My name is Peter Welburn, and I am the Vice President of Corporate Development and Investor Relations for Pegasystems. I am so excited to welcome you today to our 2026 investor session. To get started, I put our safe harbor up here on the screen. Certain statements in this presentation may be considered forward-looking statements as defined in the Private Securities Litigation Reform Act of 1995. These statements represent our views only as of the date the statement was made and are based on current expectations and assumptions. Investors are cautioned not to place undue reliance on such forward-looking statements, and there is no assurance that the results included in such statements will be achieved. For additional information about our safe harbor, please check out our disclosure materials on pega.com or the most recent information in our publicly available information such as our 10-K.
To ask a question today, if you raise your hand, we have two mic runners in the room. They will come over. If you could identify yourself, first name, last name, and the firm that you are working with. We are audio streaming today, and there is going to be a recording, so we would like that for the transcript. If you are listening on the live stream and you want to send in a question, you can send it to me at peter.welburn@pega.com. In addition to that, you can send an email to pegainvestorrelations@pega.com, and we will try to respond to those as well. In terms of the agenda for today, we are going to have Alan Trefler, our Founder and Chief Executive Officer, come up first. Alan has customer obligations for today, so he is not going to be able to stay for the entire session.
What we will do with Alan is Alan is going to come up, talk about our strategy for reimagining business with Agentic AI, and then he will take some questions. Alan will probably go about 20 minutes, and then he may be able to stay for a little bit longer after that, but he is going to have to go meet with customers after. Next on the agenda will be Don Schuerman, our Chief Technology Officer, and our head of marketing, and Carie Whalen, our Vice President of Solution Consulting. They are going to give a product and a go-to-market update. Don is going to talk about some of the most recent and exciting new product capabilities that we have, and Carie is going to do a demonstration of those latest capabilities. That will be great to see.
We will have Ken Stillwell, our Chief Financial Officer and Chief Operating Officer, talk a little bit about tokenomics, and then he will give the financial update, talking about driving durable cash flow, which is a huge focus for us, and we have had great success with that. Ken will wrap up with a Q&A session. My expectation for today is we will run about an hour and 45 minutes, plus or minus 15 minutes. Every year when we do this investor session, right after Christmas, I go to Ken and I say, "Hey, Ken, we have got the investor session coming up." He is like, "Peter, the investor session is in six months." I am like, "Yeah, I know. We got to get started. We got to get planning." I actually talk to our investors, talk to our sales side.
We try to craft an agenda that really fits your needs and answers the key questions that you have. I'm very excited to have Alan come up here and talk about our strategy. Alan, if you want to come on up.
Thank you. Do I have any slides?
You do.
I know what they are because I have a couple of favorite slides. Let me thank everyone for coming. It's always nice when, I think, investors have a chance to really understand what's going on in the company. That's especially true at this time of immense confusion in this market. I mean, just craziness. Giving you a chance to actually see how we position ourselves, see the product and where it's going, talk to actual clients and understand what their motivations are, I think that should provide, hopefully, great insight for you. I'm going to show you a couple of slides and then just throw it open to any Q&A that you might have for the remainder of the time that I've got here as well.
The critical thing about Pega, I think, was reflected in my keynote this morning, which hopefully, all of you or many of you at least had a chance to see, which is in this midst of all this confusion, Pega actually has a very cogent strategy to bring predictability of outcome and predictability of cost to a technology which, though enormously exciting, has neither. This was not something we just stumbled into. This is something that was very much planned, contemplated, tested, validated, and exercised and plays to the history we have that goes back decades.
When we saw what was going to happen with generative AI, it was hard to predict how fast this would keep happening, at the end of 2022, it was pretty clear that something remarkable had occurred that was going to give us both the opportunity and the requirement to rethink large chunks of our business. We had to answer a bunch of questions. One question is, well, where and how would this fit? The answer is, of course, that it fits in more than one way. The primary way that we thought it would fit would be to help you organize this Agentic approach to workflows.
That the workflows which we had built our company on and were such a core asset, our understanding of how workflows work, our understanding that the constructive stages, the steps and service levels, a greatly nuanced experience, having done this through many generations. We said, "We can use AI to enhance that understanding by being able to address the things people find hard about using our technology, but use the technology in a way that will give them extra advantages compared to the alternative. This idea of taking the workflow as the central element of what we wanted to create, using AI at design time, using Blueprint as its name, to both challenge and stimulate and capture our AI and our knowledge of how workflows work.
Now being able to also incorporate the IP of partners, being able to do this at design time would allow us to really exercise the workflows. Have candidly many more of them than you might have historically, so they could be even more precise, but to do it more easily with less training and education, with greater reliability, with greater quality assurance. We just thought, boy, that would just be really excellent because we've been trying for years to make our products easier to use, more accessible to greater populations, available to more of the market, and even heavy going. We've been making progress, but it was a lot of work.
We said, we can use this AI through this blueprint concept to just radically change that and move from a very conceptual way of talking about workflows to a very practical one, which is, hey, describe your business, and we're going to show you what it would look like if you wanted to run it in a predictable fashion. We said, in this environment, you will have agentics a couple different ways. One, you will have conversational agents. You will have people who want to talk to applications. Applications, the old way of applications being thought of as rigid boxes and screens, we thought that that was going to go away, and I would say that decision set turned out to be pretty good. We said, we're also going to want to, from the steps in a workflow, call agentic technology.
Call a risk control agent that one of our customers may have written for the bank, or call an agent that would do document processing and parse fields off of documents, or just call an LLM directly to summarize something and create a summary. We said, if we do this right, all of those calls should be highly deterministic. They shouldn't hallucinate, because that's not where the language models hallucinate. They do really well with micro speech translation and conversational elements or being able to move fields around. They're just fabulously at that there. Where they hallucinate is when they start to reason. When they start deciding how they're going to go about doing things. We say, well, let's use design time to challenge them to reason, to get the best thinking they have, but actually show it to a person.
From showing it to that person, be able to actually have a reliable, predictable workflow that's going to run at runtime. Then at runtime, have that work in a way that can call agents, but also just holds together in this sort of predictable system. Of course, being able to open not just to our conversational agent, but that any agent on the front end should be able to call us as a resource, knowing that if they use us as a resource, they'll have the reliability of workflows in what they do. Then to carry this along to say that we have customers with not just one or 10, even many dozens of Pega applications.
We're going to want them to be able to hook these into a fabric so that the idea of a fabric control agent, which is candidly really an awful lot like an application control agent. It basically takes a set of stimuli, looks to see if there's a workflow that knows how to deal with those sorts of stimuli. If it does, snaps to it and executes. This would let us achieve the vision of excellent predictability. It would also let us break down the cyber walls between apps, because being able to snap to different ones means you can get to the workflow wherever you want to get to and then get to another one that might be someplace else. This was the original vision.
We also thought, the good thing about this is that this is going to be very frugal with its use of all these tokens that the LLMs are giving us for free. I will confess, we were perhaps a little suspicious. We got more suspicious when they suddenly announced that they were going to build data centers for Remember they announced data centers for hundreds of millions of dollars, and then it was billions of dollars, and it was hundreds of billions of dollars. Now no one talks without using the T word here as well. Maybe we're going to send them down this because we said, they're building technology to drive consumption of tokens. We shouldn't put ourselves at the mercy of having them drive that unless there's some real added value.
By the way, not only is there no added value having them drive the consumption of tokens by doing reasoning at runtime. There's a huge negative in which you can't describe to people how things are going to work until after they happen. We went hard down the notion of burn the LLMs tokens like crazy at design time, really exercise that. For everything you design, you're going to run it hopefully thousands, 10,000, 100,000 times. When you're running it, only use the LLM for those very narrow tasks where you need to do a translation or something needs to happen. By the way, in a lot of those cases, you can use a lot cheaper LLM, because if all you're trying to do is parse some language, you don't need a 1.7 trillion parameter model. We would have all sorts of optionality.
I can't tell you how thrilled I am now that in the last six, seven weeks, this term tokenomics has come up. In the early free drugs days, which were, say, February, you never heard that. Never heard that. Now I hear it all the time. Sometimes it's nice to have an architecture that was built to take the customer someplace that should be very special for them. Candidly, that's what led us today to say that in Infinity 2026, we're not going to charge for tokens. It's not because we're using so many tokens that are going to underwrite it, which, by the way, is what the models did. It's that this style of use treats tokens as if they are something that should be treated with respect.
Which, by the way, I think they should, because the other consequence of tokens is burning forests and the incredible amount of electricity it takes to run a graphical processing unit to come to a conclusion. When I'm over in Europe and I was talking two weeks ago to some folks and describing what this architecture was, and I said, "And by the way, this burns way less trees. This is terrifically better for the environment." They said, "Yeah, we still like that." Now, here, I don't know, it depends on the times, et cetera, but given that it costs money, everybody likes that as well. I think we're in a terrific position with our architecture and with this positioning here, despite the fact I cannot overstate the complete amount of confusion that has been created by the LLM companies, basically, who we use and love, right?
Declaring war on software and trying to say that the entire TAM of the software industry is something they're just going to wipe out. I don't think they're wiping it out anytime, actually, ever. Certainly not anytime soon. There are parts that are vulnerable. There are elements of software that inevitably, when big inventions happen, have to change. The ability to manage work in a predictable way by using intelligence and then leveraging the intelligence isn't something that I see going away. In fact, I see that's something that's going to be required much more. That's what our bet is, and for those of you who are investors, I hope it makes sense to you. I will be happy to answer any questions.
Looks like Steve Enders from Citi has a question.
Great. Steve Enders from Citi. Thanks for hosting this. I want to ask just on the tokenomics discussion, as this has kind of come into fold over the past few months, where are customers at in terms of their understanding of the budget consequences, and is there any change in terms of their posture for how they're thinking about leveraging the model vendors versus looking at Pega or other software to help them through these challenges?
I think the customers are trying to figure it out. The customers are hearing amazing and wondrous things. They're trying to figure out what's right. The level of skepticism and suspicion has massively risen in the last month. This conversation that we're having, I will tell you, is resonating with the senior executives that I was talking with. I was talking with the CIO of a Fortune 50 company who has a travel ban and wasn't sending anyone to PegaWorld. On Thursday, talked to her for half an hour and said, "Look, I need to show you a slide." And basically showed it wasn't quite done yet, but that slide as it looked on Thursday. Anyway, I just met her two people who she sent. She said, "We've been talking about this a lot.
