← Becker’s Healthcare Podcast
Becker’s Healthcare Podcast
How CVS Health Is Using Agentic AI Twins to Transform Patient Experience
00:25:34
Loading summary
Transcript23 lines
- [00:00]
A
Hello everyone, this is Jacob Emerson with the Becker's Health it podcast. Thrilled today to be joined by CVS Health's Vice President of Enterprise, customer experience and insights, sri. Thank you so much for taking the time to be with me on the podcast today.
- [00:15]
B
Thanks Jacob, Appreciate the time. Excited to be here with you.
- [00:18]
A
Likewise. And before we dive into everything, we want to talk with you about some interesting new tech that CVS has been rolling out. Can you first tell our audience a little bit more about yourself, your background in healthcare and a little bit about your role at CVS today?
- [00:31]
B
Yeah, so as you mentioned, I lead customer experience and insights for CVS Health. I started my career actually as an economist, so looking more at markets and evaluating them for competitive dynamics and defending mergers and acquisitions to the Department of Justice. And that's where I kind of first got exposed to healthcare. I used to look at things like hospital mergers and medical device mergers and it got really fascinated by healthcare. And then my career kind of took a bit of a winding path. I worked at, at, you know, ge, including GE Healthcare, did spent some time at Wells Fargo, you know, in the banking industry. So got a little bit more exposure to the, to the regulated banking industry before joining here at CVS Health. And my role at CVS Health is I look at all of the consumer feedback we get, whether that's through surveys or social media data, digital session data, call transcripts, whatever we can get our hands on to understand what are some of the patient client member pain points that we have across our entire enterprise. So retail at healthcare delivery, caremark, how do we improve the experiences for those stakeholders and make sure that we're delivering best in class experiences so that those people can get the healthcare they need. When I kind of give my team a tagline, I say our job is to learn from the past, improve the present and predict the future. Particularly with respect for our members who we serve as the internal advocate for across all of our internal teams, whether that's in technology or the business or elsewhere.
- [02:11]
A
Understood. So your background is more in the business and finance sector. Now you're here at cvs, modeling what patients and customers are doing. And you're here with us today, Sree, to talk a little bit about something really interesting that we, I haven't necessarily heard a lot from other healthcare organizations, at least publicly. CVS since late last year has been rolling out agentic AI twin simulations. So talk to our audience about what that is. When you talk about agentic twins, what does that actually mean and what does that actually look like in Practice for CVS specifically.
- [02:49]
B
Yeah. And I bet on this podcast you talk to a lot of leaders who talk about being consumer centric, right? And I think that's something that everyone says. You probably see it in every single 10k out there, every CEO dialogue. I think what's hard is being consumer centric is really difficult. Getting the voice of that customer into all of the aspects of what you do is challenging because it's hard to get to customers, it's hard to talk to them, it's hard for you to test things with them and learn from them. You're limited in terms of how you can outreach these folks. You can leverage surveys or maybe the scale you have to do pilots. And being consumer centric is challenging. You have that challenge simultaneously with the advent of AI and all the advancements you're having in AI. And what we realized in working with these company from Stanford, founded by a bunch of PhDs and computer science from Stanford, is that we can actually build agentic twins of our customers or patients or clients where we can actually build the way they make decisions. So not just what they did in the past, but how do they actually make decisions, how do they actually handle trade offs, what are the behaviors that they would exhibit in a given situation? And then we can generalize that to new scenarios and new situations. So to do that at scale, in our case something like a hundred thousand plus agentic twins, you will have a set of customers with you all the time that are effectively amplifying the consumer voice. So they are with you from when you develop the product, when you want to test the outcome, if you want to test some operational change. And it creates that real consumer centricity because that representation of the consumer, how they think, how they behave is with you throughout the entire process. So the way we've gone about doing that is we ingest a lot of different information. Of course all consumer consented, but we take things like a short behavioral interview, behavioral signals, spend data, things like that, where we can build a twin of each individual consumer. And again, you do that at scale where you actually have their thought process with you. And this agentic twin of the customer lives on and is always on for you to investigate and challenge new concepts, new ideas, and to actually even perform dress rehearsals of new things you're trying to launch out. And it's really helped our decision making, our understanding of the consumer throughout the whole product lifecycle. And really just it's brought them to life in a way that I don't think was possible in the past. And now we really, we really do. It's kind of putting your money where your mouth is in terms of centricity. You're having these customers with you across the entire journey.
