On-Demand PMSA Webinar: Proactive Pharma Intelligence—A Practical Blueprint

Alright. Let's get started here. First of all, welcome, everyone. My name is Suhari Jagannathan. I am the head of advanced analytics at UCB Pharma in Atlanta, Georgia. I also serve as a member of the PMSA board of directors and the chair of education and research committee. Thank you for joining us. So today's webinar is sponsored by Tellius, and the title, for the webinar is From Reactive Reporting to Proactive Pharma Intelligence, A Practical Blueprint. And so I wanna personally thank them. And, we have two, members from Tellius that are going to, colead the presentation today. So the first one is Chris Walker, head of product marketing with Tellius, and Nick Pinero, VP of technical solutions, with with Tellius as well. So, Chris, why don't I, hand the mic to you, and please take it away. Yeah. Thanks, Shuhari, and, thanks the for the whole PMSA team for for helping out organize this. So, yeah, good afternoon, everyone. I'm Chris. I run product marketing at Tellius. I'm joined by Nick Panera, our VP of, analytic solutions. And I wanna just set the scene here for this this webinar with a couple of numbers that I think are really, really fascinating, really telling. So this is kind of the state of the play of the types of queries and capabilities that commercial pharma leaders that we talk to on a daily basis. This is sort of their their journey that they've seen over the last couple of years. You know, in twenty twenty four, you were the question was whether AgenTik AI could actually do analytics work. Like, is it gonna work? You there's a bunch of proof of concepts, lots of, you know, watching of demos, and and sometimes watching those proof of concepts fall apart when you point those to real data. Twenty twenty five, this is more of a year when people were starting to question which use cases are worth piloting. This is more discovery. I think now in twenty twenty six, this is where you're actually getting into which AgenTic workflows actually hold up in production, and what can I deploy today? Right? This is we're getting to a more of a maturity, in this space. That that stat at the bottom is telling. The seventeen percent of org this is this comes from the recent Gartner, CIO survey earlier this year, which which states that seventy percent of organizations have deployed AI agents, yet, more than sixty percent of, of of leaders are planning to deploy this in the next two years. So this is what we're gonna talk about here, this gap between the seventeen percent that have it working today and the sixty percent who are planning on having rolling this out in the next two years. Basically, the train is is is in the station, and we're you know, you don't wanna miss this. This is a really key moment, for commercial pharma. So, Nick, I'll hand over you to walk through this pattern that we're seeing for production ready systems that you have to really nail to get, AI and AI agents to be really helpful in in pharma. If you're able to come off mute there, Nick. Yes. So sorry about that, Walker. Yes. So what you're you're seeing is that every and when it comes to to what how we're rolling this out and when it comes to for you all to think about the agentic reasoning pattern, the it's really this this three step pattern that if you look at what it really takes to roll this out and what should your expectations be for having this type of technology when it comes to AgenTek AI. The first is ask. So your expectation of this AgenTek AI system that is production ready should be able to ask questions or your users should be able to ask questions. Conversational questions, answering ad hocs that come in in a natural way that where you get the correct results, and they're answering the user's questions. So the second part is around why. Now in addition to asking your descriptive types of questions, the next part would be well, now I want to understand the why. So it should also be able to answer diagnostics types of questions, doing a deep dive in terms of what's driving performance. And this is the the circuit the second pattern that should be part of your production ready agentic system. The third is the is answering being able to basically act. And so what we mean by act is give you a recommendation, not only give you a recommendation, automate that analysis. So the the deep insight, it's a onetime investigation. The when you answer why questions. But what if you want to be able to schedule this and automate this workflow so that it's sent to you in a in a proactive way. So Monday morning, you wake up, you have, hey. Our competitors growing in market share. Here are the reasons why. Or if you build a, let's say, a Friday refresh summary that gives the business the state of what is going on. Here's what moved. Here's what's up. Here's what's down, as well as a PowerPoint deliverable that is sent to you. So this is really a three part framework, asking questions, understanding why, and then acting, and that really involves the automation piece of it. So the the gut check for you all is that if you ask alone a a a a BI chatbot, then you're looking at, hey. Some chatbots can only answer the the first part, which is act. And then you look at self-service in terms of answering why questions. That is the the second part. So asking and understanding the why. Then the third part, if you combine all three of these, that's where you get a true AgenTic AI production ready, system. Thanks, Nick. Oh, yeah. And and, yeah. This slide is kind of this is yours, Nick. Sorry. Yes. No worries. So now, so here's where I want to get practical for a minute. So everybody in this in the audience, they're they're all hearing the pitch that goes something like, hey. Just put Claude on your data. Stand up an LLM wrapper, point it to Snowflake, and you're done. So I've heard it. You all have heard it, and we're hearing that from multiple customers. And on paper, it sounds reasonable. The model is the hard part, but and and everything