problem worth solving. Excellent. I love it. But you have a very broad team, too. You include clinicians and engineers, enterprise tech leaders. So I'm curious how that multidisciplinary approach has shaped the way that you develop products and services.

You know, in any startup jail, people talk about product market fit. And I actually think of it a bit differently. I think this as founder market-fit. What I mean by that is that people sometimes, for example, you have a startup, they find some cool technology, and then they try to figure out applications of, hey, where might I apply that? Whereas if you think about your career and our careers, we are looking at this from the other side, and what are the problems worth solving? And when we then apply our lived experiences of how modern healthcare has evolved, how we find that we're able to do these cutting-edge therapies, cutting edge procedures, whether that be in oncology or cardiology, one of the most powerful transformations has been the multidisciplinary approach.

understanding that not one person or one group might have the requisite expertise to solve these challenges. And I'm also a big believer that we as clinicians, we, as healthcare professionals have to be at the forefront of solving medicine's real challenges and so from that brings this multidisciplinary concept and how do you apply that to the modern company? I mean, everybody talks about AI. People have been talking about the AI now for two or three years. And my approach in that was straight from the outset was that we have to have a multidisciplinary approach to solving these problems.

That includes clinicians, that includes data scientists, and that include engineers, rather than the conventional approach of saying, hey, let's develop some technology and then figure out what we want to do with it. That's fantastic. You have quite a team. It's impressive. Thank you. Yeah, I'm curious. Can you tell, just for me and the audience, who is your client and how do they use your product? As I mentioned a little while ago, we're really trying to solve the problem of health data and, how you make sense of this.

And at a very conceptual, far futuristic state, learning what we should do before it has happened. So if you think about healthcare data, the data starts in the healthcare system, and if there are quality concerns, if they're procedural concerns then the health systems, clinicians, administrators interact with that. That process is required from regulatory bodies, called the peer review process. Those are internal. And then the same data, if, for example, a patient or their family feels that they've been harmed, that same date then flows to a litigation team, perhaps towards a plaintiff attorney, who have their ecosystem of experts, such as legal nurse consultants and health care experts who analyze that data set along with the attorneys.

And, then, as you move forward to that chain, you have defense attorneys, insurers, and peers. Fundamentally, it's the And the way we conceptualize that is that we can take the data and we take it in a very unstructured way and then really apply

Multidisciplinary Product Development 6:38

very modern AI tools. For example, some of the technologies that they've developed, even the back one of those did not exist three to six months ago. That's how cutting edge we are. And then essentially, when the records get in, we can flag what might be missing. So sort of what you don't know, what do you not know? What the gaps in care might, be what the vulnerabilities might the from a legal defensibility or from plaintiff's side. And finally, you can actually give them predictive guidance on what is the direction that you should go this from either a quality side or formal litigation side?

So essentially our customers are health systems, our consumers are also attorneys and litigation teams, and they're associated ecosystems of medical experts or legal nurse consultants. Got it. Okay. Thank you for that specificity. That helps to kind of dial it in. So let's talk about tackling healthcare's toughest data challenges. that's what your website says. And you've talked a little bit about this, but maybe drill down a bit more about why medical data is so uniquely different as compared to data from other industries.

You know, if you look at sensitive data and we were to sort of say, hey, what are the most three sensitive type of data sets? And pretty much everybody would say legal, financial, and healthcare. And out of that, I would argue that healthcare is the most sensitive type of data. It's also the more personal type data, and then what is fundamentally more person to us than our own healthcare. And it's all sort of the messiest kind of thing. If you look at the fragmented data ingress and egress, despite the advent of modern electronic health record, data is still very fragmental, very fractured, And our ability to learn from that data is not at all where we would want it to be.

For example, people have tried to tackle that problem and they say, one approach might be let's take all the data that exists in an electronic health record and let us apply some AI on it and then try to make sense of what comes out on the other end. And we fundamentally think that's the wrong way to approach these problems, because anybody who's worked in healthcare knows that data can be repetitive, data is actually not pure, and data could be multi-sourced, multi mortal. And that does not then lend itself to making good, clean conclusions out of it if you don't have the domain expertise.

