Lessons from Corporate Innovators

Decoding Analytics: Richard Clarke Unveils Highmark Health’s Strategic Approach to AI in Healthcare

Play episode

Episode description:
In this podcast episode, we engage in an insightful discussion with Richard Clarke, Chief Data and Analytics Officer at Highmark Health.

With an impressive 8-year tenure, Richard offers a unique and seasoned perspective on the evolution of analytics and data science. He and his team have built a remarkable foundation at Highmark, skillfully leveraging it during the recent surge in Generative AI, while simultaneously investing in high ROI projects across the spectrum of data science and analytics projects.

This episode isn’t just pivotal for those at the forefront of healthcare organizations; it’s also packed with valuable insights for leaders across a range of industries looking to harness the power of data and analytics in their own businesses.

Show links:
LinkedIn: https://www.linkedin.com/in/richard-clarke-data-analytics/
Highmark Health website: https://www.highmarkhealth.org/

[00:00:00] Sean Ammirati: On today’s episode of Agile Giants, I am thrilled to welcome Richard Clark, the Senior Vice President and Chief Analytics Officer at Highmark Health. I first met Richard during an executive education program at Tepper, where I was teaching. Recently, we reconnected and discussed his career journey, and immediately, I knew I had to have him on the show.

[00:00:27] Sean Ammirati: During his eight-year tenure at Highmark, Richard’s career offers a unique perspective into the dynamic world of analytics and data science. This conversation is relevant not just for leading healthcare organizations but offers really valuable insights for leaders in various sectors. I’m super excited for this engaging discussion on Agile Giants and hope you enjoy the conversation as much as I do.

[00:00:51] Sean Ammirati: So Richard, thank you so much for joining us today. As I said in the intro, I’m really excited about this conversation. I did want to start before we get to what you’re doing at Highmark and some of the relevant questions. I think there are two context questions that are probably relevant for people.

[00:01:08] Sean Ammirati: So the first one is about you. Could you talk a little bit about your academic background and your work background before you came to Highmark?

[00:01:16] Richard Clarke: Yeah, be happy to, and thanks so much for having me, Sean. I’m really excited for this conversation as well. So I always describe my career in kind of three phases, I guess.

[00:01:26] Richard Clarke: I had my academic phase, PhD in Neuroscience from the Center for Neural Basis of Cognition at Pitt and CMU, which was amazing. I always joke with my data scientists here. I was using R for all of my statistical analysis because I couldn’t afford a MATLAB license, not because it was cool, but I guess now it’s come back around a bit, which is fun.

[00:01:49] Richard Clarke: I then left academia and went into management consulting for just shy of a decade at McKinsey and Company, really focused on the application of advanced analytics in financial services. It was a fast-moving time in that particular industry. The thesis of me then coming to Highmark, a health services company here in Pittsburgh, was to bring some of that work we had done in financial services into healthcare. And my real passion is around how to apply analytic insights not retrospectively, but at the moment of decision-making inside of workflow to inform what we are doing to lead to better outcomes. That was a lot of the work we were doing in financial services, and it’s been a really fun run so far in healthcare to do the same thing.

[00:02:46] Sean Ammirati: And then the second part of this is a lot of people who listen to Agile Giants are not in Western Pennsylvania or even in the greater Highmark footprint. So maybe just a minute on who Highmark is because I think that’ll help the rest of the conversation.

[00:03:01] Richard Clarke: Yeah, I’d be happy to describe it. We do have a national footprint, so hopefully, at least some folks have heard of us. But as I said, we’re a diversified health services company, so Highmark Health, the parent company that’s the part of the organization I am in, really has five business units.

[00:03:21] Richard Clarke: The Highmark blue-branded insurance company has a primary footprint in West Virginia, PA, Delaware, and New York, covering about 6 million insured lives. We also have our health system here in Western Pennsylvania known as Allegheny Health Network, primarily operating from Pittsburgh up through Erie and into Western New York.

[00:03:45] Richard Clarke: We then have a couple of our diversified businesses, a national dental business, United Concordia Dental, and our national stop-loss business, HMIG. Finally, and very importantly, our technology platform, enGen, which provides many of the platforms for the payer-provider space, allowing us to scale the things we’re doing within Highmark Health Plan and Allegheny Health Network.