We're not sure what's going on, and you guys obviously know something that we want to learn more about." I think we have an opportunity to do some education here, but the cacophony out there is huge, and it's a very confusing moment. Having said that, when I see our customers get up, talk about Pega Blueprint, when I see what we're doing with Pega Blueprint, because Pega Blueprint, remember, only took us part of the journey. When we talk about now bringing the AI of Pega Blueprint into the whole entirety of the journey and being able to do that as soon as this summer, I think that the customers are going to be a lot more excited and receptive to that.
Thanks, Alan. Blair Abernethy with Rosenblatt. Just wanted to ask you a little bit about how you see the opportunity within your installed base now, particularly around the ability of Pega Blueprint and the other technologies you're adding to accelerate the transformation off of legacy.
I think the installed base has been hampered by the fact that when we built Pega Blueprint, we built it really as a way to get started and to conceive a new application. It was conscious decisions. You couldn't do everything, particularly when trying to do something as aggressive as that application of technology. We made a very conscious decision that we would use Pega Blueprint as a vehicle to get people to understand and reimagine what could be done. As part of that, we really didn't have a great answer for my customers' existing systems, who Pega Blueprint would not affect very much. Now, if they were building a new workflow, and many of our customers have added a new workflow, they can add a new workflow with Pega Blueprint, and it works great, and we've got clients who have done that.
In terms of adding on to an existing application, which is an awful lot of what customers want to do, we didn't have what I would say is a great answer of how we bring AI into that part of the development process. Now we do. By taking the Pega Blueprint capabilities and putting them in this Pega Infinity Studio, I think it really is going to enable existing customers to look at evolving their existing Pega systems tremendously more. I think it will make them a lot more receptive to the whole legacy transformation message because you're not making them do something new. We never intended to leave our old customers at sea. We just had to go about this in a certain way to have a chance to be successful with something that was this consequential a change. Does that make sense?
I have a question for the audience. Do people understand reasoning tokens? Do you know what a reasoning token is? Some people are nodding, some aren't. Something really wonderful happened in the last six weeks. Claude and OpenAI both added these new models. First, they claimed they were dropping their token prices. They dropped some token prices on some old stuff, but they added new models that don't have dropped token prices. These models, you'll see them if you use them, and I'm sure all of you are using these at some point or another, say things like thinking or canoodling or whatever. They're going through a thought process. What they're doing there is generally called reasoning.
If you want to talk to your LLM and ask him about this, they'll tell you exactly what I'm about to tell you, which is, you used to think you'd know how many tokens you were using. You used to think that you'd know that if you asked a question or put something in or uploaded a document, you might use 100 input tokens. Then the thing would think, and it would give you maybe 400 output tokens. Now it turns out the input tokens are 1/3 the price of the output tokens, but all right, I got 400 tokens. I can calculate the price, et cetera. I kind of know what's involved. By the way, the whole fact this industry called them tokens, I think is just an example of the BS that's going on here.
There's a mathematical relationship between tokens and words, and they could have called them words. That would have been more transparent. Tokens, that's obviously a lot more mythical here as well. Anyway, you have these input tokens, you got these output tokens. Do you think that's what tokens you're using? No. Your LLM, when it's doing that thinking, is generating tokens for itself. If it actually decides, as many of these do, that it wants to split the problem up and create what's called a sub-agent to go off and research a little bit of something, it uses tokens to go talk to the sub-agent. Then it pulls those tokens back together to try to figure out what the conclusion is.
It's not atypical for the total number of tokens that you get charged on to be 5x to 10 x the number of tokens in your input and output, 5x to 10x. You don't see them till you bill. They may have changed this in the last two weeks, but two weeks ago, I asked Claude, "Can you tell me how many tokens we just used?" "Oh, I don't have that record. You'll be able to see that on your billing." This is, I think, indicative of why when these new technologies come out, I think good organizations, wise organizations always ask, "How can this operate against us?" I'm so glad we did when we were making the design decisions about how and when and where to use LLMs. They're not done yet.
They have to raise a lot more money for them to get the IPO valuations they want. Question?
Thanks. Austin Cole, Citizens. Maybe just to continue the discussion there a little bit. On the one hand, we're seeing token usage just go up, which is indicative of Anthropic's revenue exploding. On the other hand, we talk to some of the customers here and large enterprises that have massive barriers in terms of their cyber teams and IT teams, that if AI is anywhere in the conversation, there's a million approvals that need to take place before anything gets implemented. What are you seeing that gives you faith that maybe this AI kind of token usage only in the design phase really rather than runtime? Is that a value proposition that's going to resonate in terms of getting more adoption of your solutions through those barriers in cyber?
Is it the case that people are kind of just jumping in with token usage just because AI is a mandate and it needs to happen? How do you see those two playing out?
I think that being able to point out to people the difference between design time and runtime and say that you get to curate, you get to put a human in the middle of the workflows you're doing, and then you get to execute them. Like if you liked my chef recipe analogy this morning, you get to approve the recipes, and then the sous chefs will go off and will cook them all. I think that has, and I know it has, provided a lot of comfort to some organizations who are extremely concerned about reasoning at runtime. I'm certain that that is helpful.
Having said that, there's still an enormous amount of confusion here, and it's going to take months and quarters for the confusion to abate just because, look at how obscure some of the language is that people end up using and talking about here. I think that the fact is somebody can use a Pega system to do serious automation without actually using the AI at all because we have the workflow engine there at the heart of it. I take a lot of comfort in that it's not like there's magic that's been glued in. This has been very carefully architected in to the way that workflows work. I find that customers seem to understand that.
Thank you so much, Alan, for answering a few questions. We're going to move on now to Don Schuerman, our Chief Technology Officer and head of marketing.
All right. Just a color answer to that question, just with real-time kind of feedback. I was meeting yesterday with a European energy customer and there's a lot, especially in Europe, increasing concern about even using some of the U.S.-based models, right? There are a whole bunch of European models. They've made the decision to stay off cloud and deploy entirely on their own GPUs for a whole bunch of security reasons. They love Blueprint at design time. The reason they love Blueprint at design time is their security people actually feel much safer because they're not putting their PII data in it. They're just using it as a design time tool to mark out their processes.
Actually, as our clients have increasing sort of security concerns and in some cases, especially in Europe and APJ, regionality concerns about the use of these different models, especially in the cloud, design time gives them a nice way in to be using it, where there's a lot less risk, mainly because there's no PII data flowing through it. All right. I'm Don Schuerman. I'm Pega CTO. I also head up marketing, which people always ask me about. I always say that CTO means chief translation officer, especially today. A lot of what I end up doing and my team ends up doing is translating technology into value and also making sure that we translate our clients' needs back into what Kerim and the product teams are building. The other thing I think increasingly is marketers and CMOs are increasingly becoming system architects.
We're increasingly actually building marketing stacks that connect agents and workflows and some of the data decisioning that we do to actually activate and execute our marketing. It's an interesting time to kind of bring some of these left brain, right brain things together. I'm going to be talking a little bit more about that tomorrow. I want to focus on today just sort of where we think Pega's differentiation is, where we think we can offer some unique solutions to our clients. One of the things that you'll hear throughout the conference is a real emphasis on first and foremost, this idea of reimagining work.
If I listen to what McKinsey is saying, if I listen to what folks from Bain are saying, Gartner, there is an increasing realization that the value from AI is not going to come from just dropping a model into your business for an enterprise. The value is actually going to come from the real hard work of reskilling teams, of rethinking and redesigning processes and ways of working, of in some cases reorganizing parts of the organization. There is a true reimagination that needs to happen inside the enterprise, a true, we talk a lot about digital transformation. There is an AI transformation that needs to take place.
We are really excited and have great feedback and response from our clients around the ability of Blueprint to play a role in that reimagination, specifically getting them thinking differently about how they've designed their workflows and how they start to plug agents and AI into it. We're continuing to lean into and really, I think, use Blueprint as a key tool, both early-stage conversation engagement with our clients, but also deeper and deeper into the design and delivery process to really help them through that transformational journey. The next thing we are going to be talking a lot about is this need to run predictably. Predictably both in terms of outcomes. The models increasingly, we had Gartner speaking here yesterday. Gartner was citing a stat that some of the models are getting to 85% accuracy rates.
I've heard other things from the vendors who are approaching 90%, 95. When I go talk to someone who runs operations at a bank, 95% accuracy is a 5% failure rate, right? They can't live with that. The interesting thing is if you start building out these agentic systems where I have multiple agents talking to each other, what I'm really getting is 0.95x, 0.95x, 0.95x, my overall accuracy starts dropping and dropping and dropping. That does not work for the mission-critical processes that drive regulated businesses like banking, like insurance, like the U.S. government, like you've heard from in the breakouts on the stages today.
We have an ability, because of our strong background in the workflow space, to run this stuff with predictable outcomes, to have it actually run in agents, run in self-service agents and conversational agents that talk to customers and employees, plug in agents to automate tasks that maybe couldn't have been automated before, like research, document validation, synthesizing information. Orchestrate that in a way that we are using the workflow to get guaranteed responses, to keep the things in the business that need to be deterministic. The added benefit of that is, it's not just predictable outcomes, it's predictable cost. That's what Alan has been talking about. That's what we announced this morning.
We are tying our pricing for AI built in Pega to the amount of work that you do, the number of cases, the number of customer service requests or claims you process or investigations you resolve, not the number of tokens you use on the back end. Token is a measurement of usage, not of results. We actually want to tie our pricing to the work being done, and we believe we have the architecture that allows us to do that in a way that both delivers incredible value to the clients while also protecting our margins. Some of the things that we're doing to make all that possible. What Kerim announced today is really taking that design time power of Blueprint and pulling it into Infinity Studio.
What we've done with Infinity Studio, I'm frankly amazed at the speed at which the product team was able to do this. We basically took the build environment for Pega and completely redesigned it around AI. We took all the things that we loved about Blueprint because we have tens of thousands, hundreds of thousands of clients using Blueprint, seeing that experience. We took all the things that were working in that experience, and we pulled it directly into Infinity Studio. There's now an incredible consistency across those experiences. We took the AI engine of Blueprint that has our best practices and increasingly our partners' best practices for how to design these workflows and how to apply them to different industry use cases.