else is just plumbing. So here's what the plumbing actually is. So this is why if you if you look back, hey. When iPhones were released, people thought the differentiator was the their touch screen, but it was actually the OS that Apple had been developing for twenty plus years. So it's the same thing nowadays. You have all these companies coming out with just that touch screen, just the LM wrapper. But when it comes to scaling us into production, that's where it's a a different story. So if you look at the five layers here, it's, and and, again, this is not because we made we made them up, but because when you take a real commercial analytics question, you type it into a chat box and expect a a trustworthy answer back. Every one of these layers has to to be there. It's not optional. It's not a nice to have. It's it's there. So when you look at the stack, at the top, you've got the the business layer, or you have the layer where like, what's actually what where the users are coming into the platform. And so that's the delivery in UX. It it sounds simple until your brand team wants the answer rendered in their PowerPoint template, in their brand colors with the footnote in the the the right place. That's an engineering project on its own. Now underneath that is an agentic orchestration layer. So this is a layer that nobody sees and almost everyone underestimates. So real commercial questions aren't one step. You know, why did market share drop in the southeast? It's it's a plain query or you create a plan, you query, you diagnose, you reflect, and then you maybe replan. If your orchestration can't self heal when one of the steps fails. The system then silently hands you a wrong answer. And you don't you don't know what's wrong until you put it into someone's hands, your your VP of of of the field sales team, and then they look at the numbers and it's wrong. Then you have the third layer, which is the retrieval and knowledge. This is where your semantic layer lives. So understanding what TRX means versus MBRX, how payer rolls up, what performance actually means. Every pharma company's business logic is slightly different. The foundation is the same, but you you have some nuances to your own logic. And the model has to respect yours and not a generic one. Then you have the data execution. That's a governed query against your warehouse with row level security, field sales if I'm the AGM for the east area. It has to be enforced before the l one even sees it, not after. And then you have the the last layer, which is infrastructure and security, SOC two, HIPAA compliant, tenant isolation. None of it is exciting, but these are all nonnegotiables. So then if you look and and and really just taking a step back, every one of these has a a hard problem attached. You pick any of them, take the semantic depth. That's not six weeks of work. That's that's knowing across your portfolio how TRx is defined, how it differs by therapeutic area, what happens with the payer plan roll up, which makes you require in question disambiguation, what northeast means. And so it's not engineering. That's just commercial pharma knowledge that's codified. Any and and so on. So so I'm not here to tell you not to build. Everyone is building. Plenty of great engineering teams and pharma can absolutely build this. But the question for you all is and and and put this in front of your CTO, CIO, is your your enterprise architecture team. It's it's not can you build it? It's whether that's the best use of your your team's time over the next two years and how fast you can get value. Tellius is is is eight to ten years plus of engineering that's already done. So while your competitors are spending, you know, two years fighting figuring out agentic, orchestration, semantic depth, you could just be running your Monday morning briefs, your weekly territory plans today. So it's just a matter of the the time to value for you all. So that's the trade off. But now we'll we'll we'll go to the next slide. Yeah. Thanks, Nick. And just that was the the setup, and I wanna make sure we have as much time as as possible for you to kinda actually show us and flesh out what you just kinda laid out there. I think there's a lot of meat on that. But, yeah, over the next couple minutes here, Nick's actually gonna now, run you through two examples of, agentic approaches to commercial pharma. I believe, BrandPulse and a territory intervention, use case. And then after that, he's gonna walk through kind of the components, the blueprints that you need to to to to solve and create these types of solutions. I'll talk through some real deployments and, and customer examples of of how this is working in the real world. We'll talk about what it takes to how to evaluate a solution like this. And at the end, there's a kind of a special offer, so I'll I'll leave Nick to to reveal that secret at the end. As we go, please feel free to drop your questions in the chat, and we'll do our best to to answer them live or pick them up at the at the end. And then we we will have some time at the end to do q and a, as, as our, you know, I believe leave leave those in the chat. There's a a functional functionality in Zoom to to questions and get them answered. So, yeah, Nick, over to you. Kick us off with that the first mission. Alright. Now enough architecture in in giving you the state of the land, we'll actually show you one, in real time now. We'll show you how it works. So imagine it's Monday morning. You are a brand lead. You've got leadership review in an hour. You just saw that your your your NRx, your MBRx, it dropped in in across three districts last week. And the quest the first question that you're going to get in that meeting is why. Is it payer? Is it competitive? Did our engagement slip? You don't know. And your analyst team is is good, but they're not going to have an answer for you in forty five minutes. So that's the scenario that we're going to run through. And then a few things that I want you to watch out for is, first, the