PegaWorld 2025: Top 10 AI & Automation Use-cases to Boost Business Operations with Pega

PegaWorld 2025: Top 10 AI & Automation Use Cases to Boost Business Operations with Pega

Hello everyone and welcome to our session on the top ten AI and Automation Use Cases in Operations. I'm Stephanie Hawkins, I'm on the Pega product marketing team, and I'm joined here today by Mike Gualtieri from Forrester. Thanks so much for being here with us, Mike. Hi everyone. So, you know, I know there's been a lot of talk about AI at this conference thus far. And our goal with this session. Talk about it more. Exactly. Keep talking about it. But also to really zoom in on some specific ways that you can use it now.

Um, so really look at kind of the practical angle of what are those most impactful AI and automation use cases that you can be using in operations, um, at this very moment. And Mike talks about this every day with clients. He sees it in the wild. So he's really the perfect person to lead us through this. Um, so with that said. Before we get into our top ten and I promise there will be ten, there may even be a bonus one at the end. I just kind of want to say, Mike, you know, obviously this landscape has changed a lot even in the last two years.

How have you seen it change? Um, what's different now? Yeah. So, um, what I do at Forrester as a principal analyst is I cover AI platforms and AI decisioning platforms. So these are software platforms that enterprises use to develop solutions. One of them is a little bit more generic. One is more focused on Decisioning Lead both those platforms. But as part of part of that work, I talked to, um, all of our clients and all of the vendors to understand what is actually happening in this market. So.

I would say agentic is the biggest change. It sounds obvious because we're all talking about Agentic, but it's a big change in a way that I think a lot of AI platforms are not going to be able to accommodate. And let me explain that a little bit. And maybe you could show the next slide, because I'll, um, a lot of the AI platforms focused on tools for data scientists to create models, right. Enterprises were creating models. You might want to call this statistical AI or predictive AI. Still super valuable.

Still very important, right? So then when GenAI came out, well, someone else created the models, but you still had to do some prompt engineering, you still have to test it. You still have to take it through a life cycle. But in both of those use cases, it was kind of a request and response, right? You had maybe you had some customer data came in and then what came out was an offer, right? It was a step that took place. Um, and same with GenAI. You entered a prompt, you got a response back. Well, now with the Gentic, you've got a you've got you've got that.

But you also have workflows, right? So it gets much more complicated now that you have a workflow. And the reason why this is such a big change in the landscape for AI platforms that I cover, is a lot of those vendors have nothing when it comes to workflow, right? They don't have a basis for workflow. So a lot of the vendors that came from it, from a data science standpoint, are really going to struggle because they lack that workflow capability that that the Agentic is going to use and work through.

So so that's that's the biggest change. Of course, the Agentic workflows. Use use LLM models, and it uses statistical AI models and other technologies for that matter. So it's essentially this movement from traditional to Llms and now agentic, which kind of enfolds all of those. And, you know, in the real world. How has adoption been? Is it happening? Is it slow? Is there progress being made? Well, even before Gen I, you know, this became AI was a strategic for most enterprises creating machine learning models and doing that. So and then GenAI it became even more strategic. And now with agentic workflows, um, I think a lot of companies are excited, but also more scared as well, because implementing, you know, the workflows are kind of like, that's what makes the company run. It's not just a step in a process, but it is the process. So that's why there's equally a lot of opportunity, but a lot of, uh. Concern about what that means. You know, it's interesting you bring that up. That's. Something I'm sure all of you have heard us talk quite a bit about at the conference thus far.

But really, at Pega, we're extremely focused on this tension between how you can get the productivity and creativity of agents and balance that with the control and predictability that you're used to from your workflows. So it's it's interesting to hear that that's something. Yeah. And there's nothing in the definition of agentic that says, oh, I'm just going to run an an enterprise and it's just going to figure out what to do. Right. That's those are the stories you hear about Agentic. Like for consumers, it's like, oh, I need a burning permit in my town.