That is now available directly inside of Infinity Studio, which opens it up not just for new workflows, but for our clients who are looking to evolve and extend all their existing apps on Pega. We added to Infinity Studio all the power of the AI coding agents that everyone loved. We take Claude, we take GitHub Copilot, we take OpenAI Codex, we take Kiro from AWS. We have our own offering there. We plug that right into Infinity Studio. The important thing is we're not just plugging it in, we're actually injecting through some really smart sort of MCP skills, all of the best practices of how to build in Pega directly into Infinity Studio.
As I'm going to show you in a second, I think it pretty dramatically lowers the bar of entry for people being able to come in and build meaningful mission-critical applications in their Pega environment. This is the only platform that handles that full life cycle, that lets you reimagine the work that you want to do, that lets you take AI and use that to accelerate your build. Not only accelerate it, make it better, ensure best practices are followed, ensure good architecture and design is followed as you build it out, and then run this stuff with a high degree of predictability in the outcome, a high degree of predictability in the cost that you're going to pay, and give you the foundation and the right architecture so that you can change it and evolve it.
One of the things we're seeing with some of these coding tools, they're really slick at building out a prototype of an app. You can ask Claude Code to go generate a bunch of stuff. Then a developer asks that Claude to go do something really simple, like change the label on a field. The label on that field is buried somewhere in a million lines of generated code. The more lines of code Claude has to go through, the less successful it is in consistently finding it. You as the engineer, even though you know this is a really simple fix, you didn't write the code, so you don't know where that label lives. What should have been a really simple change becomes an extremely complicated and expensive change.
Because of how we architected Pega, we avoid that because we know that our clients, the value comes not just in the first release, the value comes in the third and the tenth and the hundredth. We want to make sure that continuous change stays there. Couple of things I'll point out that we added into both Blueprint and Infinity Studio that I think are worth noting. First is this Integration Designer. I'm more than happy, I'm sure Kerim and his team would love to kind of take you through some of the gory details of how this maps into things like YAMLs and MCPs, and we could just talk about all the various three-letter acronyms this thing supports. Here's why it's important.
The number one challenge of making this stuff work in the enterprise is getting it integrated into the stuff that already lives in the enterprise. These mission-critical workflows are not self-contained inside of one system. They're not self-contained inside of a single platform ecosystem like Salesforce. They need to grab data from Salesforce, they need to update data that lives in an old mainframe system. They need to pull customer information from Adobe. They need to pull all those pieces together, right? That integration is both a large part of helping our clients understand early in the sales process the level of effort that's going to be needed in order to bring the solution to life.
It's also a big part of the work that then actually needs to get done by us, our partners, our clients during the implementation. Integration Designer both pulls that conversation earlier in the process, so we know even at the Blueprint phase exactly what the integration work that is going to be required is. It uses the power of AI to accelerate and automate a lot of that integration work so we can deliver faster and more accurately. This, I think, is a really powerful addition. The other thing that we've done is this Infinity Studio experience, which again leverages Blueprint both from a design experience but also from an AI engine capability directly inside of the Studio.
It does things like automatically generate your implementation plan for your applications, and then actually uses the AI agents that we can plug in to automate all the steps in that plan or many of the steps in that plan for you so that we can accelerate value. This is just going to keep getting smarter and keep getting better. I think this, again, in terms of accelerating that build process. I also think it's really important in AI, we love to talk about speed, and I think speed is great, and I think we want to continually compress the time between when we start talking to our clients to when they actually get value. I'm all about that compression. I think the thing that people also don't take into enough account is if you do this right, it also can help us build better stuff.
Blueprint doesn't just help you build faster processes. It helps you build better processes that are more efficient, that deliver results to your customer faster. What we've done in Infinity Studio doesn't help you just build Pega faster. It actually encodes the best practices that we know make for a well-architected application and helps ensure the application you deliver will be easier to scale, easier to maintain, easier to grow on. It's not just faster, it's better. The way I look at what we've been doing with all of these tools is trying to simultaneously lower the entry cost while increasing the value that we can deliver. Pega has always been able to deliver pretty massive value to our clients. We've been talking about 40% improvements in efficiency, 30% improvements in cross-sell, upsell rates since long before AI existed.
There's already a pretty massive return in benefit that we give to our clients. With the first release of Blueprint, we lowered that entry cost because we made it easier to conceive and understand what you wanted the workflow to do. Our initial conversations with clients got easier. We could more quickly get to the point of what we wanted the application to do. We added more capabilities to Blueprint that further lowered the cost and made it easier to get something closer to something that we could deliver. We also started adding GenAI connectors so your processes could automate more. Not only were we reducing the cost, we were actually increasing some of the value that we can deliver. With Infinity 2026, we've done that even further. Sort of step change in terms of both Infinity Studio starts using AI to automate the build.
Again, the entry cost, the skills needed, the time needed to deliver something goes down. If you start building those workflows with agents inside them, you're automating more work. You're automating things that previously were manual. We're talking to some of our clients about really massive, like eight-figure business cases floating around some of the agentic workflows that we're doing, so that value goes up, the barrier of entry goes down. That's the continued push that we want to make, not just for Infinity 2026, but 2027. That's our goal is to continue to sort of bridge that gap, because that's the value ultimately for our clients. That's why the clients choose is the gap in between there. I often get asked, not just by investors, but by customers, by my fellow employees at Pega. In a world that, as Alan points out, is incredibly noisy.
A couple of months ago, to me, the state of the market was greatly summed up when Anthropic dropped a blog post and IBM's stock price went down 11%. You guys remember that? Anthropic dropped a blog post, not a demo, not a product announcement. They put up a blog post where they said, "Yeah, Claude could help with the mainframe." Like that's literally all they said, and IBM's stock dropped 11%. There is so much noise and swings and hype in this market. It's really important for us to be very precise on where I think our differentiation is. I draw our differentiation in three areas. First, our ability to help our clients reimagine how they get their work done.
That is the profound step that enterprises need to take to move from sort of just we've deployed Copilot to we are actually transforming how the business runs. You actually have to reimagine the work. We have the best reimagination tool on the planet in Blueprint. Next, you need to be able to run that work, we focus on the mission-critical work because that's where the value is, that's frankly where the hard problems are, we're interested in hard problems. You have to run that mission-critical work with predictable outcomes and predictable costs. You guys are probably going to get sick of people on this stage saying predictable. We're going to keep saying it. Predictable outcomes and predictable costs. We are able to do that because of our grounding in a workflow architecture that is proven to operate in a regulated industry.
We are the audit system of record for how our clients do things like handle payment exceptions, how they handle fraud, how they handle due diligence and KYC and CLM. We know how to build a governance system that can withstand the pressure of a security and an audit team. We bring that now to bear on actually starting to deploy agents across that life cycle. Of course, we do it with a predictable cost because we govern the agents so that you're not burning time wasting reasoning tokens to figure out a process that you already know how to do. Then finally, we've got an architecture that allows us to continue to change. Karim showed in his demo, I think it's very interesting. It's easy to get excited about things like command line interfaces, I type into my chat, it does stuff.
Sometimes you just want to add a field to a screen by dragging and dropping it. We want to give you that flexibility so that our clients can change and evolve these systems and not build out this compounding set of technical debt that starts to happen when you've got AI agents generating hundreds of thousands and millions and millions of lines of code. We've got an architecture built from the bottom up to really support that change and support the scale that's needed to do this for real with the enterprise. That's the differentiation we are leaning into. That's the reason the clients are here talking to us over the next couple of days, and we're going to continue to build out on that advantage as we take our products and our strategy forward.
With that, I want to get Carie up here because Carie talks to our clients about this stuff each and every day, and I think she's going to just be a great person to show you what some of this stuff looks like.
All right. Thank you, Don. Hi, everybody. Nice to see you here. My name is Carie Whalen, and I lead our solutions consulting team here in the Americas. What my team does is sort of the technical side of our sale, and we stick with the customer through that sale and implementation and then obviously expand Pega with them. We spend a lot of time listening to our customers' challenges, what they want to work on, and then presenting Pega solutions to them.
Carie, I think what we should do is we should just start calling your team forward deployed engineers, since that's the term that everybody.
Right.
That seems to be the term of art for what everybody wants to do. When I hear you describe what you do, which is you're a bunch of technical experts who spend time with our clients to really understand their business, help them envision what their business could do, and then stick with them through implementing it and getting success. Boy, that sounds a lot like a forward deployed engineer in a lot of ways.
We do. We do spend a lot of time with the customers, both in pre-sales and post-sales. What I'm going to do, I know you guys have seen Pega Blueprint a lot. I know you've seen the demo that Kerim did this morning. What I want to do today is kind of talk more about what it looks like when I'm in the field with customers and how we work with Pega that way. Pega Blueprint has changed my life and my team's life because it used to just be that over there. Can you guys see that whiteboard? That was what I used to have to use to talk to my customers about their business challenges, and I'd whiteboard it out. I don't even have that great of handwriting, right?
It was circles and squares and trying to get those things in front of them, take a picture of it, document it, come back a week later and try and demonstrate something to them and see if it resonated. Now, imagine how much better my life is, my customer's life is, and my team's life is using Pega Blueprint, where I can just come in and say, "Hey, I'm talking to a banking customer in retail banking." I'm going to give you guys an example today that I actually worked with a customer on around their complaints challenges. A lot of banks get a lot of complaints, and in this case, we're talking about 2 million complaints handled per year at this bank.
Right now, or actually before Pega, they were using 19 different systems, and it would take weeks to process complaints sometimes in order to get that done.
Keep in mind, the complaints for a bank, it's actually a pretty fraught moment. One, because of the customer relationship, but two, because banks have an increasingly complicated set of regulatory obligations on behalf of a customer complaint. If complaints are related to fraud, if complaints are related to discrimination, if complaints are related to, "You sold me a credit card that I shouldn't actually have been allowed to have, and now I" All of those things are moments of truth, not just for the customer relationship, but actually for the bank's very licensure and existence. You have to get this stuff right.
