Episode Transcript
[00:00:00] Speaker A: The volume and complexity of clinical decision making that health plans need to handle has increased dramatically. There's more specialty drugs, more therapies, diagnostics. The infrastructure that these plans rely on simply wasn't built for that reality.
[00:00:19] Speaker B: Welcome to Inside the Plan, a podcast featuring real conversations with healthcare leaders on strategies shaping health plan transformation today.
Welcome to A and M's Inside the Plan. I'm Craig Savage, managing director and leader of Alvarez and Marcel's health plans and managed care practice. I'm joined today by my co host, Brian Komornik, managing director in AM's Health Industry Group. Our guests today are Caleb Gian, General Manager of Cohere Health, and Gigi Un Reed, Cohere's Chief data and AI Officer. Today we'll discuss how connected clinical intelligence can help health plans manage rising medical and and administrative costs while modernizing the UM function. Caleb and Gigi, welcome to the podcast.
[00:01:07] Speaker A: Thanks for having us. Great to be here.
Craig and Brian, good to see you.
[00:01:12] Speaker B: Good to see you.
[00:01:13] Speaker C: Thank you for having us.
[00:01:16] Speaker B: Caleb, let's get started with you. Health plans are under pressure from rising medical cost, administrative burden and margin compression. Which one or two forces are really creating the most pain for plans today? And secondly, why is the status quo no longer sustainable?
[00:01:35] Speaker A: Yeah, that's a great question.
It's an interesting dynamic and there's a lot of things changing and moving right now. I'll just call out maybe two things I think. First is the volume and complexity of clinical decision making that health plans need to handle has increased dramatically. There's more, especially drugs, more therapies, diagnostics, and there's increased demands from regulators, frankly, the providers and members too, which health plans have to contend with. And that's something health plans are feeling kind of more acutely today than ever.
And with that, the infrastructure that these plans rely on simply wasn't built for that reality. So over the past two decades, plans have kind of invested in a bunch of point solutions around prior auth care management, payment integrity, STARS, HEDIs, RISC, et cetera. But they were built independently to solve each of those different problems. They weren't built in a connected fashion that lets you actually truly understand the patient across all the different things the health plan has to do. And that's where I think you have to start to think less about how do you just upgrade a point solution and more about a model that actually has what we call connected clinical intelligence that is getting better and better with each interaction between a health plan and a provider, between a health plan and a patient, and learning from all those reactions, all those different interactions in order to really become an intelligent health plan versus just a health plan operating the different functional areas that will then lead to kind of the member outcomes. You want to see the reduced abrasion, you want to see reduced medical spend, etc.
[00:03:20] Speaker D: Caleb, I think that's super fascinating. Thank you both for being here. Gigi, from where you sit as chief data and AI officer, what does that same pressure look like on the data and AI side?
[00:03:33] Speaker C: They actually follow a very similar pattern as what Caleb mentioned earlier when the use cases and toolings were developed in silo.
That typically is reflected in both the data and AI architecture. So a lot of helplines struggle with fully seeing a member360 knowing how when we interact with a member on a simple inquiry, that inquiry staff may not be able to see what's happened with prior auth or what's happened with some of the claims records and that can create a lot of member ablation. And similarly as with really thinking about how a health plan engage with a provider, that incomplete information can create a lot of unnecessary back and forth.
So that's the data side and I think on the AI side very tempted to say hey, let's just put an AI automation in place to help what we do go faster or cheaper. Like a classic example could be, oh wow, now providers are using AI to automatically respond to all the appeals cases. That's been getting a lot of news.
A very natural response to say okay, let's see, we can put AI in on the payer side and figure out how to as quickly respond to these appeals as possible.
But that may sound great, but it actually does not change the cost structure and neither does it change the provider relationship.
So what we've been working with our customers a lot is to think about the yes, the member 360 view but also thinking about how can AI really be used as a transformative tool as opposed to making existing workflow faster. Right. We can think about hey like wouldn't it be great if we can understand truly the member needs early proactively engage with the providers through agents to seek additional information to make sure the first ask is the highest fidelity and reducing any minute information gap then you pretty much cut out all the unnecessary back and forth. Right? That is how you foundation you change the game.
[00:05:38] Speaker D: I think that's super, super thoughtful and I think what I'm hearing too from you is it's more of that orchestration layer to maintain the trust, maintain the relationship versus and build what should be a trustworthy system versus just the transaction or solving kind of that point solution that Caleb was also discussing. So, so super thoughtful Gigi.