the you'll see that the agent scans across all of your disparate data sources, so your subnational data, your MBRX data, your your payer, and and and activity all at once, not not one at a time. Second, when drivers come back, you'll see that they're ranked by contribution, a statistical number, not a plausible sounding paragraph. The third is the and this is the one that matters most. I'm going to ask you can ask the question same the same question three times in a row. You will receive the same answer every time. Now because in commercial, if you can't if you think about, hey. If I can't reproduce a number, then you can't trust the platform. So that's the the first demo that we're going to show. The second slide, this is the the second, mission that we're going to highlight for you all. Different persona, but the same the pattern is the same. This time, let's say, I'm in an AGM, an RBD. You you have eighty plus districts. It's Monday morning. You have two hours before you start making calls to your regional directors, to your reps. You don't need a dashboard. You need a rank to list Who's underperforming? Which district should I prioritize? Who's underperforming? Why are they underperforming? And and so on. So there are a few things to watch out for here. All districts will be ranked on a composite underperformance score. You'll see multiple drivers and recommended actions to take as well. So the whole thing lands in the AGM's inbox every Monday morning at at eight AM without even having to log in. So that's what we're going to show you, a couple of use cases live. So, Walker, if you can pass over the baton so I can share my screen. Absolutely. Alright. So now the first part that I want to highlight. So one, this environment, it can live anywhere. It can live behind your VPC, use your LLM keys. We can have a Tellius hosted offering, but it's a web web based browser. Now the first thing I'll show you is that what we were talking about with emissions and having blueprints. This is the mission library. So every tile comes out of the box. It's a production agent that has been built, tested, and scheduled against real commercial pharma data. So all you do is point your data at these agents, and then they'll update. You're looking at, a weekly data quality monitor. You're looking at territory risk scorecards, KOL influence scorecards, HTTP targeting, market mix modeling. So these are all, again, agents that come out of the box that you can just define. You can bring your your PowerPoint template. It'll update that PowerPoint template and so on. So now the first agent that I'm going to walk through is the the brand scorecard. So now if you if you look here, what the user would do, you can either run this ad hoc or you can have this scheduled, meaning that every Monday morning, I want this weekly brand briefing. Here's an example where I'm running this ad hoc. So a user comes in, types in, hey. Run the weekly brand briefing. You can see it gives you an executive summary. So now one click, the agent basically does all the work for you. So now what it's actually doing is if I scroll down to the bottom for a second just to show you how in-depth this is, you can see that it actually, pulls it it goes through the the the planning phase. So it looks at your data, pulls your your subnational TRx data, your payer information, your engagement data. It does all the computation. It runs the the code for you. Each of these steps, you can actually see the underlying code that was written. It pulls whatever Python libraries it needs to. It does whatever Cisco analysis. So this is that multi agent orchestration in action. It's going to plan the steps first, then it's going to run SQL agents, Python agents. All of this is being spun up on the fly. And then it's going to each step, and this is the key part, is that you have a validation agent and a reflection agent that makes sure that the numbers match. So as an example, you can see this went through three iterations. You can see step fifteen failed. It identified when it was reflecting and validating that the code wasn't right. It was producing the wrong number, so then it went through a third iteration and fixed it. So this is that self healing component that I mentioned. So enough of the technical details. Here's what you would have in your inbox. So you'll have this Intrubica weekly brand performance briefing where you can see at the top, it gives you the high level numbers. So here at the top, it gives you Intrubica recorded a hundred and fifty k TRXs in the last thirteen weeks. You can see that the the primary competitor is mycobutin here, and it's they're significantly outperforming you. So you can see mycobutin is growing three point seven x faster than Intrubica. Then you can see some more information around what's actually driving it. So this is obviously a major concern. They they they're growing. They're outpacing you. Now you want more information. Then it runs a district performance overview. It tells you which districts are are underperforming, which are which districts are declining, which are accelerating. And so the key here is that you have two high volume districts, Miami and Atlanta. They're getting hit the the hardest. The mycobutant's outpacing them by eleven point six and eleven point five points respectively. And if you scroll down, if you look at this, you can see then psoriasis. So you can look at the indication level performance, and so it's actually highlighting that your biggest indication, psoriasis, Mycobian holds a fifteen a fifty six percent market share there, and you've lost three point seven points in a single quarter. And then if you scroll down, it also gives you, again, recommended recommended actions to take. So these are your recommended focus areas where it's not generic advice. It's specific to this brand's data, competitive defense in Miami and Atlanta, a PSO, share recovery initiative, accelerate in in below median districts, and investigate your prescriber, based segmentation. So I want you