I'll just ask ChatGPT to, you know, apply for a burning permit. And that's going to figure out how to do that, right? Well, in an enterprise, we have these workflows and processes for a for reason. Sometimes for regulatory reasons. Sometimes for governance reasons, sometimes for efficiency reasons. So there's Agentic doesn't have to mean that it's just it's just figuring everything out. No, you might have an existing process that works and is repeatable but has exceptions. But then if you can put an agent on there running the first process, that's your first step in gaining serious productivity gains at some point.

And we're going to talk about all these things. You'll want to transform. You want to optimize that process. But agentic doesn't mean you have to optimize or transform first. You can take an existing process and make that more efficient. So, you know, when you're talking to clients, sort of big picture, what types of use cases are they thinking about? Um, with both agentic and kind of traditional? Yeah, I think it's very important to make a distinction between what we're calling the assistive AI use cases and the AI for process automation use cases.

It's enormously confusing, and I have to talk about it every single time. And because the assistive AI use cases are assisting a human being. So that's that's an agent collaborating with a human. So be a human interacting with it could be a customer self-service, right. But there's or it could be an employee that's asking questions and getting answers from from the agent. So it's an agent in the same way that a human would be an agent helping people, helping people or agents helping people in this case.

Then there's other the other agents which are event driven. I call them event driven, you know. So a claim, an insurance claim comes in, right. And that an agent picks that claim up and then that that AI agent is going to work that claim. Now, as a part of working that claim, that agent May collaborate with other agents. It may even reach out to the, um, to the adjuster and said, you know what? Those photos are blurry and you forgot to take the front bumper. Could you go do that again? So. So when we talk about Agentic, it's both of these big types.

Okay. That's an interesting split because I think most people really do think about it in terms of the first one. Yeah, the chat type thing. Yep. So specifically with just agentic, what are the most common use cases you're seeing enterprises adopting? So the common the common use cases are the use cases that companies have already been doing. So um, self-service. Right. That's because and I think um, uh. Self-service chat bot for customer service. And, and one of the keynotes this morning, it said, oh, you know, a few years ago, you could definitely tell that it wasn't right.

And now with the Turing test and it's much better. So there's a lot of comfort in saying we were doing self service, but these models are pretty good, so let's do more of it. And then also in the call center where there was assistive technology for the human agent in assisting a customer, let's make that so. So those were comfortable use cases because they were already using older technology non LLM technology to do it and non agentic technology. So so those are quite popular use cases. The ones with the workflow.

Those are very nascent. Companies are still trying to get a handle on on on how to do that. Okay. That's a really interesting sort of observation that people are more comfortable taking those use cases that they've already sort of been doing and then kind of identifying them. And I was kind of saying this before with existing processes. Right. Let's use an agent on an existing process. We don't have to shuffle the deck and then use agentic. So I'll say at Pega kind of the way we're thinking about these agentic use cases, you named two of them, two of the categories.

So self-service, where, you know, you actually can fully resolve a case using an agent. And then number three, really giving employees, whether they're a call center employee or an internal operations employee, everything they need to resolve a case or resolve a problem. Um, and then the other two that we are really focused on are the elimination of manual work. So things like, you know, being able to change passwords or put in an IT ticket and not have to touch a human throughout that process. And then one that we haven't talked about that is interesting, that Don actually talked about this morning, this notion of ad hoc workflows.

So taking something that's an exception and actually turning it into a workflow on the fly at runtime. So you're actually taking your tribal knowledge of your organization, and you're codifying it and making it part of your workflow rather than an exception. That happens so many times over and over again. Yeah, yeah. And I think a lot of the like consumer use of Agentic, that's what companies are focused on because it's focused on how can I get stuff done with the thousands of websites that are out there.

