Pegasystems Inc. ($PEGA)

Good morning, and thank you all for joining the Pegasystems session at our Growth Stock Conference. I'm Pat McIlwee, and I'm a research analyst in the software group at William Blair as a part of which I cover Pega. I'm required to inform you that a complete list of disclosures and potential conflicts of interest are available at our website at williamblair.com. Today, we're thrilled to have the Pega team back at our conference, including COO and COO, Ken Stillwell; as well as Peter Welburn, who leads the IR team here in the audience. So welcome to Chicago. Ken, you mentioned to me that Pega hasn't done a fireside or a roadshow in Chicago in a few years' time. So it's great to have you here. And I think especially timely given that we're just ahead of your investor session in Pega's annual user conference next week. So for investors here who are not familiar with Pega, can you just start us off by giving them an overview of the company's solutions and why the Pega platform is particularly interesting today?

Sure. And thanks for having us. It's actually been a while since I've been in Chicago in general. So it's good to get back. That's true. Good is a good month. So if you think about in large organizations, they have -- and when I think of large, I think of like banks, insurance companies, health care companies, governments, where you have either a B2C business model or you're supporting constituent management like in the public sector, there's a number of use cases to support those consumers, those customers, those constituents, things like health care claims, managing credit card approvals, loan originations, onboarding, change of agile. There's just a series of actions that need to be really structured and managed consistently, sometimes because it's regulated, like managing a credit card dispute and how Visa or Mastercard requires the banks to manage those disputes or other things that might be more driven because it's the internal control processes of the organization to do in a certain way. So when you have those deterministic workflows, work that needs to be done exactly the same way through a series of steps and stages, and be able to know on the front end and know on the back end that you actually executed that work. That's typically called enterprise workflow. Pega is the leader in enterprise workflow. And so sometimes the solutions are more horizontal. They look like onboarding that might be very similar across different verticals. Sometimes they're very specific to an industry or a vertical like know your customer in banking. And so we've been helping clients for more than 40 years doing essentially an alternative to either writing their own application or trying to buy a commercial off-the-shelf solution and try to make it good enough to meet the -- so we've kind of functioned as this low-code platform where you don't have to write code, but you can get the level of specificity and configuration that you need for your use cases.

Okay. Okay. That's great. And with more -- to kind of build on that, right? So more than 3/4 of your revenue comes from highly regulated industries like you touched on. Can you talk about why those customers rely on Pega? Is it that trust that you've built over 40-plus years, the security, the services component? What's the secret sauce?

So we tend to sit in this kind of convergence of a number of different factors. One, the scale and the volume of transactions. Not many systems are able to manage what could be billions of interactions in the course of a year in that scale and multiple kind of instantaneously having multiple threads of transactions. So there's one is like a scale differentiation. The other one is that what I talked about, about being highly configurable. Now highly configurable does not translate into customized, although some clients do like to build customization around Pega. It's really just around the ability to configure a set of work and to be able to iterate and change the nature of that work over time. That's a -- there's very few vendors that actually have enterprise workflows that allow you to do that. Another dimension of that is the level of security that we have. And security, meaning not just native to the platform, but we have over 100 different industry certifications, everything from PCI to HIPAA to FedRAMP to IRAP to ISO 27001. So in any country, in any vertical, we help our clients support their third-party certification requirements that sometimes they are because they're regulated. Other times, it's just the nature of the business that they're in. And another dimension is the ability to really drive the very robust structure of the work and be able to separate the work. We have something called a case, which is not only do we have the actual workflow, but we have the ability to contextualize each incident or each activity in a way that is unique, but also related to other incidents that look like that. Many of our competitors manage that just in a database, right? They do indexing and they have rows and columns, and it really prevents the ability to understand deeper associations and relationships around the metadata that is associated with each individual transaction. So people typically buy us for the combination of all of that. And that really -- everything that I just said really fits tightly with enterprise needs. Large companies have scale transactions, regulatory matters, consistency, repeatability and the ability to manage as the business changes.