- [05:44]
A
Wow. I mean, this is fascinating. You've literally copy and pasted your customer base and now you can see how they interact with all your different businesses. So back backing up here, Sri, why did CVS want to go down this route? How, why do you want to leverage this kind of technology and what does it actually do, problem wise for your patients and for your customers? How does it solve issues for them on the ground?
- [06:11]
B
Yeah, so I think the reason you would want to, we wanted to undertake this is we really do take consumer centricity seriously. We want to build products and services that help people access healthcare in an effective way and really solve the big challenges with the US healthcare system around, you know, simplicity, you know, ease of access, all those things that you hear about constantly from, from patients. And I think where we were landing is we have the ability to get to these folks via traditional methods, whether that's panel research or surveys, or even to a degree, looking at operational data. But each one of those things had challenges, right? Either it was speed, you know, or fidelity or any of those things that, that challenge you in terms of getting to that group. So when we heard about this methodology and this approach, it was really exciting because it would, it would really unlock what we've been trying to do, which is get the patient front and center on the way we build things. And I think if you, if you look at what we've seen so far, the four main things we've gotten out of these twins, just in terms of the value of how we build, is one, speed. You go from weeks to months to generate insights. So if you have to go get a panel and test something, whether that's messaging or some product or service, takes time. You got to go find those patients, you have to go interview them. You know, you have to, you have to distill those insights. Now we're talking about 15 to 30 minutes. And you know, I'll tell you honestly, Jake, the first handful of times we did this, I was stunned. Like, things that would take us eight weeks we were doing in 15 to 20 minutes. The second is fidelity. You know, you can only test so many things with a human being, right? There's only so many scenarios you can put in front of them before their brain sort of breaks. They'll get fatigue, the signal quality reduces. They're not able to answer with these AI representations. You can test thousands of real world scenarios and you're not extrapolating out what they're going to, what they're going to do in a given situation. You can present them with tons and thousands of different scenarios or situations and see how they'll actually behave. And that actually also solves a big problem in research where you have what's called stated intent, so what someone tells you and what they'll actually do. And here with this scenario testing, just because you've generalized how someone makes a decision or how they think, you can actually get to the actual behavior and even understand that gap between stated intent and actual behavior. The scale here you have all these folks on all the time, so you don't have to generalize the small pilots, you know, you have thousands, tens of thousands in some cases for the populations we look at agent of twins available to us at all times. So that scale of the consumer voice is, is tremendous and now available to us to access, you know, as we go forward. And then the final component is the reach, like these hard to reach populations that we often have. You know, whether that's particular disease states or conditions where it's just difficult to find those people or that can be really costly. Now we only have to really find them once because then we can create this twin of that person and then they'll be with us the whole time. And it goes back to, I think, amplifying that consumer voice with that hard to reach population, particularly because that's a group that's, you know, often marginalized because we can't get to that group to understand what they're experiencing. So if you, if you wrap all that up, we're getting away from reacting to small groups and with long timelines to get their feedback to predicting behavior, acting almost in near real time and having these consumers with us along the entire journey, it's transformative. It's, it's changing the way we deliver care. Because we're able to understand what's going to actually impact our consumers in a real way. While we're building these products and not having to do things like, you know, really narrow pilots and constant testing, we can go out with a lot more confidence in what we're doing because we've been able to dress, rehearse it with, you know, tens of thousands of customers before we actually hit the market. And anyone has to actually experience what we're building.