to see something as that is is this is all automated. Now at the bottom, it gives you more information and follow-up questions to ask as well. So now that is great to see in the platform, but you know what the teams do next. So one, listed the listed the hard part for you. So it went through all of your data sources, and it identified the the drivers. It did that whole sixteen step analysis for you. That's that is a a hard part of actually doing the analysis, so it did that for you. So it covered the insights. It put together the story, told you what's driving performance, what's going on in the latest weekly brand performance briefing. But now the part that usually takes the longest is now telling a story and putting together slides. So what you can also do is have this run automated for you, and you can actually ex have a PowerPoint deck exported for you. So here, you can see if I export this into PowerPoint, if you give us your template, you have a weekly brand scorecard. You have a an MBR, a monthly business review. You have this that is you follow the same format, same structure. If you give us that template, it'll populate the slides for you. And it's not just images. It's actual slides. So here you can see I didn't actually give it a template, so it can create slides for you automatically. But if I again, if I had given it my template, it would have populated your slide deck. So when you export this, here's what you see. So you can see that it just created this, this weekly brand briefing deck. Same insights, except it now tells a story. So you can see how clean the slides are. It tells you it gives you the high level overview of performance. It gives you the the key stories that mycobutin is dominating and they're growing. They're outpacing you significantly. It gives you the prescriber dynamics are healthy for now. It gives you the PSO insight that I mentioned. And, again, each of these is editable. It's a PowerPoint slide. So you can see this, and then it also gives you the the district breakdown, Miami and Atlanta, where the gap between Mycobuna and Truvica is is quite massive, and they're growing. And then it gives you four actions to take. So you can see that this what I just showed you, it does the automation. It does the analysis of uncovering the drivers, but it also puts together the slides for you that you can share. So this is being launched across our customer base. They're automating weekly brand scorecards, territory reviews, territory action plans, their MBR decks, and it's saving their analysts a ton of time because then they can now focus on those high value initiatives. But now that's this is the the first agent or the first mission that I wanted to highlight. Now the next part is alright. Let's say I want to axe ask some more questions here. So let's say I want to dive deeper. So now I I have some other questions I want to ask based on this. So now I can just come in, and I'll just start typing in questions, and I'll do this live. So here, we want to first start off with understanding how is Intrubica, let's say, performing on TRx this year across each district. But now if you wanted to then just answer your own questions, you can see I can just come into this conversational interface, and I can just start typing in questions so we can see, and this is the concern that the that the the brand briefing just highlighted is that Miami is your top producing district, your top performing district, but it's it's it's on a decline, and Mycobutin is growing there. So then I want to say compare that to mycobutin on the same chart. You can see you have it retains a context, and it will it will retain that context and then present that visual to you. So this is that if you look back at the the earlier slide, the asking component, this is that first part where you want to be able to to answer questions and and and and have a conversational question or conversation with the data. It retains the context, and then it gives you the output. So now I can see, okay. This is how we're doing. It gives you a narrative at the bottom as well, and then you can see the underlying SQL, the thought process as well, and the SQL that was written to generate this. But now let's say that I want to then go into the second pillar, which is understanding why. And so now this is where you can ask any type of question here. So if I want to know why is mycobutin dominating, let's say, Intrubica in terms of, TRx and NRx performance over the last thirteen weeks. Tell me the key drivers and what is impacting, performance. You give it more context just like if you're using Collateral ChatGPT today. Give it the situation, what's happening, tell it what you expect it to do, and let the AI run. So here, I'll say, give me an executive summary that I can share with my leadership team to drive strategic actions. Alright. So now when you type in that question, you can see that this is where the multi agent orchestration happens because it first identified that you're asking a more complex question, so we're going to require doing a deep analysis. So what's happening behind the scenes is that it's going to create the plan for you. So all the analytical steps that you need to go through to answer this question, then it would fire off multiple SQL Python agents, however many is needed to answer this question. It validates. It reflects to make sure it's answering the question. This is a deep insight, so it does usually take a a couple minutes to run. So while this runs, I'm going to and I'll come back to this so you can see the the planning and and what it's actually doing. But I'll jump out to the output to show you exactly what's happening here. So now this is what you would see. So, again, I gave it context. Say, mycombutyn is growing much faster than Entravica last few months. I want to know where is the biggest gap by specialty indication. Is Mycobutin winning new prescribers or getting more volume from existing ones? Which area should we we be most concerned about? So I asked a very