[00:04:15] Sean Ammirati: Perfect. Okay, so now let’s get to this eight-year journey that’s probably just getting started, but where you’ve made analytics real in healthcare. Talk a little bit about how even just your role to get started. We’ll talk about projects in a minute, but just even your role is different today versus what it was maybe when you joined Highmark.

[00:04:39] Richard Clarke: Yeah, it’s like you said, I do think in some ways we’ve covered a lot of ground, but in some ways, we are just getting started. I celebrate the leadership here at Highmark, David Holmberg, Karen Hanlon, and others for recognizing analytics as a key part of our strategic differentiation. A key part of our strategy.

[00:05:00] Richard Clarke: I’ve been lucky to have a role that is all about partnering with both our functional areas and stakeholders, business segments, to ask that exact question: How can we deploy analytics in a way that is in the workflow, informing decision-making, and thus creating an advantage for us to succeed in the market? That role has evolved, but at the center of it from the beginning is this kind of translation. How can I sit at the interface between the evolving capabilities we have, and, we’ll talk about some of the rapid evolution over the past year, but really, at the end of the day, it’s still been about how we take those capabilities and then leverage them to enable us to get the outcomes that we want?

[00:05:49] So I’ll often get the question, “Well, do you have an AI strategy?” And I was like, “Well, you know, we don’t have a defined AI strategy necessarily. What we have is our strategy is to achieve our business strategy through the use of AI.”

[00:06:10] Richard Clarke: And that, to me, while a subtle distinction, is really important and has really been a consistent thread throughout the eight years.

[00:06:19] Sean Ammirati: I love that, right? It feels a little bit like the internet, like people had internet strategies at one point and now they have business strategies.

[00:06:25] Sean Ammirati: I think AI is at a similar point. Like people want to understand this AI; it’s like, “Look, what we want is to delight customers, do it better, right?” Like that’s, you know, and we’re going to use these new capabilities to do it. Now, my intuition is when you joined Highmark. There was probably more evangelism around that than there is today, but I think you also put some wins on the board pretty quickly, early on that set the tone.

[00:06:56] Sean Ammirati: Maybe before we get to some of the more recent projects, what were some of those early projects that maybe set the tone, generated some momentum?

[00:07:04] Richard Clarke: Yeah, you’re totally right. There was quite a bit more evangelism that was required five years ago, whereas now, as you said, the recent wave really has brought a lot more people to the front door, if you will.

[00:07:18] Richard Clarke: Some of the early wins, you know, we’ve always been super practical. So, I love how you framed it. It’s all about how are we delighting our customers and our users. A couple of things I would say that we did very early on is, we launched a program here at Highmark Health called thinkUP.

[00:07:39] Richard Clarke: It was part of our overall strategy, which is all about healthUP, scaleUP, and then thinkUP. thinkUP was all defined as how do we reimagine how we do work. And for us, that was a perfect frame to go out and say, well, how can we reimagine how we do work through advanced analytics, artificial intelligence, etc.

[00:07:57] Richard Clarke: The pattern that I kept telling both my team and the organization was, “Hey, just look for people that have a scratch pad next to their computer, post-its all over their monitor. Because what they’re doing is they’re trying to go out and find information to then make a decision. And what we’re going to do is we’re going to go out, find that information for them.

[00:08:16] Richard Clarke: We’re going to apply some analytics to then suggest some action, and we’ll present that back to them to help them.” That pattern really resonated with the organization. It allowed us to tell very granular stories that we could say, “Hey, nurse case manager Jennifer used to work like this, but now works like this.

[00:08:36] Richard Clarke: And oh, by the way, it is 35 percent more productive and 20 percent more effective.” So those kind of vignettes turned out to be super powerful for us. And then the last thing I’ll say is 2018 was a big watershed moment for us because that’s when we activated our first real-time scoring engine.