- [10:22]
A
Absolutely. I mean, listening to you talk about this is incredible. It, it sounds transformative. I, I've done marketing initiatives myself in doing those panels like you said, that take eight weeks and boiling that down to 15 minutes is. That's incredible. Not to mention what to hook onto the last part of your answer there. The equity of it all, of being able to reach your entire customer base and know how they all respond and not having to put everybody into pockets. It's amazing. I can't imagine what's going to be coming from this. And so I think that's the natural follow up question here. SRI is how does this change the way that cvs, it makes decisions compared to those traditional approaches you just laid out for us, even just from a marketing perspective. But maybe you could expand a bit in terms of clinical decisions, other areas of the pharma, of the, of the industry that the company reaches. How does this change the day to day for CVS and the business overall?
- [11:21]
B
Yeah, and I can give you a couple of just concrete examples that we've, that we've tried so far, that we've seen success with. I think one, you know, one classic one for us in the pharmacy side is adherence. That's an area we've got a tremendous amount of focus. Obviously we want people to be adherent to their medication, we want them to stay healthy. Accessing medications is really important to us. That can be a difficult group to talk to, particularly non adherent patients. There's a lot of reasons people don't adhere to medication, some of them related to stigma, fear. And you're in a sense, if you're going to talk to that group, you're making them relive whatever that was the challenge that they had and that can be really difficult, that it can lead to false signals, it can lead to, you know, different types of challenges in terms of getting to the population, fatigue, lack of sample, those types of things because they're just a difficult group to reach. What we were able to do with the twins is we were able to actually go through and investigate in detail adherence issues. So get, because you don't have to face that and it's, you know, you don't have that same sort of challenge with the gentic twins because they're AI representations. We were able to get way deeper than I think we could have in a standard qualitative and even quantitative interview. What we were able to figure out is a lot of the things about adherence were the same that you would expect that we've always seen. You know, it's, you know, fear of side effects and fear of, you know, you know, not being able to talk to a clinician and all those things. That we've been solving over the, you know, course of the past few years. But a new one's emerged and it was really around the fear of getting a refill, which sounds really, you know, it sounds a little bit strange. It wasn't what we expected. But we were able to see like, you know, a lot of what we were observing was people are saying, okay, now once I start a medication, how do I make sure I'm always going to get it? What happens if I can't get that medication? Oh, you know, does that, is that going to have outcomes for me? And all those things that we were able to see that were clear in the data and I think what it pointed to is that we have a lot of best in class digital tools we're building and have built that are making accessing medications much easier. So it gave us a lot more confidence in those tools we were building. And in addition, we were able to actually look at how do we optimize some of those tools to ensure that we are making it as easy as possible on the refill. And I think what it really pointed out is this anxiety and it's patient anxiety that you were able, that you were able to see in, in the agentic twins around all the process related to the prescription. So it wasn't just the medication, it was the process and the experience that led to adherence. And you know, we were able to back that up with a lot of the actual data we were able, that we were collecting from the customers. When I say actual, I mean operational data that we were collecting from the customers. It was a way for us to unlock something that we didn't see in the past. And I think what's particularly unique about it is we're talking about effectively about robots here. Jacob. We build robots that represent customers, but they were unlocking something that's superhuman in anxiety. And so we were able to get to the emotion through these agentic twins, which I think is a really fascinating use case for us. And then just a quick related one, just because you mentioned marketing, another one was related to our pet Rx. And I'll just tell the story really quickly. You know, PetRx is our, you know, we're looking to move into pet care. And we were looking at how do we message to those, to those pet parents, if you will. I'm a pet parent, I've got a dog. And you know, there was a lot of the data. We were expecting like emotional data that people care a lot about their pets. They become like family members. One of the more interesting things we saw from the gentic twins was people were saying, you know, that even giving medication to a pet was an act of devotion, not necessarily a chore. So any messaging appealing to the emotions was going to be successful. But what the gentic twins allowed us to do is present scenarios beyond that. And what we found was it wasn't enough to just be connecting emotionally. You had to show that we had the capability to actually get the scripts from the veterinarians, have an understanding of what those scripts are, how the side effects work. That was really important to the pet parents. So by being able to drill in and get to the kind of stated versus actual behavior using the twins, we were able to uncover that the messaging needed to have both. It wasn't sufficient to just have messaging and marketing that was appealing to emotions. We had to appeal to the tactical side too, because pet owners are incredibly pragmatic despite this emotional connection that I think we all observe and see. So anyway, those are two kind of quick examples of where we've, we've seen early wins and you know, we're operating this at scale. So there's more to come, I think here as we keep moving forward.