complex question. This still falls in the bucket of a of a deep of an ad hoc analysis. It's just you're you're doing a deep insight now. So the conversational search that I just showed you is answering what questions, but now this deep insight will answer the why. And it does it by running an autonomous, a multistep investigation. It's not one query. It's a whole plan. So it's so it's planning an an investigation. It decomps microbegins growth by indication, by specialty, by district. It ranks the gap by impact score and identifies the largest threats. So now, here's what it came back with. So, basically, the headline and this is the the thing that really changes the conversation, is that mycobutin is winning through aggressive new prescriber acquisition. Ninety nine percent of their growth is coming from first time writers. It's twenty k new prescribers in the last thirteen weeks versus around two k existing prescribers. Now that's not a marketing strategy. That's a targeted launch motion probably with aggressive access, also most likely with using samples. So it's ranked by impact score as well. So if you scroll down, you can see it gives you the the key metrics. It gives you the immediate action required to launch the defense initiatives in the top three, impact segments, psoriasis, dermatology, and Denver, and then reallocate field resources to counter new prescriber onboarding and initiate a competitive intelligence audit on MicroBusen's market access and promotional tactics. Then it gives you some recommended actions to take here. And then here are the key findings that it gives you. So identifies again psoriasis and AS show the highest vulnerability. It gives you the specialty level dynamics. So this is what you would go to your your your implementation partner, your solution providers, and ask them to do a deep dive. But here, you can see you can ask any type of question. I can do a what if analysis, clustering, segmentation. If I wanted to, measure, hey. How is how are calls impacting performance? How many calls do I need to make in each territory? It can do that level of analysis for you. At the bottom, just to show you what it actually had done here, is when you're looking at this, it went through eight steps to do this analysis. So these are all the agents that it that it called in real time to answer that question. So now that is the the deep insight. It's a onetime investigation. You're doing the analysis. Now we're going to shift focus to the field sales. So I'll show you an example of an agent that we have that is for territory. It's a territory agent scorecard. So it's a different product, different persona, but the the difference here is that this is, let's say, a regional business director, an RBD. Let's say your RBD has seven territories in the south. They have one o AM per territory, and it's Monday morning. So she needs to quickly run this to know which territories need intervention this week and why. So here is the the territory scorecard is that you can configure this again in an ad hoc way where it can give you you can ask the user for prompts. So if I'm running this, hey. I want to it can ask you questions like, hey. What region? What evaluation period? So I responded with south in the last thirteen weeks. You can also configure this based on, again, that road level security. So if I'm an RBD for the south, it'll automatically know that I'm for the south. It'll run this for me. But now here's what's happening underneath. The the agent basically is composing every territory, and it's looking at it against its its prior thirteen week baseline, scoring performance decline. It's cross referencing call activity, and then it's classifying the root cause. Is it coverage gap? Call quality is an external factor. And so that's what you're seeing here. So the the the the high level summary is basically that the south is experiencing a critical performance decline. So the Dara is down sixteen percent quarter over quarter. Field activity is down twenty six percent. So you can see the the primary hypothesis forming here. So if calls are down, one point six x faster than volume, coverage is is the lever here. So you can see here, it gives you the key performance indicators. It tells you that the the territory health status, there are four of seven territories that are flagged at risk, which is obviously concerning, and that's what you would want to to dive into deeper. So it tells you what's happening. It's also telling the why. So the root cause, three three at risk territories, Houston, Nashville, and Miami, are experiencing coverage gaps driven by rep activity declines of twenty five percent, fifty two percent, and or two fifty two percent. Charlotte, North Carolina, thirty eight percent decline with stable call activity. So that signals an external factor. So you can see two of the three at risk territories, two are driven by call activity because you had a rep activity decline there. But then that third at risk territory, which is Charlotte, it's not actually driven by call activity. It's it's they're down significantly, but call there's a stable call activity there. And so then it gives you recommended actions to take. Execute urgent defensive calls on fifteen priority one uncovered accounts across Houston. And then you can see it gives you the key findings, the regional performance, the the at risk territory deep dive. So it goes into each of those those three territories. So Charlotte, North Carolina, and it's it's an external tells you what's happening here, so it gives you the headline for that one. It's an external factor crisis. Unit performance is is down significantly, worse than the region, but call activity is stable. You can see three forty b share is down. And so then you want to now probably dive deeper into Charlotte to see what's happening here. And then you can see that this is the only territory where volume collapsed despite improved call activity. So that signals that it's a it's a deeper investigation. So that's where the deep