Fireside chat · 4:00 to 4:25 PM · Thursday, May 14, 2026
So before I introduce Julie, I need your help. Raise your hand if any one of you in this room has directly influenced an AI decision in the last 12 months. All right. Keep your hand up if you can describe the outcomes in 30 seconds or less. Please. Thank you. So I am pleased and it's a privilege of HFS to have Julie Durham here with us. She is the CIO of Optum Insight and Applied AI to UnitedHealth. For those of you who do not know about UnitedHealth, they are a Fortune 3 company, the world's largest healthcare company. So a bellwether in so many ways, not only just healthcare, but all of the technology that they're enabling in order to facilitate better health outcomes. So Julie, welcome.
Thank you.
So we've got a bunch of questions, but we'll try and engage with you folks as we go along. But I really wanna ask Julie about the AI journey at UnitedHealth Group, and what were the core problems you were initially trying to solve for?
Yeah, one, just thank you, I'm so excited for this. I think one of the things I love about being at UnitedHealth Group is whether you use a UnitedHealth Group's services and capabilities or not, all of us at one point will come to rely on the US health system. So it's easy for everybody to relate to it, and it's easy for me every day to see where I can make an impact on that. So I think there is something so unique about being in healthcare that I think we try so hard at United to share our stories so we can learn from everybody else. Before I jump into that, maybe I'll just kind of give a context of how UHG is structured. I can tell you being at UHG for the last nine years, sometimes it's hard for me to kind of understand so I can only imagine how it is for you. But if you think about healthcare, you sort of put that into two parts. You've got care and you've got coverage. So coverage meaning someone has to pay for it, and that's our health benefits side of the business. Most of you will know that as UnitedHealthcare, so our health benefits. And we go to market as either employer benefits, so your employer will pick your health benefit provider for you, and then we also do Medicare and Medicaid. So think of this as our 65 and older. And then we have a ton of other programs that we've created across UHC in the last 10 years beyond just your health benefits to help you in your whole holistic journey. So everything from food benefits, travel benefits, across the board. So that's UHC, so that's the coverage side. And we're the largest insurer in the U.S. by far there. So roughly about one in 10 people in the United States will get their care through UnitedHealthcare.
On the other side, if we think of care, actually using the U.S. health system, think of that as Optum. So as we initially started as a health benefits company, over the last 10 years, we've been building out our Optum capabilities because we feel strongly that an integrated health system is more productive than those that are fragmented. And I think you can all see that too. So if I break Optum into some categories, one Optum brand you might know the most is our Optum Pharmacy Benefits Management. So Optum RX, everybody's nodding. If you've looked at your insurance card, you have health benefits, and then you have your PBM business. So that's our PBM business, Optum RX. We're not only are we doing RX benefits, but we also have pharmacies and clinicians. Next, we have Optum Health, where what we've been doing over the last five or seven years is out acquiring care delivery organizations across the United States. So in a lot of cases, you actually might be at an Optum clinic, even though it's not branded Optum. And we're at the point now where almost one in eight physicians in the U.S. is either employed by OptumCare or contracted with them. So there's a high probability that you're engaging with Optum there. We also have a bank that we call Optum Financial. So you probably know this as a consumer as your HSA or your FSA, so your flexible spending accounts. But in addition to that, we also provide working capital for providers. So think about a long-running engagement you might have. That provider's going to need capital. That's also where a lot of our Change Healthcare business sits. You might know that from a few years ago.