Okay. Yes, that's great. And so in your overview, you said the word deterministic. So I'd like to ask you to kind of elaborate on that because I think it's still not completely understood in the investor community what exactly Pega does, right? It's providing secure, reliable, deterministic workflows that work for an enterprise every time. And then what foundational models like Claude do, providing more probabilistic calculations, where there's overlap, where there's a distinction and kind of where there's a harmony between the two?

Sure. So I'll start by saying there are 2 types of work or 2 types of -- I'll use the word workflow, but 2 types of processes. One type is probabilistic, generative, where you would expect there to be an error rate. The error rate might -- you might want it to be 1% or 0.5% or 10%, but you would expect it to not execute exactly the same every time because it's going to use the data that it has. And even when given all the exact same variables, there is a slight risk that the probabilistic model will pick -- if left with 2 equal choices, may pick one one time and one another. And that's just the way the models are built. There's nothing incorrect or errodous about that they're built to be inaccurate. They're built to try to get it close enough. And then there's deterministic, which is really not focused primarily on the outcome. It's focused on the process. How is it that you will go through the work? Good example of a probabilistic action or a deterministic action. Probabilistic would be if you come to a website of a large credit card company and they want to speculate exactly what rendering of a picture or a call to action or an offer they give you, that would be probabilistic. They're not going to get it 100% right. They might see that you're coming in from the Midwest and that you're over 50 years old and you have a family, so they might show a picture of a family sitting under a tree in a corn field because they feel like that might be the most relevant. They might have that completely wrong because you're from France and just happened to move to the Midwest and actually that picture doesn't resonate at all with you. But that's not a problem. That just means they got it slightly wrong. And then there's a probabilistic workflow. I'm going to approve a loan. And I have to follow all the state guidelines, discriminatory lending guidelines, credit guidelines, wholesale lenders, whether it's FHA or VA, very, very structured, and you cannot get it wrong. Might you -- might an underwriter make a decision at a stage in that workflow that could be wrong? Yes. That's where the judgment fits. That's where the probabilistic fits with the deterministic. But the structure of how you process that loan must be the same every time.

And really Sorry, correct... Deterministic. That is deterministic. Sorry, I want to be misleading. Or that...

Probabistic. Yes, sorry. Probabilistic is where you can have an error rate and you're making a best guess. Deterministic is where you decide the work that is done on the front end. But in a deterministic workflow, you will have probabilistic AI that's used as well. The example would be that underwriting decision. In the underwriting step in that workflow, there may be some judgment. You might look at loan to value, you might look at things in there. Someone might have to make a human call on that. You could make an AI call on that. And you realize that maybe you give a loan to someone that you slightly -- maybe you shouldn't have because they -- maybe there was a credit risk you couldn't anticipate, but that doesn't undermine the process of which you went through to approve that loan. There wasn't an intentional discrimination against the borrower. So in consumer industries, there is a -- as you might imagine, there is a high bias to protect the consumer. And then if you go across the world in different countries, there's an increased bias around protecting GDPR in Europe, protecting like sovereign information, not sharing information out there. So there's so many rules and regulations across consumer industries that it's very important that you understand where does a workflow need to be deterministic and where can a workflow actually be probabilistic. And when it's probabilistic, I think that's where AI can play a role.

Okay. Yes, very helpful. I think it's an important distinction to call out. So just given this is a generalist conference by nature, can you touch on the pricing model? I think that's been a big concern across software, the software sector at large recently, and Pega's pricing model is a little unique. So can you just kind of clarify?