Back to my interaction with the customer. I sit down with them, and I say, "Okay, what is it that you want to get done? Tell me a little bit about what you do today. Tell me what you want to do in the future." We write out a functional description of the application, which I've pasted in here. You guys don't want to see me type that whole thing. I tell it, "You know what? I need something that can handle complaints, collect the information, make sure we're communicating with the customer, and we need to do it in a certain amount of time." That's what we give the LLM. That's what we're giving the generative AI in Blueprint.
That's now going to take all of Pega's best practices that Don talked about and take that into that generative AI that's going to help me design it. We can also do more now. One of the capabilities that I love within Blueprint that I use a lot with customers is this supporting assets capability. Most of our customers have process documentation. They have COBOL code. You heard about that this morning from Unum. We might send that into AWS Transform. We then upload that information up into Blueprint. In this case, I'm uploading a complaint specification. I have a walk-through video of their current mainframe application that they're doing that I'm giving Blueprint, and I've done some AWS Transform on that.
All of that is going to feed into that generative AI design that Blueprint is going to do for me, again, without my trusty whiteboard, and bring back something that my customer and I can then react to very quickly. This is something we do in first meetings with customers to really get them excited about seeing their problems in Pega very quickly. The first thing you come to in Blueprint is a case. A case, just think of it as our way to organize all the information that our customer needs to get work done. It's like a digital briefcase of their data, their integrations, their personas, all the stuff you've heard about. We put all of that into a case, and then we build out what we call the workflow details. Blueprint's going to do that for me.
I want you to look at the whiteboard and see how blank it is when I walked into the room. With Blueprint I just don't have to react to a blank page with the customer. 90% of the time when I'm pulling this workflow up with a customer, they're saying, "Oh my gosh, that is my workflow. How did you know? This is amazing." So I get them to get something to react to very quickly, where we're not just standing there and asking them 100 different questions.
I think it's also one of the things we talk to clients a lot about, having spent a lot of time talking to clients about their processes over the last decade or so, the big challenge that clients have is also often a challenge of imagination, right? The subject matter experts on their teams have been doing this stuff the same way for years, their inclination often is to just automate the thing they're already doing, right? We call that paving the cow path in That's a very technical term. This is a moment where paving the cow path doesn't work. You actually need to rethink where these processes go. You need to rethink what you used to do. I was meeting with one of the largest banks in the U.S. They're using this to reinvent their KYC CLM.
The head of operations globally for all of their CLM work was basically saying, "There's a whole bunch of stuff in my process that we do just because we've always done it, not because we actually need to do it." Blueprint is her way of cutting, it's the weed whacker for all of that. It's like cut through all the mess, get me a clean version of what the process could look like so I can build something that is now ready for me to start dropping agents in. The other thing you'll notice is Blueprint is actually even suggesting where agents, all the purple spots on this are places where Blueprint is saying, "You could drop an agent in here. You should be dropping an agent in to do this." It's actually pushing you towards the agentic future.
Yeah. Again, it gives us something to talk about, it's great because the customer will be like, "No, that's not how I do it. I do this." We go, "Okay, I can go ahead and make that change." I can actually use an AI assistant to help me so I don't have to click and go into and code anything. I'm a business user I can get my business users from my customers actually hands-on sometimes. I'll just turn the laptop around and I'll say, "Why don't you tell the AI assistant what you want it to do?" In this case, I want to create an alternate stage that gathers more data and evaluates if the complaint can be auto-resolved if there's a low-value fee dispute.
I'm able to quickly tell the AI agent how I want to edit my Blueprint, it's going to go off and work on that. It's going to check with me before it does anything, because I don't want it editing something for me necessarily. It's telling me that it's going to add that additional step, I'm going to say go ahead and proceed with adding that. For the customer, they're able to see now an alternate stage that they wanted on the screen. It's already formatted. Again, they can react to it.
Yeah.
It's given them some new ideas about how they might want to handle it, where they can put those more deterministic rules in the workflow, where they can put those agentic flows in the workflow, do all of this during design time. Don, I was thinking about this when Alan was talking about design time versus runtime.
We spent a few weeks on this, really refining this with a customer and getting the design done and interacting with Blueprint. They're running it 2 million times per year.
Every time they get a complaint, this runs. That's a pretty big difference.
I'd much rather invest all that tokens getting the design right and then the 2 million times I'm running it, I certainly don't want AI re-reasoning through that workflow every time.
Yeah. The big takeaway for everybody in this room, I think on this is the AI capability in here that gives me that assistance as I go through the Blueprint. Also the guidance that I get from where I should put maybe some Agentic processes to speed up my workflow. The next thing, this is the new thing that Kerim talked about that I am super excited about as a former implementation consultant, the thing that takes the most time when you're implementing any product is figuring out how you build the integrations to the systems that you're going to need to interact with, right? It takes a long time to get those requirements. It takes a long time to figure out what that looks like. We are bringing all of that forward with this new capability in Blueprint with this Integration Screen.
It may not look that exciting, it is very exciting to me and my team because it means that we can get to that build that much faster. I can get an offer on the table during the sales process very quickly because I now know how many integrations I have to create. It's been a huge game changer for us in the field and for our customers to be able to see this. Just a quick refresher, if you didn't see it this morning. On the left-hand side, I'm able to actually take inbound events. Of course, I'm going to get files and things like that when I get a complaint. I'm able to define what those inbound inputs are going to look like into my workflow. I'm going to be able to define what my application data is and what I need there.
The integrations themselves, I can quickly add the different systems that I integrate with. Our customers have all sorts of different things in their ecosystem. They might have Salesforce, they might have Workday, we're always working to integrate those, having these pre-built integration capabilities as well as the ability to add anything we want from an integration perspective just speeds up that process. This is huge for us and for our customers in getting these things done very quickly and having those integrations defined.
If folks were in the keynote, one of the things that Omri from AWS talked about was what we've kind of been calling three three three, right? Three hours to understand a system, three weeks to build a prototype, three months to get something going. This is the kind of capability that allows us to drive that, because we're actually accelerating the hard and important work that it takes to get a client to both Prototype that they can really understand and get their hands on, and then ultimately take that into live. That's why these investments are so important for us, because they actually drive that capability, and that certainty of outcome that our clients are wanting to get.
Once we've defined our integrations, another important step is figuring out who's going to interact with this workflow. That's our personas. Blueprint gives us the ability, again, in the pre-sales process, we're already starting to build with our customer what it's going to look like, who's going to be able to interact with this, what data will they have access to, how can I set these up in the very beginning. We've got our customer service representative, we've got our compliance officer, as well as our customer defined here. Again, this helps us capture that and design that and start the build very, very quickly. I can also turn features on and off within Blueprint if I want to. In this case, maybe I want to turn on email. Then we get to a summary.
I've been up here maybe 10 minutes, maybe a little bit more. We just built the design of an application. When I show this to the customer and I say, "Hey guys, we just built five personas and channels. We just built a bunch of workflows. Isn't that cool?" They're like, "Yes, we just got through that so quickly." I say, "Okay, I'm going to blow your mind because now I'm going to" Oh my goodness, my chat is on. "Now I'm going to show you what my application is going to look like for you." I can go directly in to preview my app with the customer and show them that we're able to see the look and feel of what their users are going to experience with Pega.
This is sort of that jaw-drop moment for the customers where they go from, "Hey, that's a cute design. That's really nice. Oh, wait, it's an application. Can I touch it? Can I feel it?" We've moved from sort of showing customers what we can do with Pega. Oh my goodness. Don't know how to get that to stop. Okay. This is how we get the customers to get their hands on Pega very, very quickly. We can see what it looks like to go into the mobile application. We can see what it looks like to actually talk to your workflow. This is my favorite thing that we do with customers, Don, I know you like this too, where they can actually come in here and have a conversation with the agent, and interact with their workflow.
I can actually ask this workflow, how do I complain, right? We just built a complaint process. Tell me what it is I can do to-
My kids ask that question all the time.
I'd like to lodge a complaint. I can also call this Blueprint. I know that sounds crazy, but I'll actually pull out my cell phone and let a customer call their Blueprint and their workflow that we just created, and they're able to interact with it and say, "Hey, I have a complaint. I'm at this bank, and the teller's making me very angry." They can see how Blueprint will react to it and how Pega will react to it. It's a pretty effective sales tool and getting them, again, to experience it, not just display it. The other thing that, both Don and Kerim mentioned that is super exciting for us is this MCP agent integration. Think of this as the way that we just created a workflow together, again, in a few minutes.
I am now able to make that available for MCP, for anybody to call me to do a Pega workflow. If somebody is using Claude, they can call me through MCP. They can run the workflow that we just created. This makes it so quick and easy. Literally, you just copy and paste these lines of code and you're ready to go.
For folks who use Claude or any of the OpenAI tools, you know how you can customize Claude, you can add connections, you can add skills. Basically, the workflows that Carie just built in 10 minutes, and really five if I had stopped interrupting her, are basically now skills that you can use inside of Claude, inside of OpenAI, instantly. Right? This means now for our clients, they've got different front ends they're using, different agent front ends. Pega's workflows will power any of them. The instant you plug our workflow in, that agent starts acting predictably. That agent doesn't do a lot of reasoning, so the cost you're spending on it drops because it actually is just following a process. It's a way to start embedding mission-critical work into any agent that our client has.
We're at the end of our Blueprint.
Excellent.
Blueprint has given me some recommendations. It's even guiding me in the design process on best practices, like, "Hey, you may have missed a few things." We're good, though. I feel like we've done a good job of building this out. I'm going to take my Blueprint and pull it into that Blueprint AI capability that Don was talking about in Pega Infinity Studio. It may look the same as Blueprint, but what this is, and that's on purpose, is it looks more like Blueprint than it ever has before. In the before days, before Pega Infinity Studio, Pega actually had two different design studios that our customers could use, App Studio and Designer Studio. It meant that you had to go in and code some things. You had to be fairly technical to kind of hook things up.
With Pega Infinity Studio, we are making it very easy to design and build and produce your applications very, very quickly. It should look and feel a lot like Blueprint to you because that's what we're trying to do. If you'll remember, we brought in that workflow that we built around the complaints, right? You'll see the intake, the evaluation, all that stuff is in here. Now that I'm in Pega Infinity, I want to kind of refine this a little bit more, and I want to think about how I'm going to put this into production. Again, in moving into a world where Pega Infinity doesn't need the most technical users to implement this, it's going to make it a lot easier, open us up to a lot more types of users for Pega Infinity.