So Caleb, then when shifting I think back to where health plans have made the investment in technology you mentioned earlier, the point solutions and those investments, whether it be point solutions or an integrated platform or suite of kind of best in class capabilities, whether it be prior auth tools, care management platforms, claim systems and back office and core admin, often over time those have been developed to meet different and evolving needs of of the business. But you know, as a result and things that we see, plans are running really largely manual disconnected workflows and really this is starting to talk more about an operating model discussion. But what does it look like when that clinical intelligence becomes more connected end to end and what's the biggest unlock from that you would see plans getting with that sort of integration?
[00:06:52] Speaker A: Yeah, I mean when the unlock is that when you actually connect the different pieces, your outcomes improve practically what that means is that I'll just give a few examples in the appeals space. For example, if you run appeals in a silo and you don't understand specifically in utilization management or in payment integrity what happened upstream and then why the case ultimately was denied and then why the provider appealed, there's no feedback loop. So you can provide process the appeal and you can overturn it. But you're not actually taking the insight from that and bringing it back into how you operate your other functions and what it's doing is causing more member abrasion or more denials that ultimately are overturned. Nothing that anyone either side of the healthcare aisle is looking for. Another good example is plans are sitting on data for utilization management that's not being used to help better drive quality HEDIS measures with stars and you can actually, if you think about the connecting the different kind of use cases and you take that data and then you analyze it and take the unstructured data and synthesize what's needed. You can use that to then drive more STARS outcomes and not have to go back to the provider to get that data already. So you're, you're creating a smarter system that ultimately it's learning from each different interaction. The hard part, like you said Brian, these plans have invested in the workflow systems over a long time and those are built up in unique ways with each function, with all the different requirements for each specific health plan. So I think often what's best is not to sit there and say let me get to a nirvana state and completely replace the operating system of the Entire health plan like that just. It will take way too long for that to happen. Instead, let us figure out how to put that intelligence on top of the systems that exist across the different workflows, but that intelligence that's connected end to end. And when you do that, you can kind of reduce your activation energy needed to go install these types of intelligent systems while kind of working with the investments that a health plan's already made. What's required though is a lot more of the change management, the process engineering. And Brian, what you alluded to, the operating model conversation of how do you actually rethink your process end to end versus just kind of what we call in coherent terms, retrofitting AI onto what exists versus kind of redesigning how that process works so that you actually rethink and things change.
You can touch 90% less cases and really focus on the 10% that matter versus shoving into the same process and getting a 5% improvement on each of the different steps.
[00:09:36] Speaker B: Yeah, yeah, absolutely. That retrofitting and gigi. I want to come back to a little bit of the conversation around data and AI and the natural inclination to think of AI as just another automation tool. The industry has been talking about modernization and transformation for years. Give us your perspective on what has really been the headwinds, what have held plans back in making that transformation and what's different today that makes transformation, transformation more urgent, number one, and secondly, achievable this time.
[00:10:13] Speaker C: Yeah, yeah. And I think both you and Brian alluded to it's modern tech, right? It's a, it's mindset, right? It's success metrics is workflow, right? The classic statistics where 1 in 10 AI experiments actually go into production and 1 in 2 of the AI that goes into production actually drives business value. So we're talking about 5% success rate of investment.
So I think that's actually a lot. That is not the technology itself is how we think about what is the organization strategy. To Caleb's point, right? Thinking about workflow transformation, how do you set the right metrics for the AI so that you know when it's doing a good job and when they should stop trying to do a good job and hand it off to a human. And then how do you get the rest of the staff to be AI savvy? All those pieces have to come together for AI's full potential to be realized. Business goals, workflow, human readiness, all those pieces have to come together.
It's very, very easy, especially from a technologist standpoint, when you're working with The CTO working with the CIO to be very much focused on what infrastructure do I put up? Do I use this frontier model or do I host my own frontier model? Those are all really important conversations to be have. But I think what's really slowing organization to really drive the transformation is let's keep in mind that while health plans function seemingly as a financial institution doing a lot of administrative work, but most of the decisions that we're asking AI to help us facilitate are clinical in nature.
So really bringing together the clinicians and the scientists and the engineer mindset, the golden triangle to me is how you can break through the 5% success rate. Is that dream team because we had to take into account clinical judgment, we had to take into account of the operational adoption and promoting the trust in the whole process.