So I'm going to -- maybe I'm going to go back about 10 or 15 years because we changed our pricing model with -- had no relation to AI or any of the things that are going on now. So what Pega does is it takes what otherwise were human activities that would be managed manually across maybe a structured set of workflow steps, and we automated that into a system. When we automated that into a system, what we would do with our clients is we would build efficiency so that if they had 10 people that might have been needed to manage a certain body of work, they might only need 5. In a licensing model that's a user model, that's kind of counterintuitive that we would go out, help a client take their headcount down by 50%, and we would then get 50% of the revenue associated with that. So we did have a licensing model that was a user-based model 23 years ago, and that -- some of our contracts still do have licensing components that are user-based. But we made a big shift to move to what we call a case. A case is a unit of measure in Pega. Think of a case as a piece of work. So we license based on the pieces of work. A piece of work could be a dispute on a credit card, a loan origination, the number of clients that are onboarded, different measures, the number of card replacements that you might have for lost credit cards. So that unit of measure of the case is how we license. And we feel like the more that the system automates work, the more cases that it does, the more that Pega should receive compensation because we're automating and driving efficiency. So we have -- so you might call that a usage model. So essentially, a case is a piece of the use of the technology. In an AI world, that become -- that's become now much more obvious. But we have 75-plus percent of our contracts that are exclusively case-based. And the ones that aren't -- that have users they're typically on purpose clause driven, like they're like -- you could use it for this purpose and this number of users and cases. We typically have both. So we've -- we were ahead of that challenge, but not because we saw AI coming. It was more around just our value proposition, made more sense to charge based on usage. The analogy I use is if you were using AWS for your cloud, you would -- AWS would never charge you based on the number of employees you have. They would charge you based on like the number of CPUs, storage, processing, et cetera. And that's very analogous to Pega.

Okay. Very clear. And so I think whether or not it's completely misled, there's a fear of disruption associated with some of this automation technology right now. But when I've spoken to customers of you and your peers alike, it seems like they're leaning more into these trusted platforms more than they're trying to move away from them, right? Can you talk about just how those customer conversations look when you're talking to enterprise-grade customers that are looking to roll out AI automation at scale?

So a few -- maybe a few thoughts on that. So one is when this -- when AI really started to get more visibility maybe in the like kind of November, December, maybe even January time period, there was a lot of confusion even with our customers. To be honest with you, I think there's still quite a bit of confusion with investors. But there was confusion with our customers around this concept of deterministic versus probabilistic work. So originally, it was -- there's this AI thing and what can AI do and we should try to experiment and see. I think that carried into the investors thinking, well, why couldn't AI just get rid of all of these SaaS companies, all of the -- all this technology. And I think that was kind of the first wave, which I think has largely been settled down now. It certainly with customers, where I think they're not confused. I mean they know that 80% of the applications that they have are deterministic and they're not going to use an agent to go execute that work. But there's probably 20% or so that you probably don't even need a software application and an agent or some type of a prompt could actually execute what you need to. So I think they're honing in on that. The next step of that was, well, okay, so I'm not going to displace it. I still need a software product but could I just write my own? Could I actually use now, then the model started to say, well, we can help you write code. And then there were tools like cursor like that would help you be like kind of almost like a development harness to be able to help you drive using the models to write code, which we do at Pega, which we've been -- we're not fully rolled out on that, but we're in that journey as well. Then that kind of takes you back to, well, why do you -- why would you want to write your own application? In some cases, you write your own application because there isn't something you can buy, to be honest with you. I mean it's just your use case is unique enough or you write your own application because what you could buy isn't quite the perfect fit or it's just too costly to buy versus you just actually writing it yourself. So I think there's definitely going to be applications where the companies decide, I can actually build my own, and it's going to be faster, cheaper, easier to support. Then you get into the bucket of why would they try. And many of our clients, we -- I've had this conversation over and over again where there's a couple of dimensions. One, error rate is one. So when you have a situation where you cannot have error rate, there's no such thing as like I'll accept a 1% error, where the system that you're building is highly complex. It's going to manage a lot of scale and may be subject to regulatory or control processes. You run into that risk of, is it better for me to build my own ERP system? or should I buy an ERP system that actually I know is hardened to be able to support all those controls. I think when I say that example, most investors would even say, yes, that's kind of ridiculous if someone would go try to build their own ERP system. But there are a lot of enterprise systems that have the same level of sophistication and discipline that you would see in some of the ERP modules. So I think that's kind of -- that's one of these decisions that companies will make. One of the biggest challenges that our clients are seeing with the AI models is -- and I'll get it to the last one, which is cost. But the middle one is the level of imperfection that you get. For example, I'll give you an example. This morning, I was finishing up one of my investor meetings and Peter, our Investor Relations Vice President, pulls up a screen. And the article was from Tip ranks, which is basically like -- essentially, it's Benzing Tip rank. And the title said, Pegasystems stock retreats based on comments made by COO and CFO. So we read the article, and it said, Ken Stillwell made comments at the William Blair fireside chat that caused the stock to go down. That was 3 hours before I'm sitting here. that was an actual article that went out. Now we called them and they took the article down, but this happens all the time, right? This is called AI slot, right? It's just -- it's out there. Nobody knows if it's right, nobody knows it's real. Enterprise company, I mean, that's just funny that, that happens today, but it happens all the time. We have to constantly be watching because the information that gets out. Now you're an enterprise company, you're Bank of America, you're William Blair. How important is it that you don't actually let AI decide that when you're trying to make a bill pay on your bank account that it decides you paid that vendor too much, so pay a different one. Like how do you catch that? How do you control that? So these are the types of decisions that companies are making like do I really want to go try to build my own? And by the way, for those of you that are not aware, you should ask around on this. AI models now build code that is 50x faster than the human's ability to review the code, which means we have no idea what it's writing. When we write it at Pega, we have no idea. So what you have to do is you have to throttle how much you can. You have to look at the code, run other models to test it, run test models. And hopefully, you actually can know what the -- that never happened before. In the whole world of coding, you never had a situation where one person could write code faster than someone could actually review it. So these are big, really big challenges that our clients are trying to figure out.