In this case, I'm going to use my agent to help me. I'm going to add an agentic investigation step, generate some test scripts, and wire in something we call the document summary agent. We have this new capability called DocAI that I'll show you once we get in there, that's going to help me as a user on the back end of Pega, analyze that stuff. I'm going to hit Go, and my agent's going to runAnd it's going to do my work for me, and it's going to come back and let me know what it did. Now, in order to go to production, I have to create some test scripts. I didn't want to have to do that. The AI agent just did that for me. It added my new investigation agent, and it's all right there.
Now, I want to give you guys the visual of what Don was talking about before. As far as hundreds of thousands of lines of code that Claude is creating. If I need to make a change, it's a heck of a lot easier to come in here and look at a workflow and say, "Hey, I actually need to change my business rules in this initial valuation." I can click on the initial valuation, I can come in here, and I can edit my business rules. Right? It's a much easier way to maintain the change that's going to happen in these regulated industries. Maybe I change my rules in the bank as far as who I want to allow to have a specific credit card, and that business logic is very easy and visual to change.
Not just change, transparency, right? Because if I'm a bank, if I'm running mission-critical work on this stuff, I'm constantly being asked, "Well, what are the business rules? What are the controls? What are you checking?" Right? I can just literally show it to you. I don't have to go find it somewhere in a bunch of code that was written by an agent where I don't actually know where the rule is anymore. Right? That, again, for this type of mission-critical work that our clients do is a game changer in terms of how they can use this stuff with an increasing degree of certainty and confidence.
We built our workflow, we're going to put it into production, one of the things I want you to see as part of this is we've got a combination of those deterministic rules, right? I have very specific rules as a bank as far as how I'm going to handle a complaint. I'm not going to let an agentic reason come up with the way that I'm going to handle a complaint, right? That's built in. We also have agentic capabilities already in here. I want to know if this is potential fraud, right, when the complaint comes in. I can actually call an agent that's going to calculate the fraud risk score for me. I can also have an agent that goes off and does an investigation for me and get that done.
After we have built this out, we are able to then launch this into production and get it ready to go for the customers who are going to interact with it. Again, that could be in any channel.
Yep.
Right? When you think about our center-out architecture, you have to think that we're building this once, right, in Infinity, and then we're able to then push it out during implementation into any of those different channels.
Increasingly now that channel is agents, right? That MCP connection that Carie showed you now means that for our clients who are saying, "Hey, we want to deploy self-service agents." In fact, we're working with a bank on building an agentic complaints workflow like this, and the way that customers are going to enter those complaints in is not by a form. They're going to go to an agent and say, "I want to register a complaint." Now we plug these workflows in, so when the client says that, the agent isn't randomly asking himself, it's going through the workflow and saying, "What data do I need to collect? What's the steps I need to do? What do I need to tell the client is going to happen next?" All predictably driven through that agent.
We're going to take a look at that live on my U+ Bank website. This is the bank that Pega's created, to show you what this might look like. U+ already has their own chat center set up, and this customer comes in, and I am upset, and I want to lodge a complaint with U+ Bank because there's been a charge on my credit card. I can very easily start a chat with this chat agent. Now, on the back end, what Don just said is it's going to use that workflow that we just built to come back to me. It's empathetic. It says, "Hey, Larry, I'm sorry that you got charged that $29 late fee. Is that the one you're referring to?" It automatically kind of understood what I might be complaining about.
It also allows me to respond and take in information from me and go back and forth, again, using that workflow. In this case, I'm asking that to go ahead and process that. That's kind of how we integrate with the different channels. That could have been, again, the phone, that could have been a different interface. In this case, that's what it looks like to be a customer interacting with the workflow. I'm going to take you to the back end of Pega if I'm in the bank. We are users of Pega in the bank, the people that process those complaints within the bank, and we are going to actively work through this complaint. What I see when I log in as this complaints adjudicator is all of the different complaints that might be out there for me.
I can see the work volume that I have, and I can even see notes on some cases that are out there. I see that there's one coming in from Larry Clark about his dispute, we're going to go ahead and get that work done. Again, on the back end, this is the workflow taking me through the process of getting that work done. One of the things I can see as a user is the supporting documents. Again, I plugged that DocAI capability into this application, it'll automatically summarize the documents that I'm feeding to it, giving me a very quick view of what that looks like. What I'm showing you guys on this back end is the ROI that our customers are seeing when we automate their workflows and make this that much easier.
Remember before they were using 19 different systems to go in and find that file. They were trying to find out where Larry was. We're bringing that all together in one place for them, and that's really the impact that we have for these banks from an ROI perspective. I'm going to go ahead and work on Larry's case. It happens that I'm kind of new to the bank, this happens a lot with our customers when they have customer service representatives with very high turnover or back office workers with very high turnover. It's hard to keep everybody trained. Luckily, we're using Pega, I have this GenAI coach that's going to use that generative AI, that LLM, to help me summarize the case. Then I don't know what to do next as the user.
I can actually ask the GenAI coach, "What's the next step I should take?" We can actually guide the back office users, again, making it easier for them to do their jobs, making it faster for them to do their jobs, and complete these complaints that much more quickly. It's telling me that Larry's a premier customer. We should probably waive the fees and say we're sorry, maybe give him a little extra. I'm going to go over here, again, a time-saving capability that we built in here is I don't have to type up my notes. I can actually just fill out the form with the AI and double-check it as a human and make sure that that looks right. We have worked on Larry's dispute. We're now going to be able to get back to him very quickly.
I'm going to submit that, and the claim is completed, right? We've addressed the complaint. To finish our story with Larry gets a text as part of the workflow that says, "Hey, Larry, we've taken a look at your complaint, and we were happy to refund you that money." You guys have seen how we are changing the design that we're working on with customers upfront, taking a lot of time out of my sales process, making it much easier for customers to get their hands on upfront. We built that integration step into the design to make sure that we can build more quickly and get those offers out the door more quickly for sales.
You've seen what it's like to be a customer that interacts with Pega, as well as a user that interacts with Pega to really get that work done more quickly.
Awesome.
Thank you very much, Carie.
All right.
All right. Ooh, yeah. Wow. There we go. Couple of quick things as we get the slides up. There are two questions that come up inevitably in investor conversations, and even sometimes when I'm talking to clients. I thought I'd just share with you how we are answering them when they pop up. The first is, we get asked, "Well, could I just code all this?" Like, "Do I need Pega? Could I just have Claude Code write this for me?" In many ways, if you look at Pega's history, we've been competing against build-it-yourself basically our whole existence. That's pretty much every year I look at our competitive data, and every year, the number one competitor that we have is a build-it-yourself option, regardless of what other vendors are out there in the market.
To me, even with the Claude Code capability, this comes down to two arguments for me, and they both are TCO, so I call it TCO squared. The first is the total cost of ownership argument, I'm going to show you a little bit about what I mean by that in a sec. I think eventually everybody has to show an iceberg slide, so I'm going to show one. Also, I look at TCO as meaning transparency, change, and operations, right? The transparency for what our clients do is so important. They are running mission-critical work. They have mission-critical rules, and having those rules buried in a bunch of code that their auditors can't see doesn't work for a lot of these regulated and mission-critical processes. With Pega, even if you're using Claude Code to write it's not writing code, it's writing visual assets.
It's writing business rules. It's writing processes. You can still go in and see a business person can review and validate that this is doing what they want it to do. That transparency also means this stuff is easy to change. As this evolves, if I need to add fields, add steps, change rules, I'm not hunting back through millions of lines of code that I didn't write. I'm actually just going right into the visual element, and I'm making the change. Or now I'm asking my coding agent to go make the change, but the coding agent knows exactly where to go because the metadata and the context is all captured in the rules. Finally is the operation piece, right? The TCO of this, I think we often get distracted by just how cool it is that Claude can write code. It's great.
We use Claude to write code. I've been in engineering for a long time. The typing of text into files is easily less than 20% and probably less than 10% of what it takes to build an app. There's a whole bunch of stuff at the top of the iceberg that we don't even see, which is actually, what do you want to build? What is the right process? What are the rules we need to obligate to? What's the outcome we want to deliver to our customers? By the way, most of these stories about people using Claude Code have them going off by themselves over a weekend and building stuff.
Yeah, how do you get 20 stakeholders in your business team to agree that you've built the right stuff, that the process is right, that the rules are right, that the teams are going to actually support and run with the new application? Well, you need a design tool that actually lets people see and test and validate and prototype and debate and arrive on the business process. That's what we're doing with Pega Blueprint, right? We can do the build stuff. We can now use Claude Code to accelerate the coding stuff. The other stuff, the under-the-water stuff of actually operating a mission critical, you have work lists, and you have to have service level agreements. You have to have all this stuff that Pega just does.
By the way, none of those capabilities are differentiating for a bank or an insurance company or a telecommunications company. Their teams are spending their time building that. They're not spending their times on things that actually differentiate their business. We want our clients to focus on designing and building processes that meaningfully change the efficiencies of what they do and how effectively they respond to and engage and support their customers. By providing a platform that does that, we allow them to focus on the stuff that matters, not maintaining the stuff that doesn't. The other question that we get is, what about this purely agentic reasoning? Do we even need workflows? Maybe in the agentic world, there is no workflow. Which sounds really cool, except the fact is that businesses by their nature are deterministic.
There is a lot of stuff in the business that is deterministic, not probabilistic. Not everything. There is some work that probably could be purely agentic. Some of the work that we do in marketing that actually involves coming up with creative designs and like that, yeah, I can use agents in a lot of that. Actually, I do eventually need a workflow because I want to make sure that the right people approve the asset and that legal's looked at it and a whole bunch of other stuff has to happen. A lot of stuff that needs to run in the business is deterministic. It needs to be predictable. I want to follow a set of rules. If I use my reasoning at design time to get those rules and processes right and reimagine for AI, then I can execute them predictably, both with predictable outcomes.
Guarantee that I'm going to follow the rules the bank needs to follow, but also predictable cost. I don't want to be burning the reasoning tokens on rethinking the workflow. We put a little calculator on pega.com that does this because it's important to understand this calculation is quadratic, it's not linear. What happens is, if you have an agent reasoning its way through a long-running workflow, at every step of the workflow, the context window that agent needs to manage gets longer and longer. Step of the workflow adds context to the workflow. Now the context window is longer. Next step adds it. Every step, the agent needs to reread the entire context window of the workflow to figure out what next to do next.
