Name: Wollman Fierce Webinar V2
Uploaded: 2025-09-25T17:35:07.631Z
Duration: 30 min 55 s
Description: Wollman Fierce Webinar V2

Transcript for "Wollman Fierce Webinar V2": Hi. My name is John Wollman. I'm Head of Revenue Strategy here at Comodo Health, and we'd like to welcome you to our webinar today to focus on where AI can and should not be used in early stage commercialization. By way of introduction, I've been with Comodo Health for about four and a half years. I've been in the industry of data technology and consulting for life sciences and healthcare companies for a little over thirty eight years. So been in business quite a bit. And, while AI, a lot of people think it's new, I was a propeller headed geek in college. I was computer science major, and, I was coding AI capabilities in a language called LISP, which is pretty much defunct at this point, back in 1985 to try and diagnose challenges with with automobiles. So it's been around quite a bit, but, these days, it's really accelerating and some of the capabilities are so exciting and the advent of these large language models is gonna change a lot of things. About Komodo Health, for those of you that don't know about us, we believe that we're the fastest growing company in our industry. Relative size, we're a little over $300,000,000 in annual recurring revenue. We believe that we also have the best representation of the ground truth of health care in The United States in our health care map. It's the largest U R US health care dataset. It's the most representative. It's the most linkable. We are native on the Datavant tokens, so we can link to virtually anything. We have established QA QC processes. We serve, I think, 18 or so of the top 20 pharmaceutical companies in The US by revenue. We're incredibly transparent with our coverage. We share that with our customers every month. But we're not just a data provider. We also happen to have robust analytics platforms, and we're really at the forefront of AI. We've just released recently, product we call Marmot. I'm gonna use that to demonstrate some of the use cases in the industry as we go forward. So that's about my company. We're really here to focus on early stage commercialization. As you all know, this is a very high stakes and complex endeavor. As you're trying to launch a product or bring a product to market, essentially, you have limited real world outcomes for payers. In many cases, all you have is what you learned in your clinical trials. You don't have that real world experience yet because the product hasn't been released. Traditionally, the payers don't pay as much behind to you until you're approved because it is a high stakes endeavor and a lot of products don't make it to commercial. So you wind up with inadequate payer engagement before your approval, which is complex. And as we all know, getting a product through the regulatory barriers that exist either in The United States or on a global basis where there's a lot of inconsistencies, what the a care the FDA cares about versus the AMA versus other regulatory bodies. There's a lot of inconsistency and stress in trying to get a product through the regulatory hurdles. And really in in today's crazy the things that are going on in our market in The United States, it's even more interesting than ever. As such, the journey to get to commercialization, to get a product to market can last over a decade and can literally cost billions of dollars for some indications, especially where you're doing very substantial Phase three trials. That being said, with all of that investment in time and money, 40% of new drug launches failed to meet their two year sales forecast that were predicted before the drug was brought to market. And about a 100% of the launches, not just this is not us. This is from an industry wide study. About a 100% experienced some type of a delay, whether that's due to internal problems, regulatory challenges, or supply chain issues. A 100% of new drug launches experienced some type of a delay. So it's costly. It's difficult. Many of these fail to meet their the forecast and, of course, there's a lot of delays. Now many people are pointing to artificial intelligence AI being as poised to reshape the pharmaceutical industry. There's no question that this is a very significant area of spend right now. If you look at 2025, there's about $4,350,000,000 worth of investment in AI, and that's expected to grow by more than 40% each year up to by 2030 about twenty five point seven billion dollars. This is serious investment. And the main point I wanna get across today is that AI will believe it. Of course, we're using it in incredible ways within my company, and our customers are using it in very interesting and meaningful ways in early stage commercialization. What we believe is that AI is not gonna replace you, people, as the expert. Essentially, it amplifies and empowers what you do best. Now let's talk about AI within early commercialization. And specifically, I wanna focus on five different use cases or areas within early commercialization. We believe that AI does add tremendous value today. One is in market research. We'll qualify that in a minute. Another is in engaging with health care