[00:12:22] Speaker B: Caleb, let's build on the success rate in the results that GIGI has been referencing and yourself. So when connected clinical intelligence really, really works well, what changes actually in practice? What are the impacts that health plans should expect in terms of medical cost, in terms of administrative efficiency and how, I mean what are their expectations around improving the provider and member experience?
[00:12:51] Speaker A: I think the first, I'll start with the last first because it's often the most important.
The member and provider experience can go up drastically. There's a lot of issues in healthcare today that are what I would almost missing information related, just a lack of being on the same page between the health plan and the doctor and that ultimately impacts the member. And there's care delays with prior auth, there's payment reconciliation issues with um claims even just matching the two. It's actually really tough and so providers don't get paid even though they thought that it was agreed upon.
So I think what we have seen is actually a really. It's kind of flipping the game on its head when you.
A lot of it in the old world was not necessarily by design but just based on how these different functions worked. Really kind of opaque, hard to understand and black box kind of thinking. And now it's like really transparent and it's more modern consumer thinking you see in most of the industries that have kind of revolutionized over the last couple decades. And so the first thing is providers and members working with a health plan that have connected intelligence. Everything just gets really clear, you know what's coming, you know what you have to do. And it's not these like litany of legal documents. It's right built for a modern age that's simple to work with and operate with. The second is you do see significant administrative cost reductions because you can focus on the work that really needs clinical judgments, needs clinicians, needs experts that know the domain. And you can automate all of the simple stuff. And so you need less human power work, workforce doing stuff that kind of grunt work versus focusing on how do I actually get this patient the right care and how do I understand the guideline in a way for this patient who has diabetes or whatever the different condition is that they're, they're contending with. And then third is you do actually see, when you really don't just talk about intelligence, but clinical intelligence, you start to see medical spend move in a way that is favorable, frankly for the patient and the health plan and the doctor. Because you're not, you're not sending patients to care that's not optimal for their specific path and that drives better outcomes and drives a patient getting the optimal path of care for them, which, which often times is actually lower cost, not always, but it can be. And so those are the three main pillars. I think the experience side of things gets a lot better.
The administrative side of things gets easier for health plans as they scale and support more members at a lower cost basis. And then the medical spend or quality improves when you bring more and more sophisticated analytics and intelligence that's learning more and more about each patient each time you interact with you change doctors. Okay, let's take that information over. You come from a different health plan. Let's take that information over that all kind of compounds to drive a better outcome for the patient at the end of the day, which that's the top goal at least that we are focused on.
[00:15:51] Speaker D: Gail, I think that's great and both you and Gigi, the change management aspect of this paired with the constant human in the loop and governance, this is still further supporting experience while also having the right type of alignment with the incentive of costs overall. So, you know, can't agree more there. You know, Gigi, I'm going to come back to you and continue to kind of build on this shift. But as health plans push AI further into those workflows and into data and the governance foundation underneath really becomes mission critical and matters more than ever.
What are you seeing plans do to kind of continue to talk about that governance?
What are they doing well and how are they beginning to navigate through when it comes to taking that foundation and beginning to use AI at scale, increasing that success percentage from proof of concept to actually push to production. What are some of those lessons learned that you wish should be more broadly applied as we Continue to evolve here?
[00:17:00] Speaker C: Yeah, that's a great question.
Let me break it apart into two pieces.
Let's think about AI governance and then let's think about the data that drives the AI from an AI governance standpoint. We're seeing a lot of health plans who really investing in the right kind of both legal and technical resources to make sure that doing the buy versus bill and all those processes to be thoughtful. So that's very encouraging to see.
It is particularly challenging because every state is slightly different, every region is slightly different. Seeing health plans putting an emphasis on that, it's great.
But one thing I would caution folks to do is to have a one size fits all approach, right?
Regardless of what the use case is for, if that shall pass a thousand criteria in order to get a green, it really should be a risk adjusted governance model. What I mean by that, I'll give you a very good example that's true. Risk adjusted governance model Within Korea Health our AI governance posture is extremely red tape.