Who CoreMetrics Serves and How It Works 9:19

And our approach to that is that we can take all of these different types of data sets, whether that be imaging, the CT scans, for example, or other types imaging. And then we can use our customized algorithms and AI to then explain to the relevant audience, which is context dependent. For example, if you're a medical expert, you might want different things out of that compared to if your health insurer compared if an administrator compared with an attorney. And that's how we have approached this, that medical data is contextual and it has to be interpreted in both the context of the situation, but also to who is accessing that data.

There are a lot of different audiences for that data. And a fascinating approach, too. I'm curious, and I promised you we'd go down a rabbit hole or two. So here's one. Because you said something important, the medical data is fragmented. Here we are in 2026. We have been talking about interoperability in medical date since as long as I can remember. And yet, here we are. We're still very fragmented, to your point. I wonder, as you look at the whole landscape, why do you think that is? Why do think it's been so challenging for health care and medical practices in particular to get to that nirvana of interoperability and kind of think about a single patient record, a simple source of truth for the patient records?

I'm curious what you thing are the barriers for that. Yeah, wouldn't we all love that if that were to happen? I think there are several reasons. One is the evolving landscape of how medicine in the U.S. has shifted, right? It's gone from the individual solo proprietor, whether that be a physician in a small town or a nurse, to this modern concept of employed physicians and large health systems. So that's one. The second is that a lot of the electronic health record market is by definition subject to special interest forces, where companies might try to maintain their own supremacy in that field.

heard is that there is not an enforcement legislated mechanism behind that our leaders in Congress can agree or get behind. For example, one of the simplest and basic fundamental things to what you're just saying is to have Let's even go a little upstream from a unified record. It's to have a unifying identifier that can you say that this forum is the same forum who might have shown up at an ER in California and we don't have that either. And so I think that coupled with the economic forces makes this very challenging.

Now, it's not all doom and gloom. I think there is some movement in that direction with evolving federal FHIR standards. And we're slowly but somewhat unsteadily moving towards that future nirvana is where I would put the landscape. I love it. You may recall, but I'm going to say mid-2000s, Microsoft took a swing at this with a product called Health Vault, which on the surface looked like it was very promising in terms of getting us to that sort of single source of truth for the patient record. My guess is that looking at it from almost inside that technology rollout, that it is a little too soon to market.

People weren't ready for it yet. Think so.

Why Medical Data Is Different 13:02

And I think our ability to do more with that data has also now improved, right? I mean, not only is just the data set important, but the ability access that ability, to make sense out of that. At a fundamental level, that's what we do. Now, even behind, beyond Microsoft, there's movement in that space. For example, if you look at Google's product, which I'm not a big fan of. I have serious privacy concerns around those, for example, to upload all of your data only to be monetized later on. And fundamentally, that seems wrong and insecure to us.

But yes, I think it's definitely a problem worth solving where there wouldn't be a unifying source of truth. And then the other related piece to this is that when we're solving for healthcare, we are also solving the entire population, which that means that we have to think about what happens to the data set in patients or people who may not be as internet savvy, for example, or not as tech literate, who might be small children or might otherwise be a wonderful population. I mean, it's a tough problem, but I absolutely do believe that they're making progress in the correct direction.

We're headed in the right direction. That brings me to a good question to ask about your platform, which combines AI with clinical expertise for medical legal review. And maybe you could talk a little bit with us about why it was important to build a health care specific AI platform instead of relying on some of the more generic AI tools. Absolutely. So, on a fundamental level, it would have customers who are, when we have our initial calls for the mask, as well, why can't I do this in Jack GPT or Claude?