The example everyone gives is booking travel. That's a that's sort of an obvious example. How can it navigate all of those websites. And that is very ad hoc, right? It's meant to be ad hoc. But those use cases also exist in an enterprise, but very different from, you know, the workflow, you know, the primary workflows that are the engines of a company. Yeah. So we have agentic our agentic use cases and another sort of bucket that we're looking at these use cases through. Um, is that bucket of, hey, once you've automated you want to optimize.

And, Mike, are there ways that you're seeing people using AI for that sort of process optimization type of use case? Yeah. So, um, when you take a, you know, any large bank that I'm sure many of you, uh, might work at, those banks usually grew through acquisitions. There's 3000 systems. There's 2000. There's just a ton of legacy in there that, uh, isn't really going anywhere. But processes were designed around people, and they were also designed around updating those systems. So to me, that means two things.

One, it means if I'm going to do a genetic AI, I need an integration layer. I need to be able to abstract all of that complexity. Second thing it means is if I'm going to use Agentic and I'm going to automate more things, I need to be able to optimize a process not around the people handling the exceptions. But now maybe I could optimize it around this agent that costs pennies. That scales essentially infinitely, right? But before you can get to that, you need to understand the existing process.

So process mining, which is, you know, a feel kind of funny for saying that because you don't say data mining anymore. Um, you know, that's kind of the in fact, the first wave I did on AI. It was like 12 years ago. It was called the Foresters Wave on data mining, but it was about machine learning. And then we changed it to machine learning and etc., etc.. Um, but finding but using traditional AI, using any techniques to optimize those processes, you can find, uh, perhaps some more efficient paths, but more importantly, find paths that are optimized for the tools that you have and maybe aren't so centered around humans handling exceptions, but agents handling exceptions.

That's a really interesting observation about process mining in general. Just the the folding in of hey, it might not be a human doing this work, it might actually be an agent that we need to optimize for. Um, but this really aligns with exactly the way we look at it with our process mining tools, just the need to use AI to be able to visualize your processes and how they're actually running, the need to use it to identify those places where things are slowing down, maybe due to handoffs or things like that, and then also the need to use AI to have conformance between processes, perhaps in two different geo locations.

Yeah. And I think some a lot of people in AI, they talk about the cost of AI and the, the llms, but it's really odd not to also talk about the benefit because, you know, you've got people saying, oh, well, you know, that that LLM they're very costly to to do the inferencing. It's like it costs $0.07. You know, it costs $0.70 or $1.50. Well, it depends on the process. But 0.00 $0.07 can actually add up if what you're doing is you're automating like e-commerce, uh, customizing product descriptions, but you're not getting the that can add up a lot.

But if you're if you're dealing with like processing a claim, um, and it costs $0.70 versus $30, then you're not going to worry about optimizing the cost. The use of the LLM in that case. Right. And I know that's not just what we're talking about here, but I just wanted to put that out there because a lot of people get obsessed with the cost of of an LLM. And for many use cases, especially these process use cases, that doesn't even matter. That's interesting. I think people, you know, sort of forget about these Process Mining use cases when they're thinking about AI use cases and thinking about just the cost per interaction, for example.

But these are really cost savings types of AI. And the other big use case for process automation is investigating a change in a process that isn't motivated by efficiency, but motivated by compliance, or motivated by entering a new market, or motivated by a hypothesis that a different form of of interaction will yield better outcomes. So it's interesting that you flow into that, because the next kind of category that we have to talk about is this what can you do at runtime with AI to make your processes more efficient.

And it's exactly kind of what you were just mentioning. It's it's these event driven ways to use predictive AI to say, are we going to miss our SLA? What decision do we need to make here? To reduce costs. Um, to be faster, to route the right resources, to the right work. Yeah. And I always remind myself and anyone I'm talking to, including you, that AI is software, after all. And everything that we've been using software for is everything. We're talking about trying to automate a process straight through processing, right?