professionals and health care organizations. Market access optimization as you're trying to make sure that you're proving the benefit of your new indication in the marketplace and making sure that you get access to the market, in brand strategy and making sure that you're on point with how you're gonna go about bringing this product to market or this. If it's not a product, it's some kind of a medical intervention or digital intervention. And funding maximization, any company that's trying to commercialize that is an emerging biopharma company is certainly looking to make sure that they have adequate resources to bring the product to market. In market research today where we see AI adding value, one area is blending qualitative and quantitative insights for market understanding is being able to take what you know quantitatively and merge that with insights that you have qualitatively. It's also used extensively in patient journey mapping, understanding how patients are getting diagnosed, how they're getting on therapy, how they're being persistent and compliant to stay on therapy. It's very prevalent these days in landscape scanning and understanding what trials are being run, what pricing and policies are changing over time. And it's certainly being being very useful in the development of materials that are used, of course, in a compliant way with health care professionals, regulatory bodies, and health care organizations. In fact, one case study that we see that from CLIP, which is, obviously, a very prevalent agency within the marketplace, show that the confluence of human and AI copywriting for medical materials cut time by seventy percent with a ninety percent MLR success rate versus twenty percent for AI only copy. As I mentioned earlier, we do think that AI augments humans. It does not replace humans. And this kind of indicates that by leveraging AI copywriting with human intervention and humans driving can get your MLR success rate dramatically higher than AI alone. If you look in the area of KOL or HEP engagement, this also permeates HCO engagements from our account based selling. But if you look at the ability to profile your key opinion leaders and your health care professionals and predict how engagement strategies are gonna work, AI is really wonderful for this. And by the way, it's so many of these use cases. I'm actually gonna demonstrate for you using our tools, but you'll get the point. There's other environments out there that that are capable as well, but you're actually gonna see this in practice. Additionally, you can use AI very successfully in understanding and doing rapid secondary market research We're using claims data or EHR, EMR data or genomic data, social determinants of health data, lab data to be able to execute secondary market research very quickly, very rapidly, iterate through it, create hypothesis, prove that hypothesis, and then ultimately standardize that and bring it into a more hardened environment. AI is also very good at contact personalization for specific key key opinion leaders or health care professionals to engage with either a single ATP or KOL or a subsegment or a broader cohort of KOLs or HCPs. And, of course, AI is very good because of its ability to simulate things at looking at personalized messages and testing and simulating how they're gonna actually work in practice. We also see AI quite prevalent within market access optimization. If you look at your HDAs trying to create your health technology assessments, being able to predict what the decisions might be, what payer behavior might be, what formulary decisions might look like because of AI's ability to do predictive models and to look at different scenarios and be able to analyze how they might take effect. It's very good at this for HTA. It's also quite good at supporting the creation proactively of value stories, creating your studies and submissions and optimizing them so that you're properly engaging with either the payers or the regulatory bodies to prove the health economic outcomes and the benefits of your drug or your medical intervention or your digital intervention. We also see AI being used successfully within brand strategy development. And a lot of times as you're developing your strategy, you wanna be able to synthesize massive quantities of data to be able to look at patient insights and HEP insights at a very granular level and model things and simulate things. AI is very good at this data synthesization. I'm gonna prove this to you in a couple minutes when we demonstrate some of these capabilities and use cases. It's also quite good at supporting the development, creation, and approval of medical and marketing content, whether those are things that go through MLR or other types of content that you're trying to create. It's very good at creating drafts and being able to optimize or review drafts that you have and make recommendations, streamline things. It's very good at those kind of things. It's also quite good at driving marketing automate automation and personalization. If you think about the three steps of that, it's audience identification. It's then activating those audiences and then measuring the results, and AI is very good in all of those