Listen and followApple PodcastsSpotifyYouTubeMore apps · Browse all episodes
This episode is sponsored by Oracle Health and Life Sciences.
Doctors have heard promises about health technology for decades. Now artificial intelligence has to deliver something better than what health care already has. Seema Verma, executive vice president and general manager of Oracle Health and Life Sciences and former administrator of the Centers for Medicare and Medicaid Services, explains where health systems stand with AI today. You will hear why ambient listening became the proof point physicians did not want taken away, and why bolt-on AI tools on older electronic health records could create more burden. Seema lays out the questions to ask before you trust an AI tool, including how it was trained, how it explains its recommendations, and whether a human stays in the loop. She also explains how AI could make value-based care work and which metrics show whether it is paying off. This episode is sponsored by Oracle Health. Find out more at oracle.com/health. Press play to hear where she hopes AI will raise the quality of care and cut administrative cost, and why it only works with the right data.
VISIT SPONSOR → https://www.oracle.com/health/
SUBSCRIBE TO THE PODCAST → https://www.kevinmd.com/podcast
Transcript
Kevin Pho: Hi, and welcome to the show where we share the stories of the many who intersect with our health care system but are rarely heard from. Subscribe at kevinmd.com/podcast. Today we have a special sponsored episode brought to you by Oracle Health. Today we welcome Seema Verma. She’s the executive vice president and general manager of Oracle Health and Life Sciences and former administrator of the Centers for Medicare and Medicaid Services.
At Oracle, she leads the clinical and clinical trials application portfolio, and at CMS she led the federal push on value-based care, interoperability, and price transparency, including the Patients Over Paperwork initiative. Today, Seema joins us to discuss what health systems are actually trying to accomplish with AI right now, and what has to be true before a physician trusts it in the exam room.
Seema, welcome to the show.
Seema Verma: It’s great to be here. Thanks for having me.
Kevin Pho: So tell us about yourself and your path in health care, and what led you from health policy to building the technology behind it.
Seema Verma: Well, I kind of think of myself as a problem solver. The early days of my career were spent working directly with providers on the front lines in a safety-net institution, so you see a lot of the problems in health care, certainly in a place like that. It gave me a lot of perspective and background.
I’d also spent some time in Washington on policy issues, and so I started my own consulting business and was working with states and a lot of private sector clients as well, technology companies around health care, but more so on the policy side, and trying to figure out what government can do to help the situation.
My last gig in government was being the CMS administrator, which was an exciting opportunity where you’re really exposed to pretty much every health care problem you can imagine. So throughout that journey of the different roles and the different places that I’ve been, it’s always been kind of the problem solver. What are we going to do? I don’t want to just run something, I want to fix it.
And so after concluding my time in government, I was thinking about what comes next, and I spent some time with private equity and investors, which was also a great experience. But what I love about tech is that I think this is the place where these problems, these very long-term problems that we’ve had, that we’ve been unable to solve, are now solvable in the advent of AI. So it’s not a panacea. It’s not going to fix everything, but it’s going to have a dramatic impact, just like a lot of other technology has had on mankind.
Kevin Pho: AI is, of course, a very popular topic on KevinMD and on this podcast. I have many, many guests who talk, through their own lens, about AI’s impact on health care. So from what you’re seeing, what are health systems actually trying to achieve with AI today?
Seema Verma: Sure. Well, I think it’s early days when it comes to AI, and I think health systems and providers, everyone, we’re just dipping our foot in the water here, right? I mean, it’s early. And so there’s a lot of bolt-on solutions. We’re trying it for this and for that. We’re becoming comfortable. We’re becoming skeptical. And I think we’re starting to now look holistically at the potential of AI. Right now, I think it hasn’t been as much agentic as it’s been more generative AI, so patient summaries, those types of things. But the future is that we’re automating a lot more of the manual work that we do today.
So like I said, early innings, people are experimenting, whether it’s rev cycle, whether it’s the ability to look up some information, OpenEvidence, a lot of great companies doing that type of work. Ambient listening, we’ve heard about those. But when you think about them, they’re all very much point solutions. I’m going to fix this problem, fixing that problem. And I think that administrators are starting to say, “Well, wait a minute. How is this all going to work, and how is it all going to come together?”