Okay. And so to shift gears to kind of the upshot of AI and how you're leveraging it within the platform. So Blueprint, it has been kind of revolutionary for you guys. It's been incredible to see how that tool has helped your go-to-market motion, taking your sales cycles down materially. Can you just talk the audience through what that has meant for you and what that is?

So prior to AI for Pega, if we wanted to work with the client, we typically had to go through a very manual and quite frankly, very human-intensive discovery session on the front end. That typically involved whiteboard sessions, operational walk-throughs, lots of collaboration, trying to get people physically together and then, quite frankly, realizing that, that took cycles to be able to really figure out like what do we want the reenvisioning of an application to be. We're trying to move something or trying to build something new, there was almost like a village that would have to build the like view of the -- and unfortunately, that could take quite a bit of time. So what that meant was slower ramp for salespeople, harder to get pipeline deals in, early-stage pipe moved slower. So these were all challenges that, quite frankly, we just accepted as part of our business for decades. What blueprint -- what Pega Blueprint is, is what we did was we took the AI models, and we built on top of it all the knowledge of Pega, specific knowledge of Pega, all the workflow history, how does the workflow work? What are the use cases? What are the personas, what are the typical integration points? So in an actual like agentic interface, you could chat with this application and build your workflow. And now the workflow that you build in that is not necessarily going to be one that you click a button and go right into production because these are enterprise companies, but it got you so far -- it gets you so far down the path compared to what we had to do in kind of before AI. So it's been a massive revolution for us in terms of how fast you can get from concept to a design where you're actually looking at the application. At the end of blueprint, you can click preview and it shows you a working application. What we're announcing I guess we've already kind of leaked this out. But we're talking about a PegaWorld next week, next week is our user conference. We're going to talk about the next phase of that, which is the Blueprint experience goes into finishing the build of the application, something we're calling Infinity Studio, which is essentially keeping that whole Agentic experience until the point where you can actually go live and into production. We know right now that, that has taken 50% of the actual time and engineering effort to just get to the point where you could decide what you're going to build. What we want to really do is make this as agentic and as automated as we can to get to application to go live. So that's kind of our -- that's what Blueprint has done, and that's how we're extending Blueprint into the build phase.

Okay. Yes, that's great. And then can you just talk about -- so a lot of large enterprises are still running mission-critical on these legacy applications. Can you talk about what the implications of this technology are in terms of your ability to go and address that opportunity in the enterprise?