[00:08:59] Richard Clarke: That was in workflow, with a millisecond timescale, decorating a particular piece of work, in this case, a prior authorization with 10,000 plus variables about a member and then making a prediction that actually informed workflow. So if that prediction was above X, it went to Y. If it was below X, it went to Z. I always joke that the lights kind of dimmed here in Fifth Avenue Place when that went live, but that was a big deal because we were proving to the organization that we could do it not only in real-time but not through an Excel spreadsheet that was emailed to someone, but actually native in the workflow so that it was seamless for the end user.

[00:09:43] Richard Clarke: So those two were really big deals for us.

[00:09:46] Sean Ammirati: That’s awesome. I think for everybody in healthcare, they followed both of those perfectly. For those not in healthcare, the second one, can you just a minute on what you mean by prior authorization? Because that’s like a term that those in other spaces may not be able to land on.

[00:10:03] Richard Clarke: Yeah, thanks for calling out the industry terminology. Please, please do continue to do so. So, prior authorization would be before a particular, in many cases, procedure. There might be an authorization that is required, and that authorization could be checking a number of things, making sure that A, the member actually has coverage for said authorization, B, that it is medically necessary based on the medical policy of the insurance company, all leading to it’s going to actually be reimbursed if it occurs, right?

[00:10:35] Richard Clarke: This is a process that we want to A, make sure is seamless and as fast as possible because we don’t want to stand in the way of appropriate care. But we also want to use it as an early warning system that something is occurring. The particular case I was talking about was, in many cases, health insurance companies will use paid claims to inform them, but the paid claims might occur weeks or months after a service actually happens, not the best for actually helping someone navigate a complicated situation.

[00:11:07] Richard Clarke: Whereas that prior authorization, by definition, occurs before the procedure. So the particular prediction we were trying to do was identify things that we thought were gonna be very complex. Because of the member’s situation, other things that are going on, that enables us to engage early to then help a member navigate.

[00:11:24] Richard Clarke: It might help us navigate them to social services that could help them with some type of barrier to care, be it transportation, housing, food insecurity, etc. It could be polypharmacy. People often will have challenges if they have four, six, eight drugs that they have to coordinate around. It could be navigating multiple doctor visits, etc. So that was really the goal is to use that as an early warning system to then help us intervene early. And our strategy is all about proactive, personalized, and simple interventions and solutions. So it was really enabling that.

[00:11:57] Sean Ammirati: That’s perfect. And my guess is everybody’s following now, right? Because we’ve all experienced this as consumers, but not necessarily with that buzzword and, and it’s super exciting. I mean, for what it’s worth, 2018 is early, but people are still continuing to get better and better at what we do around this kind of using prediction technology, analytics, AI, machine learning, whatever, but to do these things better, right? And it is one of these examples, I think, where, like, care is better, the provider has a better experience, the payer has a better experience, you can, it’s really these kind of win, win, win opportunities for ML. So I’m sure you’re not done with prior authorization, but starting in 2018 with real-time prior authorization is. I mean, that was early in the evolution of healthcare, at least for the groups that I’m aware of. Let’s fast forward to today. What’s your current portfolio of projects look like and how are things a little different in 2024 here?

[00:12:59] Richard Clarke: Yeah. So we still have a robust portfolio of, you know, however I want to call it, traditional machine learning projects. And, you know, in many ways, I feel like those are just seamlessly integrated into what we do because we’ve been at it a while. And those are really around, you know, how do we help, how do we help with predictions? How do we help with forecasting? How do we help with automation? And that’s a very robust portfolio. We’re largely organized now around analytic products that are delivering that into the organization that all have, you know, state, all have customers, all have targets, etc. And that feels pretty baked into our operating processes.

[00:13:40] Richard Clarke: The brand new thing, of course, is all the generative AI work that has emerged. And we’ve launched a dedicated program specifically focused on generative AI to really both: A) capture the use cases coming in, and then,  B) kind of from either top down or outside in, say, these are some of the transformational things we should be doing. The ramp up there has been unbelievable. In the first quarter of 2023, I think we had 200 plus ideas for generative AI. We’ve been working through the backlog, and we’ll talk about some of the successes, some of the failures there. Just had a review with the team yesterday on that. We have 50 plus that are kind of in, you know, kind of production already on that. Many are smaller for sure, but some are definitely bigger and we’re still working through an ever-growing backlog and probably will be for a long time as people understand the technology more because I will say it is even more intuitive than some of those other products out there traditional machine learning. And what we’re finding is the more people engage, the more ideas they have, which is a nice flywheel, but at the same time, a challenging one to make sure that we can meet that demand.