Kevin Pho: It’s incredible that you say we’re in the early innings of AI, considering how far AI has gone, right? Because AI has been here for probably the last three to four years, and when you think about how advanced these frontier models are, in the big picture, like you said, we’re still relatively early in the advent of AI. And I have a lot of physicians I talk to on the podcast, and they’ve all heard promises about health technology for decades, obviously with the electronic health record being a big example of that. What feels different about AI, and what evidence are health systems looking for before they fully trust it?
Seema Verma: Yeah. So I think that’s a fair point, that technology has really failed the health care industry in many ways, and in some other ways, in biotech, it’s been very helpful. But you’re right. If we think about the EHR, I think for physicians it actually created more problems maybe than it solved, right? It kind of turned them into data entry clerks, people who had spent years in medical school and lots of training, and they’re sitting in front of a computer not looking at their patients. And it really became a system of record, it’s not a system of intelligence. And I think that’s where we’re trying to get to, is that system of intelligence where it’s actually useful.
We’ve seen some early success when we think about ambient listening. Boy, you give that to doctors and it’s like, “Oh, this can’t be a pilot. Don’t take this away from me.” It’s pretty cheap. So I think when you see that type of proof point, there’s excitement of, wow, if it can do this, what else can it potentially do?
So ambient, we’re starting with that. Now you’re hearing about things like autonomous coding. If it’s listening to the conversation, can it generate the codes? OK, well now that it’s generated the codes and the note, can it do my prior authorization requests? So more and more is going to get automated. The concern I have is, OK, there’s one thing to do this and one thing to do that, and how does it really all come together in a comprehensive system?
Kevin Pho: So like you said, a lot of these tools are a little bit piecemeal, although some of the major electronic health record companies are integrating AI within their workflow. So hopefully in the future it would fall under fewer umbrellas. Now, you did mention some of the concerns that physicians have about accuracy, liability, transparency. So when it comes to health systems and physicians, what does responsible adoption look like to you?
Seema Verma: Yeah. So I think it starts with understanding how AI works, right? And AI is only as good as the data that it has. So first of all, understanding what model’s being used. There’s lots of different models out there, and you want to understand what was the background, what was it trained on, how many records, how current they are. But the value of AI is being able to take those trained models and also have access to real-time data, because otherwise you’re making decisions in a vacuum.
Think about it. It’s artificial intelligence, so it’s not unlike our own minds today, right? As a physician, you’re making decisions based off the data that you have available. So if something’s missing, it’s going to alter your course of treatment and your diagnoses. And it’s the same thing for AI. The problem with AI, though, is it doesn’t always know what question to potentially ask.
And you mentioned just a second ago that a lot of the companies are trying to embed it, and this is an important point out there. So think about it. You’re in an old car and you’re trying to get a GPS into it. So you’ve got to bolt it on, you’ve got to plug it in. That’s versus something of the newer models where it’s all built in and it works more seamlessly with the entire operating system. So that’s a little bit of what we might face in the short term, these bolt-ons, and at some point it just becomes unsustainable because you’ve got 10 bolt-ons. How do they work together? And what about cybersecurity every time you’re bolting on, every time you’re moving information from one place to the other?
And that’s why Oracle’s investing in bringing a new EHR that is built for AI and to use AI. So at the infrastructure level, our system knows how to retrieve data in a real-time way. The older models of EHRs are systems of record. They’re not built to retrieve information and bring it to a model. That’s just not what they were built to do. So in terms of building trust, you’ve got to really think about how this whole thing is working and whether it’s built for success.
I think for physicians, there’s also another layer around human in the loop. We don’t want the AI to be making decisions. It can bring information, it can make recommendations, but the AI should not be the decision-maker, and that is where we feel like humans are important.
And again, depending on the product you buy, you want to ask questions. Well, what’s the transparency? So if the AI is telling you, “Here’s a recommendation,” well, how did it get to the recommendation? The traceability, the explainability, the transparency around the decision is important. So if it says, “You shouldn’t prescribe this medication because your patient is a diabetic,” and it’s, “No, he’s not.” “Well, why?” “Well, there was this hemoglobin A1C score lab result 10 years ago.” Something like that. You’ve got to be able to know why it’s telling you what it’s telling you.