different steps. I'll show you some of that in a minute. And, of course, it's quite good at flagging compliance risks within content or within some of the other artifacts that you're creating and streamlining your MLR process that I indicated earlier. And then finally, as we talked about, the funding is obviously important during commercialization to make sure you have adequate coffers to be able to go through all of the different steps that you need to go through to commercialize. One of the things that AI is used quite frequently for is to match startups with investors, understanding who's in the market to be able to provide the capital that you need to bring your product to market. It's also quite good at speeding due diligence within an m and a process, being able to take the information that's available, synthesizing it, creating summaries from it, and now allowing the funders to make decisions very rapidly. It's also really good at pitch creation and evaluation of financial forecasting, the kinds of artifacts that you would create in order to enter into a process to get funding. AI can be used very successfully in order to streamline those processes, create very high quality artifacts that can lead to a better funding decision. These are some of the areas. There's more. I only have a half hour today where we spend a number of other use cases that could be used for. But in our estimation and in our work with our customers, this is where in early commercialization, AI is mostly being used today. Now I think it's also important to look at of course, it's interesting to know where AI can and is being used today, but it's also very important to understand where AI cannot really add as much value today, at least in today's world with today's technologies and today's large language models and capabilities. These are the areas where we don't think AI is gonna give you a lot of uplift or support. And, again, I'm gonna drill into all of these, but the first is navigating regulatory nuance. We'll also talk about complex market access negotiations, creative brand storytelling, and interpreting implicit adoption signals in emerging markets. If you look at the first one, navigating regulatory nuance, what's true is AI cannot replace the judgment and transparency and the understanding of very experienced regulatory practitioners. And if you look at a lot of what regulatory bodies are saying is that AI are is providing these blacks black box model and what's really required for regulatory bodies is transparency. I'm gonna show you how we're trying to get around that a little bit with some of our technologies, but it is a challenge and this is the kind of thing that humans are very good at and AI can't really accomplish today. Additionally, with regulatory bodies, human oversight is expected and required. I don't think you're gonna be able to expose your the output of one of your AI capabilities and give it directly to a regulatory body and expect them to make a productive ruling on your behalf in that today. So, eventually, perhaps, this will get more further along the journey as regulatory bodies get more accomplished with AI and as the industry does. But for now, navigating this regulatory nuance is not quite there with AI. Let's face it, that complex market access negotiations is a very nuanced process and it requires a lot of skill and experience to understand what the payers or the regulatory bodies are saying to you and be able to understand how to address their concerns and how to properly get them to a point where they're understanding the economic and outcomes benefit of your new drug or therapy. And, really, AI versus humans tends to lack empathy and stakeholder relationships. Although, it's funny you read some of the some of the rhetoric these days about people developing relationships with AI, large language models like chat GPT. There's really no substitute for interpersonal relationships. Additionally, in some cases, AI can lack the judgment, trust, and the ability to craft the narrative that will compel a payer or a governmental entity or regulatory body to move in a direction that you want them to move. If you think about brand storytelling and the kind of things that you want to resonate with the patients and providers and organizations that you're engaging with, a lot of that requires human crafted narrative. Differentiating your brands why not just what AI is very good at. The why in some cases it's good at, in many cases it's not that good at it. And as I mentioned earlier in the context of creating medical content to pass MLR, hybrid models tend to work best where you have AI does drafts and then humans perfect. And the last one there is interpreting implicit adoption signals in emerging markets. As we all know, AI works best when it has a lot of data to analyze. And if you're in an emerging market where you're delivering a new product, a new indication, there's not that body of evidence for the AI to introspect and to build opinions upon. So interpreting implicit adoption signals in the emerging markets, some of that is not so great for AI to be used for in today's world. So we talked about where AI is being used in use cases today where perhaps it's not