If I've got a 30-step workflow, the agent power required to think through that 30th step is 30x what it started me when I was beginning. That means that your agent reasoning costs go up quadratically as the workflow gets longer. With Pega, we don't do that. We manage the context for you because we know the context, we know the steps that need to happen, and we focus the agents when we use them on doing the very specific thing in that step. The context window stays small. The context window staying small does two things. It minimizes drift, which is what causes hallucinations and inaccuracy. Two, it dramatically reduces the cost in terms of the number of tokens that you need. With that, to talk a little bit about that, I'll hand it over to Ken to talk about tokenomics.
Thanks, Don. One real quick thing that Carie was talking about earlier. The process before Blueprint was there would be a whiteboard session to use the example. After the whiteboard session, then we would close for what would be to go away and try to build a demo that might take months, that by the time we go back and show a demo, you show it to the stakeholders and they say, "I don't remember putting that on the whiteboard." There was this iteration process of how Remember, all of that time, we're not able to put a financial offer in front of the client. That's the big change that we've seen is getting an offer in front of the client so they can consider how we can help them as fast as possible, because they're looking at all kinds of other alternatives.
I think naturally, without AI, this would have been hard to get to, but this is really not a driven-by-AI thing. This is actually the way clients want to experience transformation. This is something that quite frankly, we would have done. We were moving in that direction without AI. It's just that AI really was a big compelling event for that. I just want to clarify a couple of things on tokenomics, then we'll get into some financial discussions on the model. Don touched on this. As prompts, prompt chains, and processes expand, there is a geometric calculation of how much the agents actually work, of how many tokens are used. Because when you do one process, you have a single prompt. As you add multiple processes, the spider web of the prompts and the repeated prompts are pretty excessive.
Don mentioned we have on our website an AI cost estimator. When you think about the cost of using prompts versus using workflow, and keep in mind, it's not that we're not using tokens for reasoning all through the process versus the way Pega does it is we use AI where it is relevant and where it is necessary. Some examples of, in this model, we ran a very specific set of steps and stages in a workflow, and we modeled out what that would be if you actually re-reasoned and went through the process. It's anywhere, I think we're a little bit conservative, actually, on the 151,000. It's a multiple of cost. As the process or as the workflow becomes more complex, that gap becomes bigger. Because the more you're pushing scale transactions through, those are done through the workflow without any tokens.
The more transactions. I think this is just a kind of a framework of how our calculator works. I want to jump in. I think we're even running a little bit past our time, but I'll try to be respectful of getting everybody out of here by 2:00 P.M. By the way, Peter, what time is it right now?
Just after.
1:21. Okay, we got 40 minutes. Once again, repeat safe harbor statement. Peter talked about it earlier. The reason why it's here again is because we filed this last section as an 8-K, and I noticed that some of you have already actually pulled up the 8-K and are starting to look at the slides, so you're probably ahead of me on that. I want to go back a little bit to the conversations we've had. I think one of the most misunderstood pieces of Pega's story. I always get asked, "What do investors not appreciate? What do investors not understand?" Understanding the solution is always a challenge to make sure that because investors are at different points. Some of you are very technically astute. Many of you just look at the financial modeling. It just depends, and everybody's a little different.
I think one of the things that I think is very misunderstood is the ability for Pega in the business model we have to generate significant amount of free cash flow, and for that free cash flow to be highly leveraged as the company scales. I think that is more understood now than it was three or four years ago, but I still think it's very misunderstood. I wanted to just highlight a little bit of just going back of what we said over time and how we've actually delivered on those results in the last few years. We went through a pretty significant transformation in our business, as many of you know, where we moved to a subscription business, and subsequent to that, we moved even harder into Pega Cloud. That business in 2018 was about $570 million of ACV.
We actually modeled out pretty closely to where our trajectory has been through the first quarter in terms of we gave three scenarios, if you'll remember. We gave a base case, a bear case, and a bull case. It was basically a range. If you look at where our trajectory is now, it is pretty much at, I would call it, the base case or the mid case that we actually modeled out. Really what we have seen is as we have moved along this subscription journey, our Rule of 40 measure, which we have used and we have talked about for probably going on 10 years, you have heard me talk about this. We moved from what was a 10+ % Rule of 40 calculation to now we are trending over 40%.
In a business, it is an enterprise buyer with a very high retention rate, with highly leveraged components of our costs, like our gross margin is leverageable, our R&D costs are leverageable, our sales and marketing costs are not completely variable to booking, meaning you do get some leverage there. These are all components that help us continue to drive that free cash flow margin up. Pega Cloud has been a big part of this. Why is Pega Cloud so critical? Pega Cloud, one, it is a leading indicator on where adoption with our clients is going. Naturally, our ACV growth is less than 29% because Pega Cloud is growing much faster than the other components of ACV. As we continue to expand Pega Cloud, that becomes a bigger and bigger force for potential ACV acceleration.
You will see on this chart, from Q1 of 2024 to Q1 of 2026, we have had a pretty noticeable increase in Pega Cloud ACV. The important clarification here, some of that is because clients have begun migrating more consistently to Pega Cloud. The majority of our Pega Cloud growth is net new spend on Pega Cloud. This is not just a solely migration plan. The reason why you can validate that is you can look at our term ACV, for example, and you do not see our term ACV precipitously declining, and it is actually even growing slightly. You can tell that there cannot be that much migration happening from term to Pega Cloud, or else you would see a pretty steep drop. You will see maintenance ACV going down, but it is still going down mid to high single digits, and that is our smallest ACV component.
The reason why there is a difference between term and maintenance is because maintenance has some legacy perpetual buys that happened years ago. Maintenance would actually come down faster. What is the big lever point we have with Pega Blueprint? We have struggled at Pega to expand into net new organizations. That could be a new buyer within a client of ours, or it could be a brand new logo. Why did we struggle with that? Admittedly, the reason why was the conversation on the whiteboard discussion. If we are going to target a net new organization before Blueprint, what we would do is we would hire a salesperson. That salesperson would have to be trained up to be very technically astute because remember, there is no Blueprint.
They have to actually understand the vertical, understand the Pega platform, know the horizontal use cases potentially, and then be somewhat vertically industry aligned. That salesperson has to spend. In the past, we wouldn't even send them out into the field for sometimes six months to nine months because they were trained, they had to get certified on Pega, which means they're not building pipeline until maybe the end of the first year into the second year of their existence. When we were going to hire new sales teams and attack new logos, we had to know that the first year we were probably not going to get much production out of that salesperson and hope that we got it in the second year. That's a very costly proposition to expand into new logos. Now, with Blueprint, you saw what Carie did.
We can actually close, we have numerous examples of a brand new logo just in the last few quarters going in, pitching Blueprint in a first meeting. Two or three slides on what Pega is right to a blueprint that has led to us closing and getting clients to live in many cases inside of 90 days. I'm not talking about the entire enterprise customer service for Bank of America applications, but these are enterprise companies. These are not SMB businesses. This is a big change that we're seeing around Blueprint and the ability to use it to get in to discussions, to new discussions, to people that don't know Pega, without a sales team that has to be really technical around, without using Carie's team or our professional services team to build custom demos to shrink that time to when we can build pipe.
This is a really important opportunity for us at Pega to be able to get into organizations that don't know us, that don't know Pega. I'm not suggesting that that's easy. We don't have a household brand. We are spending time and effort to try to improve the brand, but we're not a brand that everyone knows who we are. It's not an obvious thing what workflow is and what that means and how that plays into transformation. There's an opportunity to really attack a lot more organizations than we've ever been able to do. Pega Cloud is our exclusive pricing mechanism with any new client. There are situations where a new client may choose not to buy Pega Cloud for some reason, like a government agency that might not be able to run something into the cloud.
It does happen, our primary pricing mechanism that we use and our exclusive offers that salespeople start with our clients is Pega Cloud. It really means that any of our new growth with new logos is going to come off of a Blueprint that's on Pega Cloud. It's going to lead into Infinity Studio, which is on Pega Cloud, which is going to leverage either our AI or call to their own AI models on Pega Cloud. Pega Cloud is a big accelerator tied to Blueprint. Some interesting early insight we've had around new logos. Our total pipeline year-over-year is up just under 30%. Some of that is driven by new logo pipeline being up 65%.
I will be very transparent that new logo pipeline is admittedly of a somewhat lower quality than pipeline that may be an upsell or an expansion off of an existing application if a client, say, is expanding, adding a case type, growing. This is what we hoped to have seen, which is the ability to drive significant new logos that we are engaging. When we are looking at pipeline, we are talking about where there has been a conversation with a client where we have structured an offer to that client, or we are in the process of structuring an offer where the solution is known and the size of the opportunity. This is not pipeline where it is like, I hope I can sell $10 million to XYZ company.
This is the early signs of us really seeing new logo engagement, driving pipeline that hopefully will then lead to accelerating our growth. What has to happen here is we have got to basically really prosecute this process and make sure we are learning from the engagement that we have with these new logos, and we are understanding the competitive dynamic because it might be a very different competitive set than our traditional kind of expansion with existing clients. On this slide, is not included on this slide, is any momentum that we are starting to see from autonomous partner selling. Alan showed on stage the partner Blueprints. If you remember, he talked about the branded Blueprints like Cognizant, for example. That program has just been kicked off, and in some cases, we are seeing really good momentum.
In other cases, we are still kind of working through the process of getting in front of the sellers of the system integrators. The system integrators are very much changing their business models. Time and material is going away, right? They need to sell outcomes. They need to take more risk to help clients transform. We are putting Blueprints in front of those sellers and really trying to get that co-selling motion. Co-selling meaning we help when needed, but they are the ones engaging with the client. That is largely new logos as well. That is not in any of this data. Once again, that is early. When we modeled the year, we did not expect to get any contribution from autonomous partner selling in our guide.
That is something that we are watching closely, but it is a really big opportunity because that is how we can get to the 10,000-15,000 clients through partners and really, quite frankly, leveraging their brand. A lot of times people talk about moats. Why does Pega exist? Why can they continue to exist? Why won't you be destroyed by AI? I think that is a question that, quite frankly, has nothing to do with AI. You should ask that question of yourself in any business. Why are we different? How do we win? Objectively, we are rated always with one of the best, if not the best product in all of the segments that we operate in. We are built around repeatability, consistency, governance, trust. Workflow executes consistently, predictably, and that is very critical for enterprise organizations.