The other thing you have to think about is the model monitoring. It’s not that we put these things out there and then we don’t see how they’re working. You have to continually monitor. So I think there’s a lot of questions for providers to ask on the front end when they’re using these AI tools, right? How is the model trained? How is it monitored? What kind of testing was done, and what kind of data is it using to make these decisions? What does it have access to?
So if you think about it, let’s say it’s prior authorization. Well, this one’s going to put your prior authorization request together. OK, great. But where’s it getting the data from? Is it pulling it from the electronic health record, or is it just whatever the call was about, right? So there are some nuances, and understanding, number one, how it’s built will allow you to figure out what’s the best buying decision around AI.
Kevin Pho: Now, when you were at CMS, obviously electronic health records were at the forefront. And when I talk to physicians today, they tell me that although documentation has improved over the years, and now with ambient AI it’s still improving, they feel like they’re spending too much time within those electronic health records. What’s still missing today in 2026?
Seema Verma: There’s a lot missing, because they’re still working off of very old, antiquated systems. And so even though we’re doing notes, if we think about agentic AI, which is that it can do the work, it’s listening to the conversation and it heard, OK, this is the drug that we’re going to prescribe. Why can’t it just go do all of that work and write the script and make sure the drug is available, all of those different things? And this is where, getting back to that old technology, we’re going to be hampered in really taking AI to the next level and having it do more of the work, right? The coding, the billing, all of those different things, if the AI does not have access to real-time data.
If you think about health care today, a lot of our problems are data problems, right? We don’t have enough data. We have too much data, too little data. We don’t have the right data at the right time. And we have a lot of people that that’s kind of what they do. They’re bringing data together. Whether it’s rev cycle, whether it’s a nurse, whether it’s a physician, they’re trying to track. So AI can help solve that type of problem. But it’s not happening today because right now we’re bolting it on into these older systems. And it can do one task, and then another AI agent can do another task. But how do you piece it all together? How do you bring the entire thing together?
You have to start with bringing all of your data together. It has to be in an AI platform that can bring all of the data together and that can retrieve it in real time. You can’t do that with something that was designed in the 1990s. So we’re still not getting the value of AI, and I think the next generation of EHRs that are built for AI with AI will be able to deliver a lot more automation. But we’re not there yet.
Kevin Pho: So we have a lot of physicians, of course, listening to you and on this podcast, and if they ask you, “How is AI actually helping me tomorrow morning when I walk into the clinic?” what’s your response to that question?
Seema Verma: I think that some of the basic tools, the ambient listening, the coding, the patient summarization, those are some very useful tools, right, that are able to help them go quicker and faster. The data shows that when they’re using those tools, they’re seeing more patients, they’re spending more time, they’re having a better experience, right? Nobody went to medical school to do data entry, so they’re having a better experience and are able to look their patient in the eye.
I think the issue now is they want more and more automation, right? We kind of whet the appetite of, wow, you don’t have to do this. Now it’s like, well, can it help me do all these other things that are administratively burdensome? I think the other thing AI’s been very effective at is bringing them more up-to-date information, right? No pun intended. But just to know what’s the latest treatment that’s going on, what’s the latest clinical trial, even that’s been super helpful.
What’s going to be more exciting is when that tool is built directly in, so it has all of the patient’s medical record and it’s working alongside and with and supporting, instead of the doctor having to bring this question and that question and back and forth and back and forth. Imagine a world where it’s all embedded and it just works alongside them inside the workflow. That’s sort of the future vision, and I think a lot of the older technology isn’t going to be able to deliver that in that seamless way.
Kevin Pho: You talked earlier about that distinction between AI built into the clinical workflow and AI layered on top of an existing system. And I think a lot of the decision-makers in health systems are already facing that fork in the road, and it’s already overwhelming, the number of options and the number of paths that they can take. Go into a little bit more detail about how and why that distinction matters both to physicians and health systems.
Seema Verma: Right. Well, first and foremost, you could spend a lot of money on this. You can buy 20 different, 30 different, more and more, all these different AI agents and different tools that are doing various tasks and work. But at some point, they all need to be coordinated. And they all need to have access to enterprise data.