insight is so important, the insight that I just highlighted for you, because now I can run an insight specifically on on on Charlotte, North Carolina. But I can say, for Charlotte NC territory, Tell me what is driving performance here. Is it is it, are our competitors growing? Did we lose access? Is it payer driven? Let's investigate let's investigate the key drivers here. So then you can ask questions on top of this, and it'll run a deep insight for you. So going back up to the top, again, it gave you the three at risk, territory deep dives, Houston, Texas, Nashville. This is the Houston is a severe coverage gap, and Nashville is also a coverage gap with three forty b headwind. And then Miami is a coverage gap, and it gives you again the the key drivers. You can see that for, the Houston, there are these are the top five uncovered accounts. That gives you the top five uncovered accounts as well. So you can see how actionable and how insightful this is, but this takes a couple minutes to run, as opposed to taking weeks, several hours to do this analysis. And, again, at the bottom, if I am good with this analysis, you can see it gives you the full territory performance scorecard, ranked by unit performance that highlights the the top four for you, and then priority uncovered uncovered accounts, a list by territory as well. So let's go ahead and just run this live. Let's let's have it export a PowerPoint for me. So here, I'll download this into PowerPoint because now I want to share this with my leadership team. They'll create the slides. Let's export this, and then it'll create, again, slides with this insight uncovered, and it'll tell a story in the slides as well. So that's what I wanted to highlight. Those are two production grade AgenTeq workflow systems that we have that are ready to launch, plug in your data, and and you can start automating your weekly scorecards, automating your your MBR decks. It can do all of that for you. So now the the last part that I want to highlight, before we continue is that this is where Tellius stops looking like an analytics tool and starts looking like a production app. So let's say that you wanted to, to build a an actual application, almost like by coding, almost like, building a React app that you can share with your with your stakeholders. So this is where you also have this apps section, and I'll highlight the output here. You can actually build, this type of interactive output, where this is, example, a a command center, that's centered around marketing mixed model intelligence. And so you can see it creates five tabs for you. It gives you sales decomp. It tells you the channel contribution. You can deploy this in any environment. You can publish this. They can access it through the web. They can access it anywhere that they would need to. But what you see here is that you're building this through a conversational interface. So you can tweak it just like if you're using cursor or cloud code. You can say, hey. Change the breakdown. Include this a new tab here. So you can see I just gave it a prompt to build a a market mix modeling command center, interactive analytics app for, let's say, commercial analytics team. You can see I gave it three tabs to create, or four or five tabs to create. Then you can see it created a plan for me, and then I wanted to it to add a a new tab for the summary. So then it it it basically adds a new tab based on what I requested. So this is how you can build production ready applications that you can, publish instead of having to build an entirely new Power dashboard. You can just use a conversational interface to build an actual, interactive app, based on whatever you want to. So that's a a demo of the platform where we covered how you can automate your recurring deliverables using our our preconfigured Agencik workflows. You can see that you can ask questions in the conversational interface When you have to do an investigation to understand why something is happening, you can also do that in the in the platform as well. But that's the the the demonstration in terms of of the capabilities and what you can do in the platform. So now what we'll do is we'll we'll show you some behind the scenes look in terms of what the anatomy of of having this type of platform what what it looks like. So I'll pass it over to you, Walker, for to go over this part. Yeah. Actually, I I I actually, I could just take that. Yeah. But, basically, you what I just showed I showed two demos, two different personas, brand and field. And what you just experienced isn't a chatbot with access to to your data. It's a blueprint. Every blueprint in our library ships with the same five components. So now if you're looking to build an agentic system, the the one that matters the the most is the first one. It's a pharma semantic layer that's prewired. So understand it understands TRxs, NBRxs, how to calculate it. It understands the payer and territory hierarchies, the subnational roll ups, the deduplication rules. That's the work we talked about on that, build your own slide. It's codifying how your business actually works, understanding how different therapeutic areas think, how oncology brands look at their data. If I'm in immunology, it can do all that for you, and it understands the business. So you need to first have a a pharma semantic layer. The second is the insight engine that produces driver decomposition that needs to be deterministic. You cannot have a you cannot have an l one both for creativity, so understanding the intent of the queries, and execution. So you need that deterministic layer where if I type in the same question, I get the same answer each time. That's the reproducibility moment that you watched in the demo. And then the third is the orchestration. So this is where you have multiple agents connected. It's you have the planner, then you have the agents, then you have the reflection loop. That's what led to the territory agent rank that looked at the drivers, recommended actions, and uncovered why things