quite as useful. But as I mentioned at the outset, I think the most important takeaway, if I can leave with you, is that it's not AI versus human, it's AI with human. And we believe that AI can help to accelerate strategic iteration without surrendering judgment. We know that you can use AI to challenge assumptions, but it's not particularly great at dictating a strategy. And we believe that what's most important is that US practitioners in your companies build flexible go to market playbooks that are based on AI augmented insights where it's augmented with human intelligence experience, empathy, and nuance. Well, I think about it as Nirvana in this context is that you combine AI's ability to recognize patterns and analyze vast quantities of data very quickly with human interpretation, empathy, and judgment. So if you think about that, if we look about those use cases, if we think about the use cases where AI is beneficial today, where it maybe lacks some capability today, the must haves that we believe for winning with AI is to adopt a human led AI powered model. And to do that, that requires a couple of things. One is technology and data investment. Of course, you're gonna need AI capabilities and frameworks. You're gonna need the large language models. You're gonna need agentic capabilities to orchestrate agents. You're gonna need lots of things there. But, of course, without a data infrastructure and a high quality, data fabric in order to feed and train your AI models and to serve your AI models with, you're not gonna get very far be it for a a data provider like me, to have that opinion. But we believe that it's very important to have vast quantities of really high quality data whether you get that from us or other providers, it's really important. And then, of course, you wanna make sure that you're developing ethical deployment and data integrity strategies to make sure that your AI models and your AI solutions and your agents are used in an appropriate ethical fashion within an industry that's as highly regulated as we work within. There's also the investment in people and process. We think we can't stress enough how important it is to prioritize cross functional alignment and governance. As an example, if you ask AI tools to analyze a market without providing you a market basket that's interested in governing within your environment, it could come up with results that aren't as meaningful to you as perhaps they might be. You also need to make sure that you're establishing clear oversight and human review frame frameworks both for when and where to use AI and where not to, and how to merge that notion of humans and AI is working together in concert to accomplish proper results. So I'd like to do now is illustrate some of those use cases we talked about before. I'm gonna do this in our environment, which is called Marmot. We released Marmot fairly recently. We already have a customer that's we're implementing it today, and we have a massive pipeline for this. But Marmot is our agentic AI capability leveraging generative AI and merging that with agentic. It's a full agentic model in that we built a number of agents, but we can also allow our customers to build their own or we could build them with them or for them. But what Marmot is built for us to be able to ask it a question, and it will give you results. So I'm gonna go through some examples of that. So right now, what I'm gonna do is I'm gonna use Marmot to illustrate some of these use cases and show you how these work in practice within a agentic generative AI model. Here's some examples. And, again, I'm gonna use Marmot to show these. The first thing to do is something you might do really early in commercialization, which is creating a target product profile. So you're gonna see here I'm gonna ask Marmot to create a target product profile for an oral small molecule drug that will be able to successfully enter the market in progressive forms of multiple sclerosis, then create a predictive model for the types of patients that would be most likely to use these drugs, create a predictive model for the types of HEPs that'll be most likely to prescribe it. And then from the predictive model, list the top HEPs that the company producing your drug should engage first. So I'm just gonna click on Marmot here. I'm gonna open up this this prompt and response. And here it is. And as I mentioned, this was the prompt I'm asking it to do. Now real quick with Marmot, I'm just gonna ignore this these projects over here, but I wanna show you that you ask a prompt. And what Marmot does, like any good generative AI tool, it tells you its analysis plan. So it's telling me to develop this. It's gonna look at the target profile product development. It's gonna analyze the current treatment landscape, identify gaps in current treatment options, compare efficacy, safety, dosing, and patient experience. It's gonna create a patient predictive model. It's gonna create an HCP predictive model, and then it's gonna generate a list of the HCPs. Now one of the things we have is called the research planner. This is if you turn this on, it'll ask you some leading questions based upon your prompt. So here it's asking me which progressive forms