So think about it in terms of the revenue cycle. It’s a great example, where you’ve got the front office collecting the insurance information, the back office, the middle office. In order for rev cycle to work, you’ve got to have access to all of that information. So putting one rev cycle agent to do your coding is one thing. Then there’s another one to submit the bill. But let’s say you’ve got a claim that was denied. Well, you’ve got to go back and look at the EHR data to be able to justify whatever the claim was.
So the first thing is getting all your data in the right place. You’ve got to have all of the data together so I can access anything. So let’s say we’re scheduling a surgery. A doctor wants to schedule a surgery, and we think about everything that’s involved with scheduling a surgery. Do we have the right supplies? Do we have the right staff on board? All of those things you want to know when you’re making that decision. Well, the reality is the supply chain data, the HR data, the approval from the insurance company, all of those are sitting in different systems. So in order for AI to work, you’ve got to bring it all together so that when it does that task, it can do it to completion across the entire task.
That’s instead of how it works today, where there’s a bunch of humans, right? Somebody has a surgery scheduled, somebody calls you to schedule it. Then somebody else calls you to say, “Oh, did you have your ECG? Do you have all the labs that you need before your surgery?” And it’s all a very disjointed process. It’s a very human trade-off. The reality is AI can do all of it. It can even make the phone call and do the assessment and go over, “Here’s all the things you need to do to prepare for your surgery.” But it’s not going to work if it’s piecemeal.
So the first thing is bringing all of that data together and thinking about an enterprise approach. For a lot of the software, you’re going to see a lot of the software that you already have today with people bringing in more and more solutions. I just don’t know that it’s comprehensive, right? Can it work across multiple systems? So it’s great if your EHR’s got some AI in it, but as a system, you don’t just have an EHR. You have an ERP, you have supply chain, you have HR, rev cycle. So you really need something that works all together, if that makes sense.
Kevin Pho: Now, what kind of metrics should health systems be using to determine whether AI is creating value?
Seema Verma: I think the metrics are very, very important, because let’s face it, health care’s been burned by tech in the past, right? So we’ve got to make sure that it works. And I think this is where the buyers need to say, “Well, tell me what the ROI is.”
And there’s lots of different ROIs, right? There’s what we call short-term or leading indicators, and lagging indicators. The leading indicator may be, wow, your providers are delighted, they’re happier, they’re excited. So retention of your physicians, but does it lead to retention? So they’re happy in the short term. Great, satisfaction’s high, but are they staying longer? Are they seeing more patients?
Some of the tooling, even just with ambient listening, which is not intended to be a billing tool, what we found is that it actually improves accuracy of coding. So just by having the better documentation, it leads to better billing. And so you actually see an uplift in revenue.
So I think from the very beginning, you’ve got to think about what those metrics are and how I’m going to measure it. And I think it’s incumbent on the tech companies to prove it out before you make that investment. This is what you can see. This is what the measurement shows. There’s so much data, as you can imagine, anything from just satisfaction to revenue, to quality and to efficiency. So there’s a lot of parameters, and they definitely should be measured before, during, and after.
Kevin Pho: Now, going back to what we talked about in terms of health systems adopting these AI tools and layering them on current systems, sometimes adopting them in a more piecemeal approach, and you have eventually multiple systems, multiple applications, and increasingly, of course, multiple AI tools. What do you think needs to happen for these technologies to work together rather than create more complexity?
Seema Verma: Well, this is going to be the new generation, right? This is what we’re building at Oracle, which is a brand-new EHR where all the AI is embedded. So you’re not having to buy 10 different point solutions, and it all works together. It sits on a platform. That means we’re going to bring all the data together from other parts of your system, and this allows you to build your own AI agents, right? It allows you to use other products that are out there. Maybe you’ve got a core EHR that’s built for AI, but you want to use something else, OK? So it’s got to be able to easily bring in other things. So that’s what the new generation of tooling is going to look like.
Kevin Pho: So you’ll soon be talking with health system leaders at the Oracle Health and Life Sciences Summit. What do you think is the question everyone is trying to answer right now? And for physicians listening who may not be able to attend that summit, what lessons do you think will be most relevant to their day-to-day practice?