And there are several problems with that. One is that, think of healthcare, and you say, well I'm going to the hospital, any doctor could maybe see me, if I need my appendix surgery, I don't really care whether a surgeon does it or a cardiologist does I mean, I'm a cardiologist by training, and trust me, you do not want me taking out anybody's appendix. And in a way, we have to ask the question on whether an undifferentiated tool is the correct way to solve this problem. a genetic AI that is powerful enough to be able to do all things to all people at all times.

So that's problem number one. Now, if you think about what actually GPT stands for, out of chat GP, I don't know if very many in your audience has ever looked that up. It actually stands general purpose transformer, or GP. The second is that if you don't have domain expertise, you go back to the same problem that we were talking about earlier, meaning that you can get the data and it sounds very intelligent, but then once you start interacting with it more, And hey, it's not that intelligent. It's no that nuanced.

And they're getting better at that, but nowhere close to perfect, especially as we see the differences in how our platform looks at versus if we do just a general purpose tool. Well, maybe not the most important. Most important would be privacy and trust. But the third sort of is that health data runs into thousands and sometimes hundreds of thousands of pages. And it is context, it's highly contextual. And if you start thinking about how, let's say I'm sure you have, I have most of our audience might have uploaded a file or two to GPT or Claude.

And now we will stay away from the fact that you could upload a one-page file and it can give you a 10-paged thesis on it, which most of it makes no sense. But even besides that, if you think about 100, 200, 300 pages, yeah, you might maintain context. Well, when you start thinking about 50, 60, 70,000 pages how do you maintain contexts on what occurred?

Interoperability Barriers in Healthcare 16:58

So for example, a heart surgery occurred two, three years ago. And then 50,000 pages later in the health data, some medical device failed. The problem is that by that time, a non-contextual system has lost all track of that. And that's part of where the hallucinations, et cetera, come from. Because they just start making stuff up. Too much data, but the two important points. Yes. Right. Then finally, of course, is the trust and security and integrity around that data. Do we really want to be uploading hours and our customers and clients data into platforms that have very little transparency behind the scenes.

That is a fundamental question. I know that one's getting batted around a lot. that brings up an excellent question as you and I were talking earlier about human in the loop and why that is foundational about how you're thinking about this and thinking human-in-the-loop or fully autonomous systems and from your perspective, where's the industry go with regard to that question? I think in a futuristic state, and I can't tell you when that future is, perhaps. But none of the modern AI tools is anywhere close enough.

Now, think about this in the non-health care example, autonomous, full self-driving cars. And the billions and billions of dollars that have been spent on that effort, we still need a human in that loop. And so the way I think modern AI works best, and I can't predict 15, 20 years from now, but suddenly I could over the next decade, is that there are some tasks that AI excels at. And then there's some task that humans still continue to excel at, that's what we mean by what is foundational to our platform as this human in the loop design.

which is that we will get you most of the way, we'll get to 99.5, 99,8, whatever that percentage might be. But that final step has to be the human in the loop, and it has be built in programmatically. It has have that layer of defensibility from what it is I actually think I as a human, not I, as company. So take an example for, let's say, a quality product in the health system side from a peer review. So we do have a module called PeerPro where you can essentially just AI automate all of your peer-reviewed processes in your health systems, right?

And for large health-systems, that's $100,000 of spend. But do we really want an autonomous AI system to be making credentialing or disciplinary final say? And I would say absolutely not. But we can get you 99.9% of the way. We can take all the data, we could take the relevant context, make sure that data is complete and audited. Make sure we flag where the gaps in the scare are. And then we would flag a recommendation. whether or not that is the correct recommendation and whether that's the recommendation that the reviewer wants to go with, has to be a human in the loop component.