But we could not do it totally because of these exceptions and things. But now we have cognitive capabilities. We have creativity that we can apply in these exceptions. And so we can we can further automate these processes. But another important thing to think about agents is how they can collaborate with other agents. So a lot of organizations have to reach out for a credit score or reach or reach out to some other service. Oh, we've done that before too, through APIs and service calls, etc.. Um, but with agents it's a little bit different because, um, they might be providing a stateful service. So, so meaning, um, you might ask for something and then the agent is waiting for it to return. It could be minutes, it could be days. Um, but there are standards emerging. Um, like a to a agent to agent. Google put this proposed this standard. Many, many companies have uh, are adopting it so that agents can can talk to each other. So so we should talk about process automation within an organization. But when you start to extend that to the entire ecosystem of business processes. Wow. That's pretty interesting too.

It is really interesting. I think that's going to be in our top ten for next year. I don't think we're quite there yet. Um. So I'll ask one last category question, and this is about the use of AI to transform your processes, because this is one of the biggest pain points we see. And I know we've heard a lot about this at the conference thus far. How can we use AI to help with that very painful legacy transformation process? Um, because right now, as it stands, it's expensive. It takes a very long time.

Um, you know, consultants, lots of meetings, thousands of documents. Lots of these projects really end up failing. So have you seen successes out there in the wild with people using AI to help with this? I'm not sure I've seen some successes yet with this, but the the promise of it or the potential of it is really twofold. So the first one is in. Um, the transformation of the architecture of your enterprise architecture itself. Right. Because you're always faced with this decision, do I retire systems, do I refactor, do I or do I sort of create an abstraction layer over those systems?

Well, now with um, and you know what? I love Blueprint, because Blueprint kind of uses the approach to create a process layer, right? But there's also coding agents, right, that can recode and create these, this abstraction layer for you. Um, but also start to retire systems. So there's two levels of transformation. There's the systems itself, but then there's the business process. So I don't think there's I think companies should favor an abstraction layer over, like retiring all kinds of systems.

Because the second Potential for transformation is entering new businesses and new markets because you're no longer constrained by necessarily people who have expertise. So if you're an insurance company and you're doing home and auto and maybe some commercial lines but not marine, why aren't you doing marine insurance? You don't have the expertise. The business processes is is different, right? But if you have an agentic process and you can kind of Blueprint that process for marine insurance, then maybe you don't need to hire as much expertise, and maybe you're a little bit more confident that the, you know, your AI agents can handle those exceptions.

So the business transformational potential here is to scale your business into other related businesses that you would never think of before because of the risk, the cost, and the expertise that you would need. And this is why I think we're seeing a lot of the companies like Salesforce and ServiceNow that outsource very specific. Now starting to say, actually, we can handle a lot more, like every company is saying that now, but that's because you can scale this. That's really interesting. I mean, that's a way to look at legacy transformation as sort of a really innovative revenue unlocking process where you may not have gone into a different area of business.

Yeah, because if you say, why can't we go into it? Those answers always have to do with the people and the process. Right? And so if, if, if, if agentic I can bring the expertise then you don't. The same challenges don't exist, right? Yeah. I think, you know, the way that we've been looking at it up until now, and we really have seen some, some great successes with using Blueprint for legacy transformation are sort of these two categories of um, being able to not just take the same exact process and put it on the cloud, but actually being able to take the time to re-imagine it to be the best it can possibly be to, to fit, for example, using agents or whatever the new types of, you know, parts of the workflow you need are.

And then secondly, to get things up and running so much faster than ever before. Yeah. And also hypothesis testing, um, you might have a process that works. And then you're saying, oh, well, what if we do this, this and this? Oh, well, then we have to train these people, and then we have to do all of this. And, well, if you have a genetic architecture, you can just you can just launch that process and a B test it. Right. Do a B tests. Business processes. Not not many people. That's a very costly and risky thing to do.