were happening. Now and the the two that most people don't think about is the governance, the role of security where you evaluate everything before the query runs and knows the personas at your organization, and also looks at every agent action logged. And then you have the deployment kit. So UAT scripts, validation playbooks, persona training, the first ninety day KPI template. But because this isn't a platform you install, it's a workflow that you adopt. So five components, every blueprint, every time, that's a difference between a a stalled pilot and a and a weekly workflow that's delivering value within a week. Now just to show you, I showed you some of these agents, this this blueprint library. So you saw two of these missions that that we ran, the the brand pulse and the territory investigation or the territory intervention. Two more in the commercial brand and access space, formulary impact where you want to detect a payer tier change in the feed. You want to simulate TRx impact, generate a pull through playbook in forty eight hours. Msl engagement for the medical affairs team where you want to link MSL interaction data to a scientific signal inside the medical affairs firewall. Then to that cut across functions, the lapsed writer and rising star, HEP level, flags who it flags who is who is slipping. It predicts who is emerging. It feeds prioritize call list to the field, and then you have the weekly business review. And this is one for commercial leadership, variance analysis, driver decomp, executive narrative, Monday morning, ready to go. So six blueprints, all six ship with the same five components that I just walked you all through, and all six are are are production ready. So going back to your engineering team, your architecture team, hey. Yes. We can build in six to twelve months. But while you build, hey. Let's use this platform to start delivering value today and and so on. So, all this is available, and and and and so that's what we're excited to to present to you all. But I'll pass it back to Walker. Yeah. And we'll get into what customers are actually seeing. And, actually, I wanted to maybe if you go back to that slide, Nick, or one more to that to to show those layers. I wanted to maybe address some of the questions because I know we have, we're kinda getting closer to time. There's been a bunch of engagement in the question answer, space. So I wanted to highlight these and and bring them to you know, see if we can answer as many as as we can. So I there's, like, eight or nine questions here. So the buckets I'm seeing are basically questions around, you know, building versus buying and the kind of the ownership aspect of having an initiative like this, Nick. Another one was kind of questions around, you know, why pilots fail. Another one is around the quality and the limitations of agents. And then another one was a kind of a dive into, you know, statistical modeling and of and how does that work. And then another one on the deployment timing. So I don't think we'll have time to answer all of these, but I wanted to just put forth a couple of these, Nick, to you and also see if I can tackle them as well. So the first one was around, you know, when when considering building versus buying, you know, which of the five layers, of doing it in house is most underestimated? So that was, you know, I think an earlier slide we touched on. But I would say, I mean, just looking at, your slide here, it's it's definitely the semantic layer. Right? Like, that's the modeling work that you have to get done, and it needs to be across your therapeutic areas. It can take, some time to build or, you know, in a in a platform like Tellius, we come with prebuilt. What we accept, you know, semantic models from other places if you've already got that factored in, and we use AI if you don't to help build that up more rapidly. But, Nick, yeah, what would you say you know, if you're going to build route, how which one would you say is the most underestimated time sink? Yes. I I would agree. It's retrieval and knowledge. We were actually going through an example here with, one of our manufacturers. And when you and they try to build this on on their own using cloud and and and stuff like Cortex code and intelligence. And the part that it continuously got wrong was how you calculate the actual metrics. So if I'm a field sales team, you obviously have to apply certain filters when you're calculating the numerator and denominator for different types of metrics. And and so having a platform that comes with that knowledge already that you when you're when you're calculating the numerator, denominator, you need to apply, look at the targeting universe, not the entire HEP universe. And so understanding how market share is calculated, how the brand teams look at the data. So that is a an important component. And you if if you decide to build this, that'll probably take the most time because if you open Cloud and and ask it a question to do an analysis and then you come back tomorrow, you will have to give it the same context around your business rules, how to dedupe the data. And so having a platform that can that can dedoop the data, understands how to calculate metrics for you already, it'll if you don't give it information, it needs to proactively ask you questions, and that'll that's certainly an underestimated time commitment. Yeah. That's great. I think we have time for maybe one more. I think I like this one a lot about kind of controlling the quality of the outputs. You know, can an agent, how can agents control quality of the user's questions? Questions? Can agents argue back if the frame or pre assumptions that the user is asking of the model are not valid or right approach? So I think this is actually kind of not not the intelligence of of of a platform to in understand the user's intent, but, like, maybe the agent's ability to to help the user understand what they're even working with. My my just off the top of head answer on that one is just, you know, platform like