That final step. So we see a future, or it's not even the future. It's a present where most of that work has occurred. You're not going to miss stuff. you're going be pointed in mostly the right direction, but that final layer of wisdom, judgment, and insight belongs to the human. That's a fantastic sound bite right there, the final layer. I love that. That is an excellent example of using the two, right? The AI in combination with the human in the loop. to assure that you get the right outcome because peer review can be very, it's very important to the clinician, obviously,

Why Build a Healthcare-Specific AI Platform 21:00

and also to health system in question that we get that right and that have good wisdom and good judgment included. Wonderful. So that moves us right into our next question, which in healthcare, this notion of almost right can still be dangerous. Let's talk a little bit about how your platform ensures traceability, defensibility, and trust in the insights that you generate and how good is good enough. So I think that's a very important question and a great question, that how good is good enough? Are we held to a 90% standard?

Should we get this 90 percent of the time strike? Shall we give it to you 100% of time? And I would say that 100 percent for any human or artificial system is an aspirational goal. We just cannot be perfect in all given scenarios at all times. And it's a question that we struggle with, even on the development side, on when do you roll out these tools as they go through our intense development process, vetting process etc. And we think that the way to approach this is to be transparent about what you could do well and where you can't do it.

So, for example, I would love to see the more narrow limbs publish a hallucination index that, hey, we've gotten to 90%, we're going to 95%. Now, honestly, that's not in their best interest to do that. Right. And there's no requirement to that But on the back end of our platform, we're very careful on that type of rigor behind this to make sure that we get products out that can stand behind those. But we also think that layer is then augmented by the human in the loop. That if you build those in programmatically and systematically, then you actually have the final defensibility there.

So think about this as a chief medical officer or a practice administrator. They are the ones who are actually putting the data together after having all the facts. And the AI engines might provide recommendations, they will provide very strong recommendations and probably far better recommendations than most humans do. The second part of this is that are we careful when we're vetting these solutions that they're free of biases? So here's an example. If you sort of, and I think all your listeners, if they haven't already, should try that experiment, that when they log next time into ChatGPD, they should and write a question or a prompt in a way that slightly suggests a bias.

And say something like, do you think I would have a good time going to a vacation in the Caribbean? And now you will see that the bias of all of your responses is going to be in favor of why you should go. Okay. Now, if you reword that question slightly differently, and create a negative bias that say, I really don't want to go. And the entire data set, the entirely chain or the stream of consciousness, if consciousness can exist in AI, would bias you to that direction. That's a bias. So it's going to send you Hawaii or Mexico, not the Caribbean.

Yeah, so it's going to channel your inner thoughts to the extent that it thinks, and a lot of them want to please humans. So that then is not the same level of objectivity that you would want around medical data, around clinical decision making, etc. And so we feel very strongly about integrating that at the DNA level off the platform, if you will. That is, it's fascinating to think about how the bias exists in the, just in that operator, you know, who's utilizing it. And then I love that notion of kind of.

filtering that out before it gets into these things that are so important and so critical. The end product for your end user is life-changing in many ways. That we want as clean and free of bias as possible. So that's a good segue for us to talk a little bit about one of the things stood out for me on your website is this notion of conversational data.

Human-in-the-Loop AI and Defensibility 25:10

And so I wonder if you could talk a little bit about how natural language querying changes the way that attorneys and clinicians and health care professionals interact with complex medical records and also goes to that notion of bias as well. So, if you think about how computer algorithms have worked from the very early days, it says zero or one, right? Either it occurred or it didn't occur. And in a way, you could talk about that in the healthcare landscape, that some terminal event occurred and not.

somebody survived or didn't survive. That's a zero one outcome. Zero, yes, one, no. But most of health care, most life is actually varying degrees of grain. There's 0.1, there's And then how do you capture that? How do sort of say that there are nuances to this? Take an example. If you ask a question, let's say, should a cardiologist be delivering babies? The answer would be no. Right? In the binary outcome, that'd be yes and no? What about if they were the only doctor on a plane? Should they then be deliver that baby?

And the answer will be an unqualified yes. Well, so here's an example of the same question, but it becomes context dependent. And if you didn't give that context in a conversational way, and if your ask the question as a zero one or a yes, no, then that leads to, we believe flawed outcomes. Now you can introduce varying shades of gray into this. What about if you have a cardiologist but also an OB nurse who is not independently certified to deliver the baby? What's then better? That's another level of nuance.