We self guarded it. When there's no human in the loop, anything that goes into production has to go through frankly 100 point checklist. But when we are doing something internally with a human in the loop, building something out for the first time, that checklist looks very different. So it's really a progressive checklist and it may never get to 100 checkpoints because we always have a clinician in the loop and there is downstream guardrail. So just really maturing the whole AI governance model from having nothing to having everything to then think about a tier structure. So that's one on the data front, like garbage in, garbage out. We don't have to convince anybody of that. But I think in the world of AI, especially in the world of agentic AI, for us to really understand what is hallucination, what is biases, we need to know what is grand truth, right?
So I've seen health plans who think about systems of records in new ways, right? It's not systems of record about a member or systems of record about a provider. It's systems of record about a member and all the clinical decisions that were correctly made around this member. But that's those ground truths that are typically scattered across different systems and they can still sit different places. We just need a way to bring them together at the time of, you know, agent reasoning and being a lot more prescriptive in tracking lineage. Because if you don't give the agents the metadata about who offered the data, why was it captured, when was it captured, when was it audited, all those Information needs to be readily accessible for the agent because you want to make sure that the agent is not looking at a piece of information captured for a very different purpose or at a very different time frame.
So I could go on and on. I think the context piece is just really, really significant.
There's a lot of unstructured data that is not patient centric like medical policies, state guidelines.
We need to think about them. We need to manage these content as first class citizen because without those then the agent only sees one side of the decision making. All the member data, provider data. They don't see all the important clinical, medical and financial context. So all those pieces have to come together.
[00:20:25] Speaker D: Yeah. And the aging and monitoring and maintenance of that over time has been bolted on through generations of leadership and technology evolution. So couldn't agree with you more there.
[00:20:34] Speaker B: Yeah, I think we can all agree health plans are certainly navigating. I mean it's unprecedented. That word is probably used too often. But the level of uncertainty that exists across all lines of business that's impacting health plans today is just, it's enormous. At the same time, you've got CMS.
CMS's interoperability and prior authorization rule has really put modernization front and center ahead of the 2027 deadline. Gigi, just coming back to connected clinical intelligence, how does that fit within a broader health plan transformation? And which capabilities should health plan leaders really prioritize today to maximize the impact and the success rate of that transformation journey?
[00:21:25] Speaker C: That's a loaded question.
Let's see how we can take that apart. There is a lot of uncertainty.
I'm more excited than ever about the CMS interoperability. I think going back to knowing where the data comes from, knowing who's the author, is it trustworthy? A lot of these upscaling is going to allow us to get the data by default, which is super exciting. I would say. Places to invest in is going back to the notion of clinical context. A lot of these interoperability standards are amazing and it leverages really great standard from fire and so forth.
But one big word of caution is when we, when we try to simplify things, it's easy to lose context. So let's make sure that as payer and provider interact with each other, those clinical context doesn't get lost in the simplification process. A couple of things we're seeing payers who can do a good job of is referencing back to the right clinical informatics strategy, referencing back to the right medical policy strategy and then to Caleb's point and let's lose sight of the it is the ecosystem. Every data kind of builds upon each other as opposed to one API call. It's just one handoff is how each API call built itself into that member360 and getting that feedback of what's working and what's not working. To me Those are the two pieces Clinical Context and the Iterative Member 360.
[00:22:58] Speaker D: Awesome, Gigi. I think that's great and I think it's often forgotten too in speaking about Interop is where the plans and what's required of them versus the providers and how are we getting both groups when it comes to data in the same sandbox and really sharing those capabilities and the, and the alignment of the value.
[00:23:16] Speaker C: But it's funny you say that because you were just working through, hey, like we need to make sure that we capture. If you say no to something, let's make sure you say no with really good reasons why. So. So whoever it is, the patient or the provider has to get a yes. But believe it or not that, you know, pay and provider are not looking at the same taxonomy today.
[00:23:37] Speaker D: That's right.
[00:23:37] Speaker C: Getting them on the table and say given the clinical context is the taxonomy. That makes sense. It's a clinical informatics investment that we have to make as a whole industry that is not just one payer and one provider alone.
[00:23:51] Speaker D: Yes. And I think not all providers are created equal. And so as we shift around now, Caleb, I'd love to transition to you and get your perspective from your chair. What is the next frontier in health plan transformation? And when we think of leaders, when we and laggers over the next few years, what's going to separate those two groups? What's differentiating? And then lastly, just what are you excited about from the work that you guys are doing?
[00:24:21] Speaker A: It's also a good question.