Tellius, you can even ask you can query questions about your data. You can ask, like, metadata questions and say, hey. What what data am I even working with? You know? How how is how is x y z, you know, KPI coordinate, calculated? So if you're more on the business side and you're not as close to the data, you could use a conversational interface to ask a question about the data to understand what's even possible. You could even say, hey. Like, I care about this this part of the business. What what's even possible to answer with what I've got now? And and a and a capability that we actually just recently released, called Kaya Architect allows you to if you if you find that there's a limitation to the types of answers that you're getting or the data data that you have access to, you could use something like architect and and quickly pull in additional data sources and build up that data model, quickly, through you you know, using our AI platform. Nick, any additions to that one? No. We can't continue. Cool. Yeah. I wish I wish we had time to, I love these questions that came through. Wish we had time to to answer them all. Maybe, yeah, just real quick, on this slide here. We've been using you know, we've been, working with AI approaches to analytics and and and pharma, commercial pharma approaches for a while now, and, a top twenty pharma company has identified a massive reduction in that, you know, Monday morning turnaround for their creating their briefs. Right? That used to be a four day cycle. Now they can do this with twenty minutes with the human in the in the review, like, loop. Another kind of quick use case here, Nick, if you flip to the next one, this is, market share lists. So, you know, they were able to use our, approach, our platform and see, you know, a a pretty sizable market share lift, because of just the speed of being able to act quicker by spotting that insight and making those important, decisions, you know, whether it's, deploying the right, you know, resources to that opportunity, whether it's more of a market access thing. So that's another one. And then the third one is just savings from using having to use external analytics third third party resources. Right? We've got a major global pharmaceutical company that had, you know, that was using an external analytics workbench team, and they were able to reduce that spend, in in a big way, by enabling their users to be able to just ask questions, and get deep, deep answers. So that's that's kinda how, you know, just broad swath how our customers are using, the Tellius platform and and, you know, some of the ways that you might find, savings. I mean, it it pays for itself. With the last five minutes, I will I guess we'll, I'm not sure if this if this deck will be shared, but, this is just, you know, kind of a you could screenshot this or we can share it. But these are some questions you should be looking at and areas you should be looking at when considering getting, going the route of AgenTic, and AI agents for commercial pharma. You know, is it deterministic? Are you gonna get the wrong answers if you ask it different times? Is it grounded in your semantics? Can it do the the deeper why questions, the stuff that that analysts spent days and weeks working on? Can it pull off multistep analysis and get, and and answer not just the ad hoc questions, but the deep questions here. You're looking to answer the full spectrum of questions. Otherwise, you're hopping around with different tools. And then the fifth one, governance is what the whole thing must be built on. Right? You governance should never be an afterthought. So, yeah, last, thing here is, if any of this was of interest to you, I believe Nick has, is willing to, you know, you could reach out to to the Tellius platform. This is just a quick view of what it would take to to get going. I think there are lots of questions around how long it would take to do things and how long it would take to get things into production. This is just an overview of how we if you work with us, you know, there's a discovery, a build to go live on a scaling process, then we try to compress that as much as possible, because, you know, time to value. And then the last point is, we are gonna be Yeah. I can actually just quickly take this one. But Yeah. So he here's the offer for you all. I know there are a lot of great questions, but here's the offers that we will do a a working session, an hour or ninety minutes. It's on us. No pitch deck, no discovery questionnaire. It's basically what you would walk out with is our four things, and an honest read on where your AI program is today, a map of which of the six blueprints fit your highest paying workflows, a ninety day sequence road map, what to land first, what to sequence next, and an ROI framework so that you can take to your leadership team to justify the investment. The reason we we do these is is that, basically, if you look at most commercial pharma teams that we end up working with and that we start with, they always start with a session like this. And so we we find the right fit fast. We or we find out what isn't one. Either way, you walk out with the plan. If you saw what if you if what you saw today landed, the agent's running the five components to four week path to production. Book the session. It's the easiest thing that you'll you'll be able to do all quarter. But, I'll pass it back over to you, Walker. Yeah. Yeah. So, feel free to reach out to us. Take advantage of that opportunity. And I wanted to hand it back over to, you know, to the to to Shuhari and the PMSA team here, and wanted to thank them for, making us, you know, giving us this platform today. Awesome. Alright, Chris. Very insightful presentation today, both Chris and Nick. And by the way, and also to the participants is very active, engagement. Really appreciate it. And, I hope you can get back to the questions, in the chat box, or the q and a box literally. And and maybe we need an agent, Chris, to answer those questions automatically.