[00:15:06] Sean Ammirati: So, I mean, 200 in Q1 of 2023, if people think ChatGPT didn’t change awareness around this, right? That’s amazing. So, zero to 200 in a quarter. And then, so those 50 projects, most of those were started back then? Was it mostly, mostly from that original 200 or?

[00:15:22] Richard Clarke: Some yes, some no. And important to say, they come in three forms. Form number one is really just the enablement to use generative AI that is, I would say, resident in some application that we already use. A huge use case, of course, marketing is kind of littered with enormous amounts of use cases. Anywhere you can go where kind of more content equals more revenue. Huge, huge use cases, right? So text to image, simple create early storyboards for campaigns. We don’t need to build that. We just need to enable that. So a bunch of those are just the enablement of and making sure that the enablement of kind of still works for our responsible AI framework that it doesn’t have any challenges from an ethical, legal, compliance perspective. So there’s a bunch of tranches that are there.

[00:16:23] Richard Clarke:Another tranche is working with specialized vendors. That might come to us that have a solution that is either trained in a very specific domain or corpus or has a particularly challenging set of integrations that we don’t think we should do. 

[00:16:32] Richard Clarke:And then the third, of course, is where we decide to spend our own development time to build some bespoke solution. One of the biggest challenges I’ve seen, frankly, is deciding where a use case falls across those options because we’re trying to, of course, deploy our own development capacity at the most specialized, kind of most strategically advantaged. So we talk a lot about being ruthless curators of that backlog, and that’s been hard.

[00:16:57] Sean Ammirati: Yeah, I was literally, as you were talking about that, I was trying to picture myself in your shoes and figure that out. Like, how do you generally think about, okay, this is something we should do bespoke versus use a specialized?

[00:17:07] Richard Clarke: I’m sure the answer I’ll give now will be different than the answer I’ll give in two months, three months, four months because we’re figuring out. Plus the market is moving so darn fast. That because you would love to say, well, you know, this is something that I don’t think is going to be available to me anytime soon, whatever it is, right? But it’s impossible to predict. So the way that we’re really thinking about it is, really, number one is value, of course, is it, you know, is it incredibly valuable for us? Of course, a second one is also, are we aware of any current offerings? You know, that’s the one that’s the hardest to know because the offerings are so vast and changing so rapidly. But I think the third one that has been most important for us is how much does it require our own data to fuel it and, you know, certainly those ones that don’t require our own data are right before some other external party. But the ones that really need to tap into or and kind of be integrated with either our own data or our own data legacy platforms. I think those are the places where we see the most impact because they’re the hardest to implement as well.

[00:18:24] Sean Ammirati: But by the way, I mean, I want to keep going on this conversation, but I would just say to listeners, I think there’s some real business model innovation needed for AI around that as well. Like, that should not be driving Richard’s decision-making process, but clearly, given how people are approaching this today, it is, and I think there’s huge opportunities there. So for all the entrepreneurs listening, let’s get a little bit more creative about how we’re bringing solutions to our customers here. And Richard, I don’t know if you have any reaction to that, but I do want to get to wins and fails in a minute more broadly and lessons learned there.

[00:19:03] Richard Clarke: No, I would like just quickly, I agree with you. This certainly is a ripe space for innovation and kind of turning the conversation to not that would be good. I mean, the other, I guess, reaction or observation I would have for any of the entrepreneurs on the, you know, listening is, you know, obviously the easy to demo but really hard to implement is incredibly frustrating for those of us that are stuck with the executive that got super excited by this thing that they saw and now expect it to be able to be simply delivered in two seconds. And so, um, continuing to be practical and upfront about what, you know, getting to real production and kind of the definition of done, I think can differentiate some folks versus just leaning on the easy to easy to demo right? The sort of hymns effect in health care, right? So for those not in health care, like there’s this conference every year called hymns and people come back super excited about this thing that shows really well on a trade show, but maybe not as easy and in reality. Um, so we’ve kind of gotten to two lessons learned, I think already.