I think one of the toughest things in deploying these intelligence and AI systems is it's really hard to see the value come through in an incremental pilot and just slapping kind of AI on top of what already exists. Going back to the language I was using around retrofitting AI into what you have, you can do that. I think the magnitude of value you see might be sub 10% versus if you, you actually take a swing, take a bet and say, let me work with the team I have internally with a partner, whoever it may be, to really transform the process. But doing that is, it's a, it's a harder thing to do. You don't have to replace your underlying systems and engage on this three year IT project that takes forever, but you do have to actually say, I'm going to invest in the appropriate change management, really process engineering the domain expertise that's really key in all these use cases. The one thing we've learned is like you can't just do this with AI wrappers that don't understand the domain, the nuances of payment integrity coding or risk adjustment or hedis rules and measures. There's so much depth in these areas. So you need partners or teams that know the domain deeply, are willing to engage in kind of enough of a swing or take a big enough risk to actually reinvent what happens, redesign the process versus just put the little incremental pilot per se on. And I think that's, that's what we've seen. I think the plans furthest ahead for us are the ones that said, let us take a bet. We know it's going to take a year to put the system in place, but then not just put it in place, optimize it. As the data comes in, it does get smarter, it learns. I think that's, those are the plans that will be honestly orders of magnitude ahead of the ones that say let me try three pilots and three use cases with people that don't know the domain super well and you just see incremental value.
I think what I'm excited about a lot of different things, but ultimately I think this intelligence type of system is leading us to a world where there's a lot better data on patient care and getting patients to the right care. And the experience which everyone in has families are or whatnot, has dealt with the insurance companies over the years. And it's never an easy process. And so I think that the care delivered to patients will get better because of that data and intelligence system. And at the same time it won't be this frustrating process working with health plans where you don't know what's going on. It's a black box, it's opaque, et cetera. And I think that's a much better world for all of us to live in.
[00:27:07] Speaker B: Caleb and Gigi, thank you so much for joining us today and sharing your perspectives. Today's discussion really reinforced that modernization is not simply about adding more technology. The importance of data governance I think was really critical in this discussion today. It really is about connecting that clinical intelligence across workflows, strengthening that data and governance foundation and delivering measurable improvements in cost, efficiency and experience.
Before we wrap up, I'd like to give each of you a chance to share your final Thoughts And Gigi, let's start with you and then go to Caleb.
[00:27:48] Speaker C: I'm very optimistic.
I think when we face the challenges we always step up. That's how it works and I'm seeing that kind of popping up all over the place and people willing to adopt and embrace innovation. So I'm super excited about that and I do truly believe that AI, while AI itself is not a strategy, but AI as an enabler is truly a way to completely shift the cost structure. Right. We are level upping and how we, we're not no longer locally optimizing. We can really operate in a very different level.
Member transparency, better communications amongst payers, providers and patients. All the things that Caleb referred to earlier. It's completely within our reach.
Yeah, that's awesome.
[00:28:38] Speaker B: Caleb, over to you.
[00:28:40] Speaker A: I don't have a ton to add. I think I agree with Gigi's sentiment, optimism, super optimistic about where, where things are heading in the industry.
I think there's a, the, the technology advancements create a world where things can get a step function better than they were and so that's exciting and it's always fun to be on the precipice of kind of massive change. I think that being said, it's important to come into each of these knowing we're honestly only ending probably two or three still really early. There's a lot to figure out. So I think we come to the table with a measure of humility. Knowing each of these use cases, there's still a lot of unknowns, a lot to figure out the right course forwards and it's going to take kind of no one company, team person to figure out. It's going to take everyone working together through these tough problems to get to ultimately an outcome that's better across the board for patients, providers and the health plans. So very excited for where we're heading.
Awesome.
[00:29:38] Speaker D: And Caleb, Gigi, we can't thank you enough for your time. I think I'm energized and excited as well.
And really I think a great takeaway here too is that without redesigning really the work end to end and being consider of that with keeping the right type of humans in the loop constantly, AI has that chance to increase complexity, not reduce it. So that operating model shift is not a technology problem, it's, it's a people problem that we're going to solve and we're going to continue to do great things with. So can't thank you enough for your time today.
[00:30:11] Speaker A: Awesome. Thank you both for having us. Craig and Brian, great to see you. Great to see you both.
[00:30:16] Speaker B: Gigi, nice meeting.
[00:30:17] Speaker C: Thank you so much.