Seema Verma: We look at it as, our theme this year is vision into reality, because we’ve been talking a lot about AI, and we want to give the reality check of here it is, right? We’re bringing lots of new products to the market, so they’ll get an opportunity to see it.
But the conference is more than that. This is a global conference. It’s not just physicians and health systems. It’s also life sciences companies. Big pharma companies are coming. Big payers are coming. So it’s really the intersection of the entire health care system coming together, and thinking about AI and what impact it’s going to have.
And there’s a lot of questions that we all have to answer that go beyond the new gadget or the new tool that’s coming out, right? It’s governance. It’s ethics. It’s policy. It’s regulation. It’s funding. And so we hope that this is a learning conference for people to see the tools, but also thought-provoking. Well, what is this going to mean to my business process? How do I need to change things? What kind of buying decisions?
We have regulators coming to this meeting as well, because they’re rethinking how AI should be regulated. Or things like interoperability. Well, AI’s only going to work if it has the right data, the right information at the right time. We need interoperability, so we have a lot of conversations about interoperability, why that’s so foundational, why it’s important, and how we build it so that the AI can work more effectively. So I would say that it’s a health care conference for payers and providers in life sciences, and it’s really about the transformation of health care that we want to deliver and everything that it’s going to take to do that.
Kevin Pho: Now we’ve talked about the clinician experience, workflow, accountability, and value-based care. As health systems move from experimentation to implementation, where do you see the greatest opportunity for transformation?
Seema Verma: I think the transformation is, and you said it earlier, what is it going to deliver? If AI’s not going to deliver something better than what we already have, then it’s not worth doing and it’s not worth investing in.
So I’m hoping that it does a few things. Number one, that it increases the quality of care for patients, that somebody in a rural area is going to get the same level as going to a big, large academic institution, so that the providers all over the world understand what the best treatment is. That they can ask the AI a question, they can present their patient, and instead of having to do it in a formal kind of conference, that you’re able to exchange ideas and get input right away. You can say, “Tell me about patients like mine,” and compare it and get better insights and better clinical decision support on the front lines to raise the quality of care.
The other thing I would say is efficiency. We spend, doctors, nurses, administrators, we spend way too much time and money on administrative work, a lot of paper pushing going around. How do we reduce that? How do we take, instead of having people manually reporting quality metrics, how do we automate a lot of that so that we can spend more time on patients? So reducing our costs on administration. And I hope that that helps us financially, because I think we’re at a point now where health care is costing everybody more. It’s getting more costly to deliver. We’re seeing impacts on premiums. You’re seeing governments that are having trouble affording it, and we’re having to make really tough choices. So I hope that it drives a level of efficiency that creates more resources for actual patient care instead of administrative costs, so really trying to strengthen the financial stability of the entire system.
I also hope that it leads to better cures and treatments, and that it accelerates the pace of innovation. It takes so long to bring a drug to market. It’s hard to find people to participate in clinical trials. It’s hard to understand, even once the drug has been approved, what are the other indications? What are the safety concerns? I think AI can play a dramatic role in automating a lot of this and making all of this go faster and quicker, so that we have better quality of life across the entire population.
Kevin Pho: We’re talking to Seema Verma, the executive vice president and general manager of Oracle Health and Life Sciences, and the former administrator of the Centers for Medicare and Medicaid Services. We’ve been discussing what health systems are actually trying to accomplish with AI now and what has to be true before a physician trusts it in the exam room. Find out more at oracle.com/health. That’s oracle.com/health. The link will be in the show notes. And Seema, my last question always is to share your take-home messages with the KevinMD audience.
Seema Verma: Well, I’d say I’m excited about AI. I know that there’s a lot of skepticism, but we have to remember that this is early days, early phases. We need everybody to come together and help make these solutions stronger and better for our patients and for the experience. I have a lot of respect for physicians, who spend probably more than anyone does in terms of training and the time that they take, and it’s a vital, important profession, and I hope that AI’s going to make their experience, as well as the experience for their patients, better. So thanks for having me.
Kevin Pho: Seema, thank you so much for sharing your perspective and insight, and thanks for coming on the show.
Seema Verma: My pleasure. Thank you, Kevin.
Listen and followApple PodcastsSpotifyYouTubeMore apps · Browse all episodes