[00:20:33] Sean Ammirati: But I am curious, like you’ve been you’ve been at this for a while. You’ve certainly ridden this wave aggressively over the last year. When you think about what’s working and what’s not, maybe talk a little bit about a couple of examples.And then some some lessons people could take from that would be really helpful, 

[00:20:36] Richard Clarke: Richard. Yeah. And I think, um, I always find that maybe some of the Um, you know, challenges or failures are most instructive. So I’ll start there and I’ll start there in kind of two dimensions. So, you know, the biggest lesson learned on the failure is the readiness of the data is by far the single biggest, um, challenge and thing that we are now doing a much better job of assessing.

[00:21:01] Richard Clarke: Um, and it’s not just the readiness of the data for the initial idea, but really got to get deep into it. So I’ll give you just simple example. We, we, one of our initial ideas was, uh, hey, let’s build a, uh, AI assistant for some of our strategy analysts that have to go out and, um, collect information about competitor products, um, because we’d much rather have them spending all of their time.

[00:21:24] Richard Clarke: Analyzing and interpreting the implications of those and not collecting that information. Sounds simple, but turns out a lot of those product details sit in complicated tables. They sit in images. They sit in things that are not ready to feed into a large language model or a pre trained transformer. And so, Uh, now, the good news is we learned that, and then the good news is that we kind of failed forward in terms of building, uh, table extraction algorithms, you know, really learning a lot about, you know, uh, RAG infrastructure and other, other, other things very rapidly, which now will help us in the future, but at the same time, That business area got really excited because the first question they asked worked, but the second, third, fourth and fifth did not.

[00:22:11] Richard Clarke: And so now we’re kind of having to manage the emotions of that business unit of, uh, Oh, well, maybe this gen AI is just hype because it didn’t actually work. Um, you know, so that was definitely, uh, that was definitely, uh, you know, a lesson. Ah, lesson learned. Um, I think the other lesson learned is frankly one that is the exact same lesson that we’ve been learning since the start of this journey, which is, uh, you know, business engagement and adoption is still is still king.

[00:22:42] Richard Clarke: Like that is still king. It was with traditional machine learning, and it’s with, uh, it is with, uh, it’s the same with general generative A. I. It’s no different. Um, and if we can’t get folks to iterate, engage, iterate and have a continuous improvement mindset. Uh, it’s not going to work because this is not magic.

[00:23:02] Richard Clarke: It’s not, it doesn’t work perfectly the first time. It requires iteration. Um, and so that is still, uh, the case. And I think we lost, I’ll be honest, we lost a little bit of that because everyone was so excited by the technology that I think we did lose a little bit of focus. We’re back now on it. Uh, but it’s amazing to me that that’s been the continuous thread, you know, 

[00:23:22] Sean Ammirati: throughout.

[00:23:23] Sean Ammirati: I mean, in some ways I think that’s a continuous thread. And I’d actually say on the readiness of the data too, right? That’s a good point also. Hey, like machine learning has always been. Past is predictive of the future. And so you need to have good, clean, large data sets. And I think, I think this is why you see companies and you’re a good example of this, who’ve been working on this for a long time, not getting caught up in hype, but actually delivering value because they’ve got the infrastructure in place, right?

[00:23:54] Sean Ammirati: If you don’t have a mature data science organization, you can do some, you could, you know, put a thin wrapper on top of an LLM and you can, you can do some things that feel magical. But They are going to get to question three, four, five, and then all of a sudden fall flat and then be like, oh, this is hype, not, not something that’s, that’s really valuable.

[00:24:16] Sean Ammirati: The, 

[00:24:16] Richard Clarke: um, I, I was sitting with some other healthcare leaders, uh, at a, at, you know, a conference. And honestly, it doesn’t even matter which one because I guarantee 90 percent of the sessions were about generative AI because that’s, that’s pretty much it. And the really great way that this healthcare executive framed it was once you can fully represent the problem you’re trying to solve in data.