And then the last major business unit is the one that I'm in called Optum Insight. And if you think about what Optum Insight is, think of it as taking the best we've seen across Optum and UHG and being able to put those together in commercially their software or services and provide those out to any payer or any provider. So even if you are not part of UHG, we have parts of the UHG ecosystem that likely could help you. And that is my best summary of what is United Health Group. We're about $430 billion. We're about 450,000 employees globally. A little over 150,000 of those are not within the US. And our Optum Tech is about 40,000 engineers. I lead about a fourth of those in my Optum Insight role. And I also lead Applied AI, which I would love to hear two people describe Applied AI the same way so happy to kind of frame that. So our chief AI scientist Michael Pencina reports into me and I also have our shared AI platforms as well. We'll tell you I feel strongly no one person can own AI, like it's impossible. So even though I have a little AI in my title, I always joke it's like saying I'm going to own math. I don't know how you're going to own math, but I do know at this moment we just needed a little more coordination across those topics. Okay, that was probably a longer intro, but I thought that that framing at least is helpful for kind of where we've been on the UHG journey.
And, you know, AI is not new to UHG or healthcare. We've been using AI for some time, but I think what is different is this massive step change in both technology of what the models are able to do, the accessibility of how well we can interact with them, and now the opportunity where we're seeing the scale at a pace I don't think any of us were expecting. I think when we started, we sort of did the natural thing and put it into three kind of buckets. We said, okay, we're going to focus on claims because there's a lot of claims. So how do we automate that? That's a huge part of waste and not new to us, but certainly an area we knew really well. We said we were going to focus on consumer. And actually prior to this role, I was our chief digital officer for the last three years. So that was a huge part of my agenda where how did we start to kind of bring healthcare into the digital age, lack of a better word, and AI being a huge boost for that. And then the third bucket we had was care. And that was really our Optum care delivery capabilities as well. And I think it's natural for all of us to first take the processes you know and make them go faster. And that was probably a lot of our first stage AI investments where we started to put them into those particular buckets.
So the way you describe UnitedHealth Group and all of the different components of that, clearly a very complex business, which I would argue would also raise a variety of complex challenges. So when you think back a couple of years maybe now, what kept you awake at night when you knew you had so many of these different challenges to solve? How did you pick the one that you wanted to solve, and where are we in that journey?
Well, I will tell you, I think we still struggle with the right set of prioritization capabilities where there are so many opportunities in health care that you see every day. I'm going to go to when was the first time I think I moved from, yeah, yeah, yeah, we've been doing AI a long time to this feels different to me. So I had spent about three years helping us build out our digital capabilities on the UHC side first before I became the CTO. And we did a ton of investment into our UHC mobile app. And we had to even change status quo where I remember that one of the CEOs was like, hey, our M&R population will never even use the web. And actually, they were the fastest adopters. And same thing with CNS of, hey, our community state, our Medicaid, our most vulnerable population, they won't use this either. And actually, they're likely to only have one digital device, and that's their phone. So they were high adopters as well. So it was like the start of this challenging status quo of how consumers do want to interact digitally within health care. And we had a huge boost for what we did. Normal things you would think of, make a digital ID card available offline, just like you can see your boarding pass. So we actually even put into some of our plans saying, hey, we'll give you a discount if we no longer ship a physical card, right? So some of these things, like these building blocks had to be there.
And we had just reached number one on the mobile app store. We had just done this huge redesign of something called an explanation of benefits. So you guys get a lot of mail from the U.S. health system. In fact, UHG is the largest user of the U.S. health system of everything we have to push out, where we had done this beautiful explanation of benefits that made it easy for you understand what you had and we were really proud of that and we had digitized it and put it on the app. I was in India and we were doing a demo with the team and I said well you've got this really rich document why can't we use an LLM to ask it questions and interrogate it right and put a RAG around it so it's not going to hallucinate and see what that looks like. And I realized at that moment that we had just gone from painting screens to forever living in a conversation, because of how quickly and easy it was for that experience to exist. How many other questions people wanted to ask that document was a huge trigger for me that we were kind of on to something to the new step change in this technology. So it's like these little sparks that tell you like, wait a minute, everything's about to change differently. And then from there, we actually rerouted all of our investments into moving into conversational AI. So investing into chat, investing into our voice channels, partnering with some of the best frontier model companies, and securely and safely trying to push into that paradigm.