Don used the example, I think, on one of his slides, he talked about a 98%, or he might have used 95% success rate, which meant a 5% error rate. Alan and I met with one of our larger banking clients, the CTO and the CIO, and he talked about how his team was bringing these AI projects to him. He said, "One of the struggles I have, though, is they're bringing me use cases saying, 'AI can get this 90% right. AI can get this 92% right.'" I don't doubt those numbers. His comment to us was, "What am I supposed to do with the other 8%?
How do I actually go to a regulator and say my system will execute this way, and I'll screw up 5%-8% of the time?" Now, there's a difference when you say, my system is built to execute 100%, but we will have errors. We will have humans. We'll have fraud. We'll have mistakes. That's different than actually saying the system is built to not execute at 100%. I think for regulated, for deeply critical process or control workflows, which is the majority of what we do at Pega, that's critical. Build for change is one of those really underplayed pieces of If you want to change something, Don and Carie talked about a little bit. You can go in and change it right in the UI.
Imagine having to go in and deprecate code or try to deactivate code and write new code, and how would you know what's actually caught? Even if you look at the AWS Transform tool, it ingests COBOL code. One of the things it can't do well yet is it doesn't know how much that code is being called. It actually can tell you what the code is supposed to do. There's other tools that you have to do to try to see how often is the code actually being used. Very complicated to try to change anything that's code. By the way, was one of the fundamental differentiators of Pega and the low-code or no-code solutions that we've sold for decades. We understand industry use cases. One of the biggest value propositions of Blueprint is when you go into Blueprint, you're not doing a Google search.
You're actually calling on Pega's proprietary knowledge of how to do that work. That's a very important differentiation. We view Pega as being the orchestration of these agents that are kicking off or asking a question or being requested by a user to actually go into some level of work. Pega is the guardrails, the harness, as you might call it, to ensure that AI is driving or initiating the right. Then when the activity happens, the activity actually happens in a predictable way. One of the important takeaways I hope that everyone will take away from today is we've made commitments on how we're going to improve the structure of the business, and we feel like we've delivered on them. We're going to continue to focus on the importance of us delivering on what we said we were going to do.
The 2022-2025, those are actual numbers. Back three years ago, we set targets for 2027 and 2028. We said we wanted to be a gross margin business of 80%. We're approaching that now. We said sales and marketing, we wanted to get to 30%. At the time, we were in the 40s. We're down to 30%. I think there's more room to go there. I think AI is a powerful enabler for sales efficiency as well. R&D, we said we were going to be at 17%. We were above 20%. We're down to 16%. I think there's a lot of AI efficiency that we've started to see. AI in development can come one of two ways. Either you can actually get more done in your roadmap and/or you can actually reduce the amount of people needed to actually get the work done.
Right now, there's a kind of a balance that companies are taking to try to figure out where they should go on that. I think we'll get likely a little bit of both of that. Our operating margin, which was single digits, is approaching 30%. We feel like we've made bold commitments on what we said we were going to deliver, and we feel like we've delivered those. I'll show this as a cash flow slide. In 2022, I said Actually, it's 2021. 2021, I said, "This business can generate $300 million of free cash flow in a few years." Some of you in the room, probably met there's a lot of new faces here, but a few of you in the room and other analysts or investors came to me and said, "Why would you say $300 million?
There's no possible way you'll generate $300 million." It's almost you're going to lose credibility by saying that. When we got to $200 million, I said, "This business is going to generate $500 million in a few years." I had some of those same investors, by the way, some of which are actually longstanding still investors of Pega, came back to Peter and I again and said, "You can't. That's a ridiculous number. You're not going to get to $500 million." We did $491 million last year. We put $700 million out there. We had people come back to me last year and say, "Why do you keep, like, $700 million? I get $300 million, congratulations. $500 million. How are you going to get to $700 million?" That was before we guided. That's before we finished with $491 million.
At the time we did that, we were tracking to $440 million of free cash flow for 2025. We put $700 million out there. The reason why we are confident in our execution is because this business model is a model that has a level of predictability and visibility to it. Not every business model does. If you're Dunkin' Donuts and you have to bank on weather and how many people are going to buy coffee every day, that's not as predictable a model given macroeconomic changes. Our clients are not subject to deciding they're not going to do disputes anymore, or they're going to shut down their customer service application. We are always at risk to competition. We are always at risk to switching. We're always at risk to pricing pressure. The foundation of what we have is a very sticky platform.
The ability for us to have that visibility and manage our cost structure, which admittedly over time was not really one of our core competencies, I think it is now. I think that we've got to just keep executing. One of the really important takeaways is we've made commitments on how we are going to generate free cash flow. We've delivered on those commitments. We continue to be focused on this slide. This is one of the most important slides to me personally around how we can show to ourselves that we can run a good business. Naturally, we're doing it for all of our investors, but there's a pride level here of running a good business, and this is very important.
If you go through the halls of Pega, and you ask what Rule of 40 is, I guarantee you every single employee will know. I guarantee you they'll understand specifically the calculation, and they'll know how important this slide is. This is a hard needle to thread in an organization because you want to grow. You want to put investment into growth. It's critical. You have to have a business model that allows you to leverage the investments. That's why we look at Rule of 40, and that's why we're so committed to managing the business with this level of discipline. The next piece of this is, you're starting to generate cash now. Let's go back three years. With $600 million of debt, we had maybe $300 million of cash. Our net debt was between, say, $300 million of net debt.
We weren't generating a ton of cash. The cash flow started to increase. We paid down the debt. We started buying back. Last year, we paid down $470 million of debt, and we bought back a few hundred million. This year, I think we bought back $170 million in Q1, I think somewhere around there, $167 million. You're starting to see share count come down. The reason why you're seeing it is because the other piece of this is we don't have a significant amount of stock-based compensation as a percent of revenue. If you compare us to our peers, we're in many ways half of what some of our peers are in terms of the amount of stock we'd have to buy back just to offset dilution.
Our focus is to pay a dividend, and we've heard from a number of investors that having a consistent and methodical increase in the dividend shows a level of confidence in the free cash flow. To optimize debt, we don't want debt. If we have debt, if for some reason we decided to do a transaction or accelerate buybacks or whatever we might do, we'll pay down debt. We don't believe that debt is necessary as part of our capital structure. We will use a substantial amount of our free cash flow to opportunistically buy back shares. Naturally, the way we structure it is we do the typical 10B5-1 plans that you've seen other companies do. We buy more when the stock is lower, we buy less when the stock is higher, and we have a lot of flexibility there.
There's a meaningful opportunity for us to take down the amount of shares, which increases free cash flow per share, which is, I know, important to all of you. Just a quick snapshot of that. Over the last three years, we've returned $600 million to shareholders. In Q1, it was about 80% in the first quarter. Naturally, that will change because our cash flow is not consistent quarter-to-quarter. Q1 and Q4 are our two biggest cash generation quarters. Just highlighting the use of proceeds, so to speak, from this significant cash generation business that we have. This is really just highlighting that point of. Some of you have asked a very good question, and I'll maybe just give you maybe one-minute clarification on something.
You said, "You bought back approximately $500 million of stock last year, but the share count stayed relatively steady." What happened is our compensation plan typically gives up to 50% of our compensation is in options. We had a series of options that got exercised in 2024 and 2025, which, as you can imagine, took our share count up. That's not something that would happen every single year. You saw in Q1 where you had more of a normal quarter where you actually took down. By the way, Q1 is our biggest vesting of RSUs because we have an annual vest that happens in March. The other quarters in the year are much lower. You're seeing this trend now of as we buy back shares, the share count coming down. I think you'll continue to see that.
There's a really meaningful opportunity here to use this free cash flow to take share count down. There's another question that. To be completely honest with you, I wasn't as in tune with how egregious some of the stock-based compensation is of some of the other companies in the technology space until I continually heard the point in the last year. I know all of you know that with multiples coming down, there's a lot more focus on the after-diluted free cash flow. I think many of you are focusing a lot more on that. As I started to hear that, independent of that conversation, I've always focused on the way to think about our measurements is the percent dilution of our market capitalization and the stock-based compensation as a percent of revenue.
When you compare these numbers to our peers, I think you'll find that we're probably in the top quartile of companies that sit around Pega. I think that does not mean that we are underpaying our teams for stock-based compensation. What I think that is really a signal is that I think we've used stock-based comp as a proper and equitable part of compensation and not over-enriched large populations of the workforce with not paying attention to this number. I actually applaud all of you to actually look at the after-diluted free cash flow, because I think it's been ignored for way too long, and I think it is a number that's of use.
We feel like we're in really good shape in terms of how we think about our equity pool and how much we use for stock-based compensation and how that helps us be able to buy back more shares after we offset dilution. The summary of this is we have a recurring business model that is heavily influenced by Pega Cloud. Pega Cloud continuing to track towards that 75% number that I had talked about a few years ago. The operating margin, free cash flow margin, those actually are starting to become closer aligned now because the volatility of the model is not as bad on a 12-month basis now that we've got through the major pain of the cloud transition. The operating margin will continue to expand. We'll continue to get operating leverage, which drives then into our free cash flow margin.
Our free cash flow margin right now is about 30%, ballpark. There's no reason why this business shouldn't be a 35%-40% free cash flow business. I think we have all the dynamics in the business to be able to support that and still be a double-digit grower. I think that's a very valuable business that we'll have if that comes true. Then using that capital to be thoughtful about where we put it to use. Naturally, if we saw an acquisition that made a tremendous amount of sense, then we would decide to do that and slow down buybacks or take on some debt or whatever was required. It's hard in the environment we're in, where multiples are, and certainly Pega's multiple, to argue that you could get a better return than actually using that cash to buy back the shares of Pega.
That's certainly been our priority over the past year. Key takeaways. Pega Blueprint gives us an opportunity to get into markets that we couldn't get into before. This is not a Q1, Q2, Q3, Q4, Q1 of 2027. The business can be structurally changed to be able to attack a market that we could not attack. I think it's really important to not think about this as Blueprint will just somehow help us in the next 90 days or the next 180 days. This is how we want to change the nature of how we target the organizations that we can attack. The subscription transition, I would say it's largely done. Naturally, as you're moving to Pega Cloud, you're going to have a little bit of variability there. But this has really helped us anchor significant profitability and significant free cash flow.