- [16:22]
A
Oh, absolutely, I can imagine so. And for pretty much any company listening in right now, I would imagine they are taking notes on how to replicate something like this Sree, because the, the possibilities seem endless just listening to even those, those two examples you shared. And I really like how you framed it of unlocking something that is human. You're unlocking something that the humans on the other side maybe hadn't been able to before, which is really fascinating.
- [16:48]
B
It's a really unique AI use case. You know, it's not, it's, in a sense, it's, it's not one of the ones that I think are traditional. And it's, it's using AI to understand human behavior, which again, sounds a little counterintuitive, but I think.
- [17:03]
A
No, it makes sense.
- [17:04]
B
Yeah, yeah, it's really creating this always on approach with our customers.
- [17:07]
A
Definitely, definitely. Is there anything you'd add for, for the health care audience that's listening in terms of the value prop here for the wider health care system?
- [17:18]
B
I think, you know, if you, if I were thinking about more broadly, you know, even things like patient experience and I'll use the experience hack because that's, that's where I tend to focus. It's ability to test messaging like how do we change in particular in health care is really painful. That's observed everywhere. You can see that across multiple populations. When you change people's medications, when you change formularies, when you change the way they access care, it's painful. But what this allows you to do is test different messaging and different approaches to see how do you, how do you smooth that, how do you make that easier on the patient. And you're able to do that here in a controlled, safe environment that you're able to see the results before you actually have to go into the market and test that. I think that's a really big unlock for healthcare. The other one I think is understanding journeys. Healthcare journeys are incredibly complex and you know, it can be difficult to actually get patient feedback along every aspect of that journey. But here you can, you can do that and you can see how people are experiencing every aspect of where they engage with you. And I think understanding quickly where you have pain points and where you may need to do some work and may, where you're excelling and maybe where you need to double down healthcare in particular. To me, this is the way to test and learn quickly with those audiences. And I think it's a major unlock because I think that's where, and I know I keep using the word unlock because I think in these populations traditionally it's really difficult to get the feedback and they're also really difficult to pilot with because if you're running tests, you're actually running tests on someone that's experiencing something or trying to access care and that's not always easy and that's not always right and it can be difficult. Here you're able to do those tests in this controlled environment without having to worry about the patient risk or outcome. It makes it safer and it makes it more controlled. And I think, you know, the follow on effect is you're going to deliver better care for those patients and for the stakeholders.
- [19:27]
A
Sherry, let me ask you a bit of a prodding question. Just as this technology continues to develop, we'll continue to see it across many companies, different industries. How are you thinking about how CVS wants to ensure responsible use of agentic twins at enterprise scale? You know, you're able to now test an external pressure on, on your customer base. How do you balance some of the potential negatives of that or, or ensure that there's going to be human judgment, trust, good governance of this technology as it begins to, as already has started to influence real healthcare on the ground?