So I think the point is that we should be thinking about how humans are approaching this data set. We're asking the correct questions. When you and I are talking to each other, we are asking follow-up questions, and we're following those strains of thought. And that's what we mean by this natural language ability to interact with that data. that you don't want to go on a tangent with the data, but you do want the ability to have back and forth, you want context to that data. So for example, if you say, show me the EKG in our platform, it will say well, what EKGs do you need?

And then it'll still show you the last two or three EKTs because it thinks that's probably what you wanna see. But then say, no, I really want to see the EKG from the ER visit on April 23rd of 1996. I'll actually bring up that EKJ right in front of you. So we, we're also not thinking about this as the old word of PDFs, where if you, let's say search for the word EKGE, you have 10,000 references on somebody wanted to do an EK, somebody thought about an EKG, someone built for EK where you just want see EK.

Yeah. And so that's what we mean by the ability to. have a conversation with the data. I think that's critical in modern systems. Yeah, I love that notion of having a Conversation with data because in most of these cases the Data is kind of a thing in and of itself, right? It's and the ability I loved the use of the technology in giving us the Ability to have A conversation with that data. And to your point, it's all so contextual that I love that you've gotten something that really allows people to do that.

It's fascinating. I Love it. That's why it becomes so critical to maintain that context to our earlier point about that if you have data into hundreds of thousands of pages, then you want to make sure that the context exists within the platform. Yes, absolutely. So I want to explore a bit about security and compliance, because I'm guessing many in our audience are probably having those thoughts, especially as you're talking about an example of let's upload a file into chat GPT with how does core metrics approach HIPAA compliance and SOC 2 certification and the other protections that the industry is going to require.

Traceability, Bias, and Conversational Data 29:28

That's a great question, and I think a very relevant one with the advent of AI. There's deluge of the AI systems now, right? Every clinician gets approached maybe a hundred times every day. And it would be a funny part of this system when our customers ask us, why should we trust you? You know what our response is? Please don't. I think you should trust no AI system unless you've done the vetting yourself or unless your confident that somebody external to them has vetted them. For example, if I were to just say, hey, we're built with security in mind.

What does that really mean? I mean, yeah, I built a system. I'm invested in it. We're biased, right? So you want to make sure that there are external auditing and transparency around what those processes are. And the industry standard is that you use neutral third parties who actually serve as auditors. So in the healthcare environment, think Jayco or think CMS, who are neutral, third-parties in a way, so the hospitals could say we provide great care. But you say, yeah, okay, great, but prove it.

Because we need an external auditor. Whether you use J.C.O. or another accrediting body, same thing happens in medical residency programs or training programs. Like the training program says, oh, we do a great job, and ACGME will say oh yeah well let's prove. And the same type of rigor has to be applied to any systems. that any good modern system or technology should be able to say, we are HIPAA compliant, not because we say so, but because some other external auditor says so. And then in the tech space, HIPAA is obviously a healthcare-specific standard, but in a tech-space, that same principle is SOC 2, which is the US standard.

The international standard is ISO, for example, if you've heard of that. And the GDPR is European standard You need somebody else other than yourself or somebody unaffiliated with you who's certified that, and that's what SOCT 2 means. So to me, I think it's just stable stakes. That's the first step in a conversation. If that's not even what discussing in my mind in that sense when we are evaluating these platforms. Wonderful. I love that. You just said something that I find very true. Every health care leader I'm in conversation with right now is getting barraged with requests for, you know, take a look at our new AI.

So there's a lot of hype about that from your perspective. Where are organizations making smart investments in AI and where are they getting distracted by the hype? Yeah, I would broadly categorize that into two categories. Errors of omission and errors of commission. And the biggest one is the errors om omissions, that not believing that this is not coming. I see that as the number one error. And I also have that analogy of that we could be at the side of the ocean and have our head in the sand and think waves are now going to come by.