Um, but when you don't have to train people because you have agents and when, and when it's scalable, you can just start putting out new processes and testing those in the market. So I'm curious, as we've been talking, I have a few things have come up. I'm wondering in the companies you're seeing putting these use cases in place, how are things being organized? Who owns Agentic, for example? Well, in terms of AI, there's been a whole, um, a transformation of who owns it, right? Because when I first started covering this, uh, ten, ten years ago, ten, 12 years ago, uh, these decisions were AI decisions were very focused within the data science teams and the data science teams.

Though they knew how to code, they were smart, they understood data, but they didn't understand the entirety of the enterprise and how to deliver software systems that that worked. Um, and that's why you kept hearing, oh, they created the model. And then it took six months to get into production. Well, it wasn't. And I would I would always say, well, why can't you just deploy it this weekend? Like, what's the big deal? Well, we didn't think of this. We didn't think of that. We didn't. Do any change management.

It wasn't in the SDLC. There's other software that had to change. But then when I became more strategic, then it started moving more towards the technology, uh, groups. So enterprise architects, office CTO office CIO and then with GenAI. And so that they were now involved in the decisions. And as someone who covers AI platforms, I saw most of the AI platforms all started to add new capabilities that weren't what data scientists wanted. They were doing that too, but it was what these technology people wanted security, scalable, that type of thing.

And then with GenAI, more business people now got involved in this. So I would say now where the people are making decisions on Agentic are now that same, that same group of people in the business and the technology leadership are now making those decisions. And they are very mature and sophisticated in making those decisions because they've been doing it for regular software for for years. Do you think it'll be the sort of thing where we'll see a Co type of model emerge for AI in general? I think that's company specific.

I mean, I don't I'm not the companies I talked to. Many of them have a COE approach to this type of thing. Um, but I'm actually seeing a lot less of that now. Um, and I think a lot of that, though, is because this is also, um, it's kind of a career maker for many people as well. So people are trying to own it. They want to own it now, like I think before they were afraid to own it because it's a good way to get in trouble. Um, but so, so I see some companies have a COE culture, others don't, and I don't.

I think it can work both ways. Okay, that makes sense. Um, so you mentioned earlier that self- service was a great use case and sort of contact center agent assistance was a good use case. If you had to kind of recommend one other use case to start with. For those who want to dip their toes into an AI use case and operations, what would it be? Well, I mean. Whiteboard. What are your critical business? You know, what are your most critical business processes? It doesn't have to be most critical one.

It could be a costly one. Right. And then just walk through that business process and look if there are any steps in that business process where you could predict something. There's your predictive AI where you could generate some content or extract content. There's your GenAI or where you could make some decisions. Right. And this is the whole thing that I cover is AI decisioning. And this is what Agentic is particularly good at. Um, make decisions in that process to route that, to maybe route the workflow, uh, handle exceptions, that sort of thing.

So the use cases are very easy to find because they're there in your existing business processes. Okay. Yeah, I think it's interesting. People often start with the question of what use case should we start with. But to your point, it's more a question of what is your what are the pressing business problems that you might be able to solve with AI? Yeah. Okay. Well, I wanted to point out before we move to audience questions that we have a great resource on this very topic. Uh, a white paper on different AI use cases and operations.

So feel free to grab the QR code and and get a copy. Um, but I will open up the floor to questions if anyone has one for Mike. Or for Stephanie. Or for me. You hear me? Okay. Yes. So you guys talked a little bit. Mike, you mentioned a bunch of different technologies. Pega is obviously workflow automation company at heart, right? That's what kind of what we do. Um, you mentioned things along the lines of AI decisioning. You talked a little bit about process mining. Steph brought that up. There's other technologies that are out there like robotic process automation and things like if you're really thinking about a case management, things like that, if you're thinking about kind of an end to end process that will eventually become agentic.