[00:24:36] Richard Clarke: Then it is magic. But getting to that point is a lot of work. And you know, I, I totally, I totally agree. And oh, by the way, getting to that point also requires the engineers and scientists are working out to really understand the business problem like deeply, like deeply understand the business problem and deeply understand the business processes, which then goes to the other point of in order to understand that you need deep kind of collaboration and connection to the business.

[00:25:03] Richard Clarke: So, you know, a lot of those are, are still present and still really important. 

[00:25:08] Sean Ammirati: Yeah. And the reality is when you’re doing this well, it plugs back into a lot of the classic or traditional ML projects. I think these things are less islands than people, people perceive as well. 

[00:25:21] Richard Clarke: That’s a great, that’s a great point.

[00:25:22] Richard Clarke: And then, um, that ecosystem then still needs to work with the legacy infrastructure that every company. You know, kind of has going right? Because at the end of the day, it’s gotta help the person doing the work and you know, that person is likely in some system that was, you know, engineered 20 years ago.

[00:25:43] Richard Clarke: So those integrations are still very real. 

[00:25:46] Sean Ammirati: So how does the special purpose stuff you’re doing around Gen AI fit into that? You talked about kind of your analytics projects are organized around these kind of core offerings like What’s the interaction between, between those pieces 

[00:26:02] Richard Clarke: of your group? Yeah, they really start to bump up in the way you were just describing, right?

[00:26:06] Richard Clarke: When they need to sew. Uh, exactly like you said, we are seeing these, um, new generative AI solutions will need to leverage other components that we already have, so be it our natural language processing engine, or they might be calling our, um, endpoints for some of our predictive models, or our ID and strat engines, and it has been a really interesting, um, um, Uh, journey because this is becoming just more and more of a team sport.

[00:26:36] Richard Clarke: I was kind of just talking about this with, uh, our program lead for the gender of a I program. I’m like, my goodness, this is such a I have not had so many interactions with Not only all the data and analytics folks, but privacy with security, our infrastructure team, engineering. It is, uh, remarkable to me that, um, that this is all occurring.

[00:26:57] Richard Clarke: And of course, it all goes to these kind of, whatever you wanna call it, fusion, uh, team models or hybrid team models, agile models, whatever, you know. But this is like the, at least for me. Um, the most obvious kind of manifestation of that, um, that, that I’ve seen to date. 

[00:27:13] Sean Ammirati: Yeah, 100%. Um, completely agree. How do you think about it from the other way?

[00:27:21] Sean Ammirati: So how do you think about the people who are these end users? How much they should be doing versus leaning into to your team to do? 

[00:27:31] Richard Clarke: You know, I think, so I would, I would answer that in two ways. One way would be. Um, our our view has been we want them to be starting right now, yesterday to talk to machines, if you will, right?

[00:27:46] Richard Clarke: Because this is going to be part of their job, everyone and how we do that is going to be different. And frankly, how folks do that is going to be tailored to their job. And the quicker we can get to that, the more distributed Innovation that we can have. I think the better off that we’re going to be. So the onus on me is to provide them a secure, um, pathway to do that.

[00:28:11] Richard Clarke: Um, you know, because we’re pushing that and thus we want them to then go and do that in a way that again doesn’t have any data leakage or, uh, you know, challenges for us. That would be one answer is like, go, go, go. If there’s one thing that I could tell all of them is just start to, you know, start to talk to machines.

[00:28:28] Richard Clarke: It’s really important. Um, you know, the second comment. Um, and probably harder, um, way that I want them to do that is also to take a step back and say, Well, how is this going to change my workforce? And, um, I can’t do that from my seat. I don’t know. Their roles. I don’t know their work and really start to ask the question.

[00:28:49] Richard Clarke: How much of my work can either be, uh, you know, done by a I supported by a I or not. Um, and really start to think through that and start to think through how they’re going to redesign jobs, uh, that. Uh, take advantage of that and then pull us into that. But I can’t do that from my seat. Um, and I think it’s going to be a really, um, big challenge for leaders, uh, to uniquely be thinking through that for their organization.