And I know it might seem fairly obvious, but remember a couple things. One, I operate in the most regulated environment across everyone. I got an answer to a bank. I got an answer to medical devices. We take this with care. There was also this hypothesis of, would people even want to interact with a voice bot in an LLM about their care? And I think the next early signal that started to keep me up at night it was about to be different, was seeing the types of honest questions people were now having in that digital channel. You've probably all been there, you go see your PCP and it's like, well how long has this been bothering? And you're like, oh just today, you're like, no. So I think like that was probably the kept us up at night where we started to see that you as consumers were sending us signals we were missing and you desperately also wanna simplify healthcare with AI. I think we have that moment where we have the right assets, the right people, the right rigor across it to truly change some of our own status quo. So over the course of 12 months ago of being in front of BT at the time, Brian Thompson, our UnitedHealthcare CEO, talking all about the app to basically kind of killing my favorite things so I could go chase what was next was probably where we all realized that I was up at night realizing there was a future that was gonna come next that was gonna be pretty exciting.
And where you are now, I would argue that you must be feeling pretty proud about the journey and the outcomes that you're beginning to deliver. Can you talk a little bit about what those outcomes look like and how is it being manifested on the ground, especially from a membership standpoint?
Totally, and I know that's just like one use case, but I think for us, we sort of started to see this new kind of like, as opposed to saying, yeah, AI is for other people or this won't work, to just how much it was gonna work. So maybe I can give you kind of just breadth and depth of just across UHG of how we're trying to manage this because I think there's sort of three healthy kind of transitions of AI now we're starting to see. By the way, yesterday we did our monthly AI MBR with Stephen Hemsley, who's our CEO. And I will tell you, while I feel like I am proud of the progress we are making, we are in no way satisfied, nor is he. Which is natural of anyone who has a CEO in here. We know there is so much more that we should be doing that we're not doing, and how do we continue to challenge ourselves?
So I think there is opportunity number one, which I think is new and emerging, 'cause a year ago we would have been just talking about the best model running in our software. Think of that chat bot I mentioned and what it looked like versus really what does this look like for empowering our employees? And I think that started to change as Claude Code and Cowork and Codex started to come out. So there's sort of like bucket number one, how do we make sure that we empower all of our employees? And to date, most of the employees that had access to AI were either sitting in the C-suite like me or working on an AI project in IT. And I think this kind of democratized access to AI where every night nurse and every claim operator eventually will start to have access to these frontier tools is incredibly encouraging. So we're watching and encouraging that because we fundamentally believe that they will be able to help us do the second part, which is where a lot of our investments were, which is how do you kind of do faster processes? That makes sense, right? Like the reality is you do have to automate the world as you know it so we can do the next part, which is to start to really get into how do we redefine healthcare. And honestly, I think we were naturally in that middle bucket for a while that said, okay, how do we do faster prior authorizations? Not even faster, maybe eliminate some of them, which you can see we just did. How do we move into the contact center? So if you do pick up the phone and talk to us, there's voice bots there. And how do we optimize that with a new way of working where all of our employees now have the ability to create not only their own agentic workforce or worker, contribute to that, and then start to actually reimagine healthcare?
So I didn't imagine that that first bucket was going to be as big as I thought. I'm sure those of you that are seeing what these tools can do is realizing that's a whole other paradigm of unlock and access that I think we're just starting to scratch the surface on. And if you look at some early reads on these tools, first off, they're pricey, so watch them. That's first lesson, is uncontrolled token utilization in your employee base is both amazing to see what they can do and another paradigm you have to watch. But man, can you imagine if every employee at UHG was twice as productive? Like, how great would that be? Think about how hard we all work for those 15 minutes with that PCP and how much that they're just doing this or have someone else doing it. I think that is starting to give us a little bit of new creativity. We spend a lot of time having some discussions about, should we even bother automating or putting AI into the current processes to speed them up, or should we just kind of jump to this next frontier, right? So here's my best advice on at least how I'm trying to sleep at night with that is, you know, let AI run healthcare in the way we know it today so we can create the next one. Because if you can reconcile with that and say, hey, you are going to need to understand your current processes deeply before you can move to that next step, I think that's some of the advice that we're starting to give ourselves.