Using that free cash flow to be thoughtful about capital allocation can help juice up the free cash flow per share by actually taking down the total number of shares and creating a natural buyer in the marketplace, which is us. This is who we are and how we're executing for all of you, and I just wanted to reinforce how committed we are to continue to drive value for our shareholders. We believe you cannot drive value for shareholders without running a good business, and a good business means that you're making difficult trade-offs and that you're driving margin and free cash flow in the business. I'm going to stop there. How many-
1:50.
Okay. I'll open it up to any questions that you guys might have. Devin, why don't you-
Steve Enders has a question .
Yeah.
Wait for the mic.
Yeah. Why don't we just start with Devin, and then we'll go to Steve.
Sure.
Yeah.
Great. Devin Au from KeyBanc Capital Markets. Thanks for doing this. Really informative. I just want to maybe go back to the slide of where you show all the go-to-market metrics influenced by Blueprint. It's great to see total pipeline up 29%. I think new logos up even more.
Yep. 29, 65 and about double in the number of new logos in the pipe.
Yeah, really great to see. David, my question here is, when I look at those really strong metrics and when I look at your near-term guide, ACV in the mid-teens, can you help us understand where the main gating factors of conversion lag or maybe what you're doing to accelerate that lag in the pipeline?
Admittedly, much of this pipeline change is early stage. Early stage, and we don't have great statistics or data on the conversion rate. If we had five years of data, we'd have a lot more confidence, Devin, in terms of how we think conversions. I would say that this is a great metric. It's also riskier pipe. It could turn out that the win rates are pretty reasonable on this. It could turn out that our win rates are not as exciting as we want them to be. We're not really there yet to know. The point of this is the number of logos that we're engaging with, and we are growing pipe significantly, and that will impact the overall pipe because new logos will become an increasingly bigger percentage of our overall pipe.
Thank you.
Yep.
Steve?
Great. Steve Enders from Citi. Going back to the free cash flow discussion.
Yeah.
It seems like there are a few areas where you were calling out there's incremental margin opportunities for sales and marketing and R&D and getting efficiencies there. I guess, why not maybe update some of the guidance on the free cash flow side if you're seeing that incremental opportunity, and maybe how does that kind of inform how you're thinking about incremental investments as you kind of get through the efficiencies versus letting that flow through to the bottom line?
We wouldn't be messing around with 2026 at this stage, Steve. What you're really talking about is the $700+ number, which is why I put a plus there, right? Which is naturally how fast our ACV grows will be dependent on that, and how much AI actually drives efficiency in our business will drive that, and naturally, us running a good business and making the difficult trade-offs. Because we don't want to get in the habit of constantly changing the number just slightly. We're just highlighting that we are still on track for that commitment that we made last year. You can kind of see our, I would say, our credibility building to where we are now.
Hi, thanks. Pat McElwee, William Blair. First off, is there a plan to monetize the Pega Blueprint song you were playing earlier because it's shockingly catchy?
Sorry, Pat, what?
A joke. The Pega Blueprint song you were playing earlier shockingly catchy. We heard customers on stage today talking about how scary it can be to endeavor into the modernization of legacy apps. You've long talked about the kind of pot of gold that is that legacy modernization opportunity, and it seems like the capabilities of Blueprint and Infinity are pretty consequential in actually unlocking that opportunity. My question is, how much of this recent acceleration in your cloud ACV growth is attributable to that? How early are we in addressing that opportunity, and as you think about the trajectory of the business, how much of that is still to come?
I think very little of what you've seen so far would be material legacy transformation that's been driven to Pega. My read on talking to clients is they are very focused on it, more than they ever have been, but still not moving as fast. It's not like they've a light switch, and they're just transforming. I do believe over the next five years, it's going to accelerate. You heard on stage, one of our clients talk about COBOL developers and the fact it's out of the curriculum of 85% of college classes, and that there's nobody graduating with COBOL experience, and nobody wants to do it. Quite frankly, COBOL engineers are retiring and, quite frankly, dying because the code because it's so old, right? I think that is a very big catalyst.
I also think Mythos and some of the other tools that are out there that have really been able to expose vulnerabilities. All these systems have unsupported open source that's unsupported. Old versions. Lotus Notes. There's 50,000 companies that have Lotus Notes. That isn't even a company anymore. These are tools that you can't really let accessible to the internet. You can't then tie it into data lakes and AI and all the things you'd want to do. I think that is a very big compelling. That's where AI is the most compelling because it makes you get your systems modernized, get your data together so that you can actually mine your own data. I think that's what's different. We've talked about legacy transformation, and we've talked with AWS, and Matt would say, or well, originally Andy, when Andy was running AWS.
He would say, "We're 10% of the way there." I think AWS would still say we're 15% of the way through the transformation, but I do think in the next five years, it's going to accelerate noticeably.
1:56:10. If you want to wrap up at two, you might have time for one or two more questions.
There's another question. Yep. Thanks for the time check, Peter.
Thanks for doing this. It's Blair Abernethy with Rosenblatt. Just wondering if you can walk us through a picture of your install base today. What sort of versions are they on? How many are very current or within the last, say, two years? Because it looks like Infinity 2026 looks like it's a pretty substantial change. I think hugely beneficial if you guys can get more of your base there faster.
Yeah. Pega Cloud is probably 80+% on the last two releases. It's pretty significant. I would say the other 20 is in the process of getting onto one of those. If you go off of Pega Cloud, they're not as current, but still they would say the shape looks similar. I would say we probably have 60%-80% of our clients are going to be able to get to 26 like if they wanted to right away. I would say the other ones, you're probably looking at realistically probably a year or two. There's an important point on the ones that move slower, and this will probably not surprise you, but it's worth mentioning. Clients that move slower for upgrades are typically much stickier clients.
That's a good and a bad thing. It's a good thing because they're sticky.
It's a bad thing because you can't get them real-time access as governments, for example, people that tend to move a little slower. It's not that clients don't want to move, it's that sometimes they have a lot of process to get through to be able to go. There's some rules. Some companies will not go on a current version. They'll only be one version back. That's typically more like high security, high levels of third-party security. I think Lucky had a question, too. Yeah.
Great. Lucky Schreiner with D.A. Davidson. Thanks for taking my questions and great presentation. A philosophical question. You emphasize the high cost of the frontier models and how Pega lets customers drive value out of that at a reasonable price, which makes a lot of sense. Can you maybe level set for investors how you view the relationship with the model providers moving forward, particularly as you talk about using them to drive internal efficiencies and the model improvements help customers get more value out of their Pega applications. I think the goal is to clarify frontier models aren't going to eat all the software, but maybe is it a case of classic coopetition moving forward? What's the relationship there?
That's a great question, and I will draw a parallel to, you shouldn't use AI for things like deterministic work. We've talked about that. We strongly believe that. You also should not use frontier models when they are not needed. Frontier models are incredibly expensive, 10, 20, 30 times as expensive as maybe some of the more older models that might be able to be just as helpful to be able to solve certain use cases. I think the best-in-class companies are going to think about what am I trying to do? Should I use AI? Should I not? Where do I use AI? Then which model do I use? Do I always want to be on the current model? That's a silly answer. Anybody that says that is going to change quickly because the ROI is not going to be worth it.
I think people are going to think about the models. We actually look at each model. Although they are commoditizing, they still do have differences. We have model optionality. In Blueprint, we use four different models, depending on different activities and different things that you're doing. I think my view, and this is somewhat philosophical, but I do think grounded in feedback we've heard from clients, that I think people are getting smarter around tokens, smarter around the models that they use, where to use AI, and there's a whole ecosystem of companies that have been built to manage AI spend, right? I think we are in a mode where clients are going to leverage AI, but they're also going to be really thoughtful about what they're getting for it. I use two final points on this.
One is, the perfect example of the hallucination that can happen, which is well intended. I was at the William Blair conference last week, and apologize if some of you have heard this conversation. My fireside chat was at 12:30 P.M. At 9:50 A.M., Peter Welburn opens up his screen, flips it over to me, and there was a TipRanks press release that said, "Pega stock drops based on negative commentary made by Ken Stillwell in his fireside chat at William Blair." I hadn't presented yet. What AI did was it saw the stock price, it knew I was presenting, and it drew a logical conclusion that that must have been why the stock price dropped. It's completely AI. We notified TipRanks. They took it down right away, and we understand that is a very common thing that's happening. That is not really a hallucination.
That's just AI doing a probabilistic association, they just got it wrong. Imagine if they did that for loan originations, for credit card increases, for payment processes, for ACH exceptions, for debit card replacement. Just think about all the use cases. We would be a disaster of a society if that's actually allowed. Imagine, though, if that TipRanks one went out one minute after my presentation. I wouldn't have really had any way to dispute that. It actually said, based on investor sentiment, they didn't like what Mr. Stillwell had to say. Thankfully, it happened before, so I could actually refute it. This is a problem. If AI is just going to go out there and just make things up because they think probabilistically it sounds right, you cannot run scale transactions at companies through that.
I think that's the most important anchor of deterministic versus probabilistic. When you get into probabilistic, it's like use the right model, just be thoughtful about it. I think clients are trying to figure that out.
Just one point of clarification, Ken, that was a news story that was generated. You said press release.
Oh, I'm sorry.
Yeah.
Sorry. Thank you for the clarification.
Certainly, we can keep going answering questions, but I also wanted to point out for the people that are in the room.
Thank you.
The Innovation Hub is open today. I've highlighted on the screen what I consider to be the top five things for you to visit. If you guys want to take a quick snapshot of that with your phones, if you're going to go down to the Innovation Hub, I would go to those areas and focus on those. We can certainly answer a few more -
I actually have to leave because I have to go meet with the non-sell side analysts. Our industry analysts I have to talk to them. Thank you, everyone, for coming. For those of you that are staying, enjoy the Innovation Hub. We actually have a concert this evening. Please partake if you're around. Naturally, we're here if you guys have any questions and to reach out. We're still in our open window for another 10-15 days or so. Thanks, everyone.