So I don't know what the percentage is, but I've heard different percentages, anywhere as low as 10% and as high as 25%, which is the percentage of applications that have actually been modernized in large enterprises. So I've heard Amazon talks about between 8% and 10% of applications have been modernized. I've seen more aggressive ones in the 20% to 25%. Whatever you believe the number is, it certainly is nowhere near 50%, and there is a lot of work to do in terms of getting these typically like homegrown systems running on ancient infrastructure into a more modern world, whether that be on cloud, public cloud like Pega Cloud or whether that be managed on a virtual private cloud. What Blueprint does for us, which is just massive, is it allows us to go after new logos and new workflows in a much more aggressive way because the upfront selling process, the upfront solution process is so much faster. If you think about in the previous world before Blueprint, if Pega wanted -- if we wanted to target a new organization, the first thing we had to do was hire a salesperson that would target that. The next thing we did was train the salesperson for 3 to 8 months to get them certified on Pega, then they would start calling the company. But remember, the whiteboarding session example, that might be a 6- to 9-month pipeline building. So we had salespeople that we would hire and they might not build pipe until their second year working there. And that's if they follow the path. Now we can hire a salesperson, they don't need to be certified on Pega. All they need to know is how to get to pega.com/Blueprint. That's the extent of what they need to know. Blueprint is right there. They can engage with the client in a first meeting -- the other thing is with new logos, if you think about a company that knows Pega, like Bank of America, I'll use that example because they're a large, many decade client of ours. We don't go into Bank of America and say, let me tell you what Pega does. They already know. If we go into a brand-new client, they don't know what Pega does. So we're going to go into that brand-new client. Blueprint is an easy way to show it and say, let's walk through one of your problems, onboarding a client, managing a dispute. You pick whatever that vertical might be. It just makes the whole conversation. You're immediately going into a demo that's very specific around the customer use case. And that -- the level of confidence that gives our sales teams, how fast we can ramp our sales teams, how quickly we can attack new logos, these are all brand-new things for us.

And you can correct me if I'm wrong, but you guys have actually quantified your sales cycle might have been 12 months before, on average, it's been cut in half. Like last quarter, you highlighted some deals that went live in 90 days.

We had one that went live in 42 days which is -- may seem like 42 days for not knowing enterprise software, may say, well, that's still 1.5 months. But like, I mean, to take an enterprise application and actually go from whatever they had before into a working application inside of a quarter is nearly unheard of in enterprise. So we've had a handful of those in the past 2 quarters.

Yes, pretty impactful. Yes. Okay. So we can get more into the financials in the breakout. But just one -- so there was some noise in the first quarter on the ACV growth. There was some noise around the license revenue, the renewal timing, a little disruption within your federal pipeline. How should investors be thinking about current ACV growth versus the growth that you expect over the next few quarters?

So when we guided the way we -- so there's a little -- in our business, many -- much of our growth comes on the back of a renewal cycle. So if a client has a renewal event, typically, that's when our ACV is our equivalent of ARR. When our ACV increases, the customer typically makes that commitment based on the usage or systems that went live in the previous year. So we're tied to renewal cycles. In 2025, our renewal cycle was not back-end loaded. In fact, actually, there were more compelling events in the first quarter, last year, meaning 2025. And then in 2026, when we guided, we had exactly the opposite. We have more compelling events in the back end of the year and not as many in the first half of the year. So it creates this dynamic of just difficult compares in the first half of the year and easier compares in the second half of the year. So it makes our growth rate kind of bounce around a little bit because we measure a trailing 12 months. And so that's really what we had talked about. Now separate from that in Q1, we had a few kind of isolated incidents that caused our bookings to be slightly lower than even what we would have modeled. Like we would have modeled about $25 million of net new ACV in the first quarter, and it was about $20 million. So we were at about a $5 million gap. Some of those were some of the government shutdown and changing to the processes that they've had, had some deals slip a little bit. Just these are renewals with expansion. So these are not deals that we have to win. They're just paperwork situations. So we had a couple of those. And so we've had a few situations in Q1 that were for various reasons that caused Q1 to be slightly lower than what we had modeled. But we've -- we had said from the very beginning, and we still feel that way that first half of the year, tough compare, back half of the year, easier compare. So it will cause some growth gyration through the year.

Okay. Got it. So there's some more mechanical factors at play than anything necessarily concerning and you guys still feel pretty good about that