[00:29:16] Richard Clarke: And it’s going to be a big change for us because the, uh, usage of AI isn’t going to be coming out of some central group, right? It’s going to be very distributed and disseminated throughout every, uh, area. And that’s the right answer, but frankly, it’s also a more challenging answer. Uh, yeah, 

[00:29:33] Sean Ammirati: I, I, I love both of those.

[00:29:35] Sean Ammirati: And I’ve, I’ve got a follow up question, but I want to make an analogy first for people. So we talked about this being like the internet. So when I started working, I remember I had a boss two levels up who his EA would print out his emails for him and then he would dictate emails back to her to send, right?

[00:29:52] Sean Ammirati: And like, there’s like, that’s the email thing. I don’t do the email thing. I think people who are like, Oh, I’m going to stay away from. The A. I. Thing are kind of like that colleague. Now he was close to retirement. He he fared. Okay. But I think this is pretty quick. Like you want to be you want to be thinking through both of these things, uh, to make sure you’re not you’re not caught into that.

[00:30:13] Sean Ammirati: I’m I’m curious, Richard, practically on both fronts. How do you do that? So on the on the talking to machines point, are there tools you’re giving them to do that? And then on the future of work thing, are there resources you’re giving them to help them sort of understand what AI is capable of, what it’s not like, just how do you, how do you practically help these leaders do both of these things?

[00:30:37] Richard Clarke: Yeah. Um, we’re certainly not perfect on either dimension and we’re, we’re, um, pivoting and, uh, zigging, zagging as we go here, but on the first one, uh, yes. So we have, uh, provided, uh, our internal Uh, team members and we’re rolling, um, rolling this out further our own kind of internal. We call it sidekick or, you know, chat, chat, GPT ask, um, uh, interface.

[00:31:04] Richard Clarke: And for us, it’s really important, especially in health care that that is built on our own, um, You know, secure infrastructure that we know there’s zero data leakage, and we can have confidential information or P. H. I. Um, protected health information go into it, and it’s not going to go anywhere. Um, and so that is one.

[00:31:23] Richard Clarke: Um, that’s how we’re enabling the first on. We’ve seen amazing engagement. It’s been awesome to see not only amazing engagements. We have, uh. Data scientist in my group is the kind of product owner of that. She holds office hours. I think they’re at 10 a. m. this morning. Super well attended people talking about, uh, how they use it sharing.

[00:31:44] Richard Clarke: And you know what we tell everyone is try to break it because that will help us make it better. So that’s how we’re enabling one on the second one. Boy, that’s a tough. That’s that’s harder, right? We are. We are certainly encouraging people were trying to do a lot of, uh, training. Um, uh, both technical training and otherwise to help people understand, um, and we’re certainly providing people, you know, various forums and then chances to talk to each other.

[00:32:09] Richard Clarke: But that one’s harder, and I don’t think I have that one fully, uh, figured out yet. Um, and I think it’s gonna really be a bit of, You know, how H. R. also kind of evolves to, to help people think through their talent strategies that now incorporate A. I. into, into that. And I know our team’s thinking a lot about that, uh, but that’s definitely one that’s gonna, I think, see a big change over the next 12 to 18 

[00:32:33] Sean Ammirati: months.

[00:32:33] Sean Ammirati: Yeah, um, both of those are cool. Just, just on the first one, too, uh, for those listening, the next episode after this will Moderna who’s gonna talk about. A tool he built inside Moderna that’s a lot like their sidekick and, uh, interestingly, somewhat similar profile, I think, to the, to, to your colleague, like he was a data science product manager, build a bunch of, so, so I think that is a.

[00:32:58] Sean Ammirati: If you’re not doing that right now, you’re like, you should, you should think hard and a bunch of different vendors that can help you do that. You can roll your own, but, but thinking about how you put kind of enterprise versions of these things in your, in your company makes a ton of sense on the training thing.

[00:33:14] Sean Ammirati: You know, I think we are in the first inning here. We got to figure it out. Um, one thing I’ve found, and I’m curious your reaction to this, Richard, and then we’ll pivot to talking more about the future is. Vocabulary is really important for a lot of these people. Like I, I just think a lot of these executives, they kind of know, they’ve heard the terms, they’ve read them in the wall street journal.