- [20:08]
B
Yeah, it's a really important question. I think one, we do, we have very strong controls around this and there's a couple different ways we do it. We continuously test These twins against not only our own data, but third party studies to ensure that we have the highest quality information. And you know, one of the things that we're doing that these twins do as well is they actually live in the world, so they ingest current events. You know, if you're a, if you have a twin that's a Democrat, they'll read Democratic news organizations. If you have one that's a Republican, they'll read Republican news organizations. So you're able to actually keep them up to speed, you know, even on current events. But regardless of that, we're always back testing, we're always making sure that they're meeting our standards of, of matching known human results that we have. The second is, you know, right now we have my team of researchers, I have a team of, you know, analysts and market researchers that we have organizationally that are, are the ones who are administering this test because they represent human behavior. So you can lead the witnesses. So if you don't have someone, you know, at the controls who's engaging with effectively, the way it works is an agentic chat and you're not asking the questions appropriately, you can get results that are misleading. So right now it's my team of researchers that are, are looking at that information and validating it and making sure it's meeting the standards that we would have for human studies. But one of the things I point out there though, is not to live in an echo chamber. On occasion, the agentic twins will teach us something new we didn't know. And we shouldn't just be sitting there saying, oh wait, it didn't match exactly what we thought in the past. So therefore it's wrong. What we need to do is investigate it further so we make sure we have that process built out. And then I think the third thing, and probably most importantly, I feel like everyone always frames AI as replacement and as like, what is this going to replace? It's always the first question you get when you're deploying an AI solution. I don't think this is replacing talking to our customers. It's augmenting our understanding, it's amplifying their voice, as I've said. But ultimately we're going to have to always talk to our customers because that's important and that's what good companies do. You need to listen to people and you need to understand what their friction and their pains are. And it also is important to do that, to potentially catch emerging issues and things like that. So I view this as additive to what we're doing. With our human populations. And I think if you combine those three things, you know, the, the, the back testing the researcher at the wheel and then always talking to the human populations as well. It's a major, it's a, it's a set of governance that gives you a lot more confidence in how the twins are operating and it also helps you amplify the impact. Because when you're going out to various stakeholders and you're doing the change management of saying, hey, listen to, listen to these agentic twins, they're telling us, you know what to do. The people you're working with have more confidence in you because they know that you're taking the care and you're doing the work to make sure these things are accurate.
- [23:14]
A
Absolutely. And like you said, sri, I mean this is, it sounds like this is unlocking things that your customers were already feeling and thinking anyway. This is now just another way for you to, for you to reach, to reach that. So really, really fascinating. Anything else we're missing here? Any final thoughts you want to share with our audience?
- [23:33]
B
Yeah, I think at CVS Health, it's not just that we're experimenting with this because I think a lot of people will frame this as it is new and I get it's an emerging technology, but we were at the forefront here and I think it's gone beyond experimentation. We're, at this point, we're operating at scale, as I mentioned, 100,000 twins. My hope is to get us to 300,000 or more in the future. But we're using this responsibly at that scale to improve people's lives. And this is giving our teams unprecedented clarity and foresight into, you know, what's happening in the industry, what's happening with our consumers, how can we better serve them, how can we build better experiences, how can we learn faster and design better? It's, it's helping us, it's a step change for us in terms of how we actually deliver healthcare. And I think in the end it's going to help us deliver better health and support for every customer member that we serve. And we're going to do it with more confidence, empathy and understanding. And it's going to help us achieve our goal of being the most consumer centric company in healthcare. So, yeah, I hope everyone listening out there, particularly those in healthcare, understand that this is a tool that I think can really give your customers an unprecedented voice in the way you operate.
- [24:59]
A
100%. I mean, this does seem to be the, one of the largest commercial rollouts we've seen of this technology or that we've been we've heard about in healthcare on our end. So SRI really appreciate you taking the time to chat with us about this, and please keep us updated as this all evolves. We'd love to hear more about it.
- [25:18]
B
Yeah, always a pleasure talking about this, and it was a pleasure talking with you. So excited to come back and tell you all the great things we're doing
- [25:25]
A
in the future and to our audience. If you'd like to listen to more podcasts from Becker's Healthcare, you can visit Beckershospitalreview.com.