I think yesterday Alan Trefler got up in the keynotes and talked about the idea that that needs to be well controlled, at least for to start with, organizations that are well regulated, regulated have to be very much in control. Probably everybody does. What do you see that looking like over time though? Like how does that the orchestration of all that different type of activity, how do you feel that's going to change? Because I feel like it's not where now is probably not what it's going to be in 18 months or three years.

But do you have a directionally. Well, first of all, absolutely. It should be controlled, right? I absolutely agree with that. There's there's and this is what is driving me crazy about how some people are talking about agentic, as if to do agentic means to give up control. It doesn't have to mean that. Um, so, but so I think in the future it's going to still mean having confidence. I think it's not just about control, right? It's about confidence. Just as leaders in an organization have confidence in their people and in their processes, they work and the customers to have confidence in it as well.

Um, so I think there'll be continued confidence and control, but it's going to be at a higher level, right? Because as they start trusting lower level components that just work right, then your attention focuses higher and higher at higher and higher levels of abstraction. Um, but I don't think there's going to be a time when you're ever just going to sit. Well, okay. Do you want to get into artificial general intelligence and superintelligence and all that stuff? Um, but I don't think in the next five years there's going to be a time where you're just going to have the, you know, you're just you're just going to have an agent that you buy.

Agents Gone wild. Yeah. That it just just like you're just going to basically say to say to what? Run my business. However, I do see Tools that in the future that will allow business people to, um, uh, talk about the metrics they want to measure and the outcomes they desire. Right. And then I see that I see advice being given by the agents of saying, well, just like just like a team would internally at a company. Well, you know, here's an idea. What if we do this? But I still think the person's going to say, yeah, that looks like a good idea.

Let's do that. Makes sense to me. The other one I was going to ask you is I think yesterday, um, Alan talked a little bit about it was a phrase he used. It was the right AI for the right moment or the right time. Some stuff. You know what I'm talking about or something along those lines. Um, do you have a sense of. So you do AI decisioning, right? You cover that, you cover. This whole market is there. Where does AI decisioning kind of fit into this? Because if you think about it, I think of that as a surgical saw.

And I think from a genetic perspective, I think of that as a lot more general. So where's that particular kind of technology fit? Well, when an agent has to make a decision. Right. And and when a transaction comes in, um, deciding whether it's fraudulent or not is a decision, right? When when a customer is shopping for something, deciding what you what their intent is and deciding what to show them. So so there's decisions in everything. Um, and the AI decision platforms are focused very specifically on making those decisions.

And what distinguishes them from AI platforms, which I also cover. So for example, a so Pega is an AI decisioning platform. Datarobot is an AI platform, right? But the AI decisioning platforms allow for multiple decision intelligence technologies. There's three flavors of AI. I. There's mathematical optimization, mixed integer programing. If anyone uses that, there's backtesting, there's math, there's human logic, there's compliance rules. So so you need multiple technologies or you may need multiple technologies to arrive at that decision.

Thank you both I don't you guys let me steal this I will do this all day. So you got to stop me. So I'm just going to I'm going to. Hi. I'm from new Jersey courts, and we have about 90 Pega applications. And one of the biggest challenges we're facing right now is we have a lot of mainframe applications, and we're trying to get rid of the legacy applications and transition into Pega based platforms. And in terms of the presentation that was I was attending that I attended yesterday. It showed that you can actually take copies of the mainframe application data, and then have Blueprint actually identify and generate a more Pega application, Pega based application.

So from that standpoint, what type of process would that allow or require? I mean, if it if the mainframe application is already integrated with Pega applications or Java based applications, does that actually have the capability of identifying the best option of creating a Pega based platform and getting rid of legacy? I think it depends on kind of the combination of applications that you have. Because we have a few different flavors of how you might do that. But there are definitely ways to upload documents, you know, screenshots, policies, um, of whatever resources you have that describe your application and then model out that application and reimagine it with Blueprint.