So, you know, when I look at the audience that we have in this room, it's a cross-functional, cross-industry audience. So if you were to strip out the healthcare pieces that you spoke to, what kind of advice would you give many of these folks who are in different levels of transition of applying AI and doing interesting things?
Yeah. I'm going to give you some of the advice I have to give myself, too, is it is early. I think that everybody is hoping that there is some sort of trend line or playbook of what this looks like. I think there's also this fear that regardless of where you are on it, that you're falling behind and that you're not able to catch up. I don't think that's true. I think at the moment that you start to be curious about it and closing the gaps, that you can not only catch up, but everybody needs to have their moment where they start to see AI differently and go from kind of fear to empowered. My other advice for you would be, actually I just was in New York last week with Anthropic with their head of applied AI and his description of his job was way different than mine, which is good actually. It's like where do we see the end of this productivity boom we're in? And how big is it gonna be? And how many tokens is that gonna take or whatever we wanna use as our unit of measure? And he said, you know, this is probably the first time where we just haven't seen enough continual data points to really know that arc.
So my advice is a little bit of like, this isn't going to move like other things. Like all of us are probably executives in here who need to create a CFO defensible business case. So what I try to do is to say, look, there are things that are moving maybe too fast that we believe in, and let's call it that exponential productivity gain of that bucket one. There are enough data points, I think, in the second part to say automating your current processes should come with some level of consistency and no regret move. And then you've got to watch for those signals in your either employees or your step change that start to fundamentally change your business. We'll also give you one last bit of advice. I just was talking to one of my CEOs. I know that we all have to do a cost justification of any investment we do, since whatever technology is. I think we're all underrepresenting the productivity gain and the boost as opposed to just the cost takeout, that new market entry, that new idea, the ability to iterate really fast through a market pitch, I think is the other lever to it. I don't know if that was helpful. I'm just telling you what's on my mind here. Maybe we get to questions and you can see if I've got a better way to refine that.
Let's engage the crowd a little bit. So can you raise your hand if you can describe in 30 seconds if you're doing something using AI within your company that is actually improving one of your colleagues' life? 30 seconds. Anybody? Go ahead, Russ.
Yeah, we've used a common skill across go-to-market, so we've standardized how we do call plans, but we force people to ask one question because we don't want this to just be AI dumb, and that is, what is your objective? What is your end in mind? So that's standardized, especially for new employees, to be consistent.
Thank you. Anybody else?
One more.
I'm going to take your question slightly differently. We helped a client change the basis of competition in their industry, and they've dramatically improved market share and EBITDA.
Pretty good. All right. All right, Julie, another question. So you spoke a little bit about lessons that you've learned, but let's sort of dig deeper into that. Key lessons that have emerged along the way, and where do you expect yourself to be a year from today? And how do you think healthcare will be meaningfully different as a result?
Ooh. I'm going to go with the pace to which this is moving, just kind of predicting. So if I go back a year ago, we were just looking at, so we run a responsible AI review board across it, not only from a compliance perspective, but any kind of new idea that's going to use Gen AI, how do we not only vet that across clinicians, everybody else to make sure it has responsible use and accuracy, et cetera. So the number of AI projects we did last year in just Q1 has gone up by 300% just through that board. So those actually have to be new ideas that we haven't vetted. So that's a tremendous amount of signals of just how many new creativity ideas are coming through. The second thing I think that will start to happen and I don't know if this is a year from now but I do think it's in the not too distant future, you heard me talk about prior auth, and if you've ever tried to use the U.S. health system and waited up all night to know if your procedure is going to get approved or not, I think is another step forward that's going to happen in the next year. And the reason for that is I'm watching how good the annotation services are happening while you're with your clinician, so you can see, do you mind if I transcribe this, right? So that quality of that data, that assistant with them to make sure that when they code it, it's right and you have the right procedure. I'm watching how well some of the data is getting digitized, for lack of a better word, across the payers. And that friction in prior auth will take some time, but I think we'll be at a point where we almost don't even know what prior auth is. Like many of you don't know that Visa's processing your credit card payments in the background. You just see it work. Doesn't mean that all that went away. I think like some of that type of friction, which is hard to see, will start to change dramatically. And I think AI will play a huge part of that.