[00:33:35] Sean Ammirati: But like, just getting them comfortable with like what these terms mean and what’s possible and what’s not at just a sort of 101 level, not at the deep level that you and your team understand it, I think is often a way to at least get people off the, off the blocks here. 

[00:33:51] Richard Clarke: I totally agree with you. That really resonates with me.

[00:33:54] Richard Clarke: And I would say not only, uh, which has been another change, not only the kind of C suite. Right. But heck, our board, I’ve done more education sessions with our board than I’ve ever done on this, uh, before because they just have a lot of very valid questions and really want to understand, like you said, the, the state of the possible and, you know, partially just so they can feel equipped to ask the right questions going forward.

[00:34:21] Richard Clarke: Um, and I think, um, that kind of seeking to understand is a really good sign, uh, inside of a company that they’re recognizing the level of value and transformation that this can provide. 

[00:34:32] Sean Ammirati: This is a board issue. Period. 

[00:34:33] Richard Clarke: Full stop. Period. Full stop. Yep. And I got a lot of the, uh, well, isn’t this just blockchain?

[00:34:39] Richard Clarke: And I would talk to a lot of people and be like, you know, we had zero board meetings on blockchain. Zero. We’ve had lots of board meetings on AI, so that in and of itself should tell you it’s, it’s different. Yes, 

[00:34:50] Sean Ammirati: yes. This looks more like the internet. So we’re going to create a bunch of pets. coms, but we’re also going to create a bunch of Amazons.

[00:34:56] Sean Ammirati: And that’s a, that is definitionally kind of a full stop, uh, board issue. So, Richard, I could do this all day, and I think I’ve actually already gone past what I committed to in terms of timing. But I do want to ask one last question, and then we’ll do kind of a quick wrap up. But when you look forward and think about the future of healthcare and AI, what are the things that you’re particularly excited about today?

[00:35:22] Richard Clarke: Boy, so many, uh, so many, but um, I’ll, I’ll force myself to give you a short answer. So, one, I am so excited to see how AI transforms. Uh, the patient experience. Yes. Um, I, I think how we experience, um, interacting with our clinicians and our care teams is going to be dramatically different. And frankly, I hope people demand that.

[00:35:47] Richard Clarke: Right. Um, it should be dramatically different, um, through ambient listening, through automatic documentation, through, uh, you know, Digital avatars or interface. I just think it’s gonna look totally different. So we can finally start to get to that always on kind of always available, much more kind of personalized interaction.

[00:36:09] Richard Clarke: So that’s probably the biggest piece of excitement that I have. Uh, I think the second thing I’m really excited about is there’s a tremendous amount of expertise, um, inside of the people that work in health insurers and health systems. And that expertise is often wasted on tasks that do not require that expertise.

[00:36:32] Richard Clarke: Yes. And, you know, this is going to allow us to really unleash that. Um, and hopefully bring joy of practice back to a lot of clinicians, bring, you know, joy of practice back to a lot of administrators that are, you know, burnt out by a lot of that. And so I think that’s going to be a tremendous, um, a tremendous change.

[00:36:54] Richard Clarke: And I think those two things are, they’re not like, “Oh, I hope it’s going to happen.” I really think that those are going to happen, and they’re going to happen soon. And it’s going to be pretty amazing when they do.

[00:37:04] Sean Ammirati: I love that. And you can tell you spent, you know, a decade in finance before, right?

[00:37:09] Sean Ammirati: Because financial services have, have, I think in many ways demonstrated this operate where you do the things that you’re good at and you let computers do what computers are good at. And I think healthcare is, we’re, we’re right on the cusp of a similar experience.

[00:37:25] Richard Clarke: We’ve, um, we’ve used very effectively this, like, you know, let’s enable people to work at the top of their license as kind of a, uh, a talk track that I think is really resonated and gets people excited because you’re saying, “How can machines propel my kind of work and expertise?”

[00:37:44] Sean Ammirati: So amazing. Well, Richard, I really appreciate you joining us today. To our listeners, keep track of, keep following along with what Richard’s doing on social media, stay aware of that. So to you, Richard, what’s the best place for someone to kind of keep aware of the things that you’re doing?

Join the discussion

More from this show


Episode 1