Thanks. Sure. And we also think that agentic AI is emergent AGI. So artificial general intelligence is where, um, it's that AI can be completely autonomous and it can do three things. So the characteristic of AGI is three things. One is self-learning. So it it can if it's pursuing a goal and it has a gap in the knowledge it can pursue and self-learn. Right. That's as opposed to models right now that get trained. Right. So ChatGPT the next version of GPT, it's trained. So models are trained so self-learning.

Right. So and there's already some some models that do a limited self-learning. The second thing is that it has to collaborate right? So if you look at Agentic, it has to collaborate with people on other agents. Well, if you look at Agentic, right, there's these collaboration frameworks that are emerging. So that agent to agent framework I mentioned from Google. Um, and then the third characteristic is it has to be able to take actions in the real world. And that's one of the characteristics of Agentic as well, to do that digitally and physically.

Um, the thing about um, AGI though, is that you give it a goal, right? You give it a goal and then it uses those characteristics to pursue that goal. Non AGI or Agentic. The AGI we're talking about right now is very narrow. The focus it's it's focused on this workflow or it's focused on this task. Even though the models are general the solution is narrowly focused. So I think we're on the cusp of of that as well. We're calling that emergent AGI. And I was really fascinated to see that up on the screen too, with the superintelligence as well this morning.

I've never been shy of getting on a microphone, so a few people aren't going to steal this from me. That's fine, that's fine. Okay. So I really had two. I know we're running out of time stuff, but two questions for you. So, um, a lot of when we talk about agentic, people talk about large language models, right? But it feels like the market's kind of becoming more domain specific. Is that fair, Mike? Like, do you see this, like, from a small language model? A much more kind of, um, like a how is that changing?

Well, I mean, yeah, you may have seen like Google recently released some medical model that they've released those in the past to medical models. Um, so there's two schools of thought on that. One is that models can be trained on very specific data, and they're going to be smarter about that. But the other philosophy is that you just keep training it with more and more data, larger data, and it gets more generally smarter. The way that most people focus those models now is through the prompting.

Right behind every, um, uh, request you make to an LLM is a system prompt. And that system prompt is what can focus, what focuses the model on the knowledge. And then there's Rag retrieval augmented generation where you're getting the data. You're getting specific data. You're giving it to the model and you're having the model use that as a source of truth. Do you see it? So the way I think about it is if you have a large language model, there's only so many people that can train those large models with that much data, right?

That's a very time and time intensive and expensive endeavor. Do you see the market wanting to shrink down a little bit? Do people want to build their own smaller like, you know, slms, if you will, or small language models? Do they want to go that way? Uh, a lot of companies in, um, that do scientific research. So some hedge funds want to um, some pharmaceutical life sciences companies during drug discovery. Because remember, when we talk about language, it's not just human language, right. There's coding models, but there's also the language of every discipline in science, including DNA and molecules.

So for for very special use cases, um, where it's not human language, there's companies that want to develop models. Yeah, I'm rooting for a baseball history one way or another. Can I ask questions about batting averages? Um, and I know we're out of time, basically. But, Mike, my last question for you is, is there somewhere that myself, these other people you write a lot like you output a lot of research, right? Is there somewhere they can go to learn more about this, to learn more about it? And do you have any research coming out like kind of what's the what's the scoop there?

What do we need to know? Well the I decisioning wave is out next Tuesday. And what I do there is I evaluate 15 vendors in the AI decisioning. Pega is in there and 14 other vendors. So um, and, and I've been doing that for a long, long time. But the criteria changes and some of the things that we did test for in terms of capabilities is agentic capabilities. Um, and that is really starting to distinguish the vendors in an Agentic is having the workflow capabilities versus just sort of a request and response type mode to a decision.

So yep, that's perfect. So I need to know I'm good. Thank you guys. Thank you. All right. Well we will all look out for the wave on Tuesday. Um but thank you so much Mike. This has been a great conversation. Stephanie. Yeah, really appreciate having you. Okay. Bye. Everyone.