of MS should the drug target, just primary progressive MS, or should I look at primary progressive and secondary? I chose that. What key efficacy endpoints should be prioritized in my target product profile? I'm gonna balance disability progression with the other endpoints, like brain volume loss, cognition, and quality of life measures. How do we define early adopters? I have some options here. Of course, you can just accept the default here, but it prompts you based upon the context, which is pretty cool in and of itself. And then based upon what I said, it's gonna go about looking at this target product profile for an oracle oral small molecule drug for progressive forms of MS. And here it's done that. It's developed a product overview here. It's identified the clinical need and opportunity in the market based upon Komodo's real world data, looking at approximately 715,000 MS patients, and it's telling us where these patients the average age, the progressive MF MS patients have slightly more complications versus one point two for relapsing, higher proportion of progressive patients have Medicare coverage. Then it's telling me about the product differentiators that might be here, the mechanism of of action, the clinical efficacy targets that we should look for based upon what's in the market today, its primary and secondary endpoints. It's looking at the safety profile targets that we should strive for based upon what's in the market and etcetera, etcetera. So you can see that with just a simple prompt, Marmot is giving me just about everything I need to know for my target product profile. I do wanna share something over here on the right. You'll notice that it's showing here every single analytic step that Marmot went through to arrive at these conclusions. As I mentioned earlier, one of the inhibitors to AI adoption and usage today is a lack of transparency. We think we've solved for that because every single analytics step that Marmot went through is in here. It tells you every single thing it does in order to arrive at a conclusion and even gives you notes about why it made that decision. You can see here as I scroll down as well, it also generates code. So it's showing me the code that it generated. So if you ever have to reproduce an analytic that Marmot did for you, assuming you have our data, you'll be able to reproduce that analytic. So every single step is in here, and we can output that to a Jupyter notebook. You can run it in other environments. So Marmot addresses a lot of those transparency concerns. Let's look at another use case. Let's look at adherence and discontinuation. So in this case, I'm gonna ask Marmot to evaluate the persistence and adherence patterns for patients that are on current lipid lowering therapies, particularly focusing on combination regimens, identify the rates of discontinuation of three, six, and twelve months, the reasons for discontinuation when that's available, and what patient or provider characteristics might predict better adherence. Again, the kind of thing you might wanna do if you're launching into a market, it's a competitive space, and you wanna understand what are the existing persistence and inherent patterns for drugs that are already on the market. Again, Marmot creates an analysis plan. It's gonna create the patient cohort. It's gonna baseline the patient characteristics. It's gonna do persistence calculations. It's gonna do adherence measurement. It's gonna look at reasons for discontinuation, create predictive factor analysis, and then visualize and report everything. And, again, it's gonna ask me some questions, which lipid lowering drug classes should be included, how should combination therapy be defined, what gap in therapy should constitute discontinuation. One point I wanna make here is I didn't have to tell it what persistence and adherence, what that means, how to analyze that. It's smart enough to understand what that means and to come up with an analytic here. What time period should we use for the analysis and how should we identify new therapy initiators versus ongoing users? I accepted the defaults here and then it goes about its business. And really quickly, it was able to look at the persistence rate comparison of monotherapy drugs versus combo therapy at three, six, and twelve months intervals, and it creates a nice graphic along with the table here. It looks at the adherence rate analysis by lip lipid lowering drug classes. I didn't have to tell it any of this. It knows or it interpolates what are the different drug classes for this particular therapeutic area and how do the adherence rates in PDC differ across these different regimens. It looks at adverse event parent patterns and lipid lowering therapy discontinuation. So when you're either on a combo therapy or monotherapy, what are the main reasons why people discontinue the adverse event types, muscle pain, GI symptoms, memory issues, liver issues, and drug adverse events, and you can see what those those rates are. And then it looks at age and persistence a little bit lower in therapy. Is there anything going on here that would indicate, as you can see, as you move other than between this dip here, tends to get progressive as people