Yeah. But the waves already here. I'll also get conversations where people say, oh, I don't use AI. Well, I hate to break it to you. You do use AI. I say, well, not in my clinical work. Okay, so your email is coming via snail mail. Listen, no, actually not. Well all of the modern email systems, Microsoft, Google, Apple, your corporate accounts, they're all AI Okay, how do you write? Maybe you're writing on a paper and pen. No, we use Microsoft Word. I hate to break it to you. That's AI. And what do use for your spell check?

Grammarly. What do we do for PDF documents? Adobe. So the point is, it's a threshold. Then no longer is it a binary question. And then the second piece of this, which we already very strongly believe in, is workflows. Because what we're seeing is that any time you adopt a system that doesn't make your life easier, that's a problem.

Security, Compliance, and AI Hype 33:38

These, by definition, should make our lives simpler. And if it's not solving that problem, if you're saying it'll take you six months to get your workforce trained and then it will take another six years to generate some ROI, that's just not an acceptable standard anymore. Yeah. Yeah, the timelines are much, much shorter. There is still a lot of hype out there. But thank you for making the point about just our general use. AI is kind of everywhere. It's in the air you breathe at this point. Well, I have loved this conversation.

One last question to bring us home here. And I would love, from your expertise, if you look ahead three to five years, which is a really long time in this place where we have this innovation happening so quickly. How do you think AI-driven healthcare intelligence platforms like CoreMetrics will reshape healthcare organizations, legal review, and decision-making for physicians and healthcare organization? One, you know, as they say, it's hard to make predictions, especially about the future. Here's where I think that we are headed towards, though.

I mean, if you think about a lot of AI right now, It is still predictive. So think of health data at scale. You say here's 100,000 patients with high blood pressure. And we can predict how many of them will have stroke events, right? So that helps at a public policy level, but does that help the individual patient or an individual community? That becomes much harder. So I predict that AI goes from predictive to more prescriptive, that what are the interventions that you can take to do that? So now apply that same analogy to quality.

Right now, all quality, well, not all, most of quality is retroactive. that events have occurred. And then you're sort of judging on what the impact of those events were, whether the right thing was done or not done. Whereas I would say a modern futuristic three to five year platform is continuously integrated into your decision making model and say there is a quality concern that's going to occur if you do this. Therefore, don't do that. And then the final layer there still remains the human in the loop.

So a surgeon can operate. But think of the system, perhaps, that says, hey, if you use the scalpel, then this is likely to occur. That's fascinating. And that has huge implications on our medical education systems and all of it. Fascinating. I cannot thank you enough for being with me today. This has been a fascinating conversation. Thank you again. Thank you so much for inviting me, and this has been such a pleasure. And I'm just impressed by your depth in this field and these questions. Oh, thank you.

AI is coming, whether we're ready for it or not, right? So. Wonderful. Thank You. Thanks very much.

The Future of AI in Healthcare Intelligence 36:38

Before we wrap up, if you're listening to this and thinking, I know I need to understand the business side of medicine better, but I don't even know where to start. I want to share something we've built specifically for you. It's called Physician's Edge. Physician's Edge is designed to give you the practical business, financial and leadership skills that most of us were never taught but are absolutely expected to have. This is not theory and it's not overwhelming. It's a structured, physician-friendly way to learn how your practice really works, how money flows, and how to make better decisions with confidence.

Its online and on-demand so it fits your schedule. And just as important, it's not something you have to do alone. Physicians Edge is also a community of physicians and leaders who are asking the same questions, solving similar challenges, and supporting each other along the way. We host a monthly live journal club session so you can pop in if you'd like to ask your questions live. Because one of the biggest risks in medicine today is trying to figure all of this out in isolation. The goal is simple.

More clarity, more control, and less stress when it comes to the business of your practice. So if you're ready to build that foundation and step more fully into the leadership role you are already in, I'd encourage you to learn more about Physician's Edge. You can find details and join us through the link in the show notes or at healtheps.com.