That's encouraging. So I'm looking at the clock. Why don't we do a couple of rapid fire questions for you? Okay. And then if we have time left, we can go back to the audience for some Q&A. So one AI tool you personally use every day that isn't ChatGPT.
Ooh, okay. So right now, I went heavy into Claude. Not in my personal life, or this is in my personal life, and here's one last bit of advice. So I got really into Claude. I'm originally from Wisconsin, and for some reason I was into, like, cheese in France, so I named her Claudette. I also have my partner, Sandy. We've been together 15 years. I got her a Claude subscription, too, so Claudette not only is my personal AI but also does a bunch of stuff across the house, bugs the kids, et cetera. So not only do I use Claudette every day, but she does too. And I'm realizing now too where if I'm ever doing something with Claudette, I get the like, you talking to Claudette kind of like vibe from her. So if I could go back, I would have named it a different persona. Because if I named it Ralph, like the number of eye rolls in my life would be way, way lower. So, but I'm definitely right now very bullish, just personal side. I think on the enterprise side, all of these frontier model companies are highly capable, so don't take that as Julie's one or the other, but I think I'm watching how much and how easy it is to have, whether it's managing our kids or our lives and use that in a much more meaningful way. So that's my favorite other AI tool right now.
Let me ask you one more. In 2027, leaders who win with AI will be the ones who?
This is going to sound a little bit corny, but be adaptable, right? Like find a little bit to enjoy the ride. I had to throw out a whole bunch of 25 years of software development, estimating skills and things I believed in and become really adaptable and pivot. And that's gonna sound corny, but just I would stick with that. And at any moment where you feel like I thought this was gonna work and I went for it and something different happens, just find your way to be adaptable. Second part is put more faith into what your employees can do and how creative they can be if they weren't bogged down by all the operational stuff. Less time in war rooms means more time to build, et cetera.
I got one more thing that's bugging me, and maybe we could all talk about it. Remember how I talked about how addictive these are? What I'm also seeing is some of my high-end engineers who now are doing 10x the work, they're not sleeping anymore because they're up all night building like crazy. And it used to be just me because I had a global team and, you know, whatever. I was like, oh, my God, what is everybody doing? And I'm watching that they're getting up at 2 a.m. to go prompt it because they feel like they're going to lose so much more productivity of building. And I asked, I said, do you feel pressure from me that I'm expecting all of this output or that you now have to work 24-7? And they're like, no, I can't stop because now I'm building, right? I'm creating new things, which is way different than my teams through events that, you know, we're working all night long or trying to get something shipped. So that's also very interesting. So my advice for you is, you know, we probably need to find another end to this hype and balance because when it's going to nudge you and you're, so as there's a couple people nodding, you're like, you know exactly what I'm talking about, right? It's two o'clock and Claudette needs me. Like one more prompt and we're out of here. I would also give you that advice is be adaptable, know that we're early, trust your employees, but watch out for them because the ones that are going to adopt this and have a constant AI genius that they can go to, whether in their work persona or others, it will be hard to shut off. It will.
I wish we could continue to have this conversation, but we are out of time.
Okay.
I really appreciate you coming and spending some time with us. Thank you very much.
Thank you.
If you don't have an account, Register here |
With the exception of our Horizons reports, most of our research is available for free on our website. Sign up for a free account and start realizing the power of insights now.
Our premium subscription gives enterprise clients access to our complete library of proprietary research, direct access to our industry analysts, and other benefits.
Contact us at [email protected] for more information on premium access.
If you are looking for help getting in touch with someone from HFS, please click the chat button to the bottom right of your screen to start a conversation with a member of our team.