get older, they're more persistent. Looks at insurance types by insurance groups so you could see where those are, and then it gives a summarization of all of this. Another use case where AI is quite good. Another thing people do quite a bit these days is looking at prescribing patterns. So just like staying in this little bit lowering area, we wanna identify the top 200 health care providers who most frequently prescribe PCSK nine inhibitors to patients with elevated LDL C despite maximally tolerated statin therapy, analyze prescribing behaviors, patient populations, blah blah blah blah blah. And, of course, we can use AI here to do that. By the way, having been a consultant in this space and having built and sold to consulting companies, we used to charge our customers. It would take us about a month to do this type of an analysis. We project probably charge them $200,000 or more, and this is the kind of thing that you can do right here in Marmot in minutes. So, again, it'll look at my question. It'll create an analysis plan. It'll ask me how do we define elevated LDL? Is it greater than 70 for highest risk patients with ASV ASCVD or diabetes and greater than a 100 for other patients? How should we operationally define maximally tolerated as a high intensity statin, like the highest dosing for atorvastatin or rosuvastatin or some other metric. And, again, I accepted the default here, and here it goes about doing its business. So it's looking at this PCSK nine prescribers, analyzes whether they're in the academic or community settings. You can see largely in the community. It analyzed the top 200 PCSK nine inhibitor prescribers, shows me their specialties here and what the distribution of those specialties are. And then here are those providers. It's showing me their unique patients, their practice setting, their name, their NPI, their state, etcetera, etcetera. So with very little work on my part other than to create a very interesting prompt, it's gone about answering that question, which you might do a lot in early stage commercialization. And, again, every step that it take is is logged here. It's auditable and it's available to you. And then another thing you probably do a lot or we do a lot, you do, I'm sure we all do, is looking at patient journeys. One of the things you might wanna look at in this particular therapeutic area treatment journeys of patients that are that have hypercholesterolemia, who have failed or are intolerant to statins, identify what treatments they receive after statin failure, how long they stand each therapy, what percentage you see their target below seventy mg per DL within twelve months. Then they do an analysis plan, and here it comes up with the treatment journey analysis for statin intolerance first people that discontinue with a non statin switch. So it's showing me my different categories of patients for documented statin intolerance versus statin discontinuation with non statin switch, showing me the patients within there and their prevalence. And then it's giving me the treatment journey analysis across lots of different things, identifying where they receive their targets and where they don't, what the treatment durations look like, time to treatment. Lots of interesting things that you might do in your early stage commercialization efforts. Now I didn't show you that to demonstrate our product. If you wanna see our product, please come by our booth. I can't say that in the webinar. I didn't demonstrate that to you just to be an overt demonstration of technology. I was really trying to show you how these use cases can be enabled and fully fulfilled with today's technology, especially if you have data like we have and technology like we have with Marmot. So the bottom line with AI, as we know, early stage commercialization is high stakes and complex. We think that AI offers potential to augment, not replace human strategy. In this context, AI provides the intelligence, the speed, the scale, the insights. Humans bring the empathy, the vision, the nuance, and and the experience. And knowing when not to use AI may be just as important as knowing when to leverage it, try to provide some color on that today. Additionally, we think the future of commercialization, early stage, late stage, all throughout is not AI versus human, it's AI plus human. The intent is to make sure that you augment your team strengths with automation, but don't try and automate judgment. Your team is really good at that. And of course, success depends on a lot of cross functional planning and a very high quality data infrastructure. If you don't have one, we're happy to provide one for you. But the key here is to plan early, build trust, execute deliberately as a team cross functionally, and make sure that you're empowering your teams to leverage AI responsibly and ethically for patient impact. Thank you so much. We hope this has been insightful for you. If you have any questions, please feel free to reach out to your friendly neighborhood Komodo representative. If you can't find one or don't know one, feel free to find me on LinkedIn and hit me up. We'd love to talk to you about your specific use cases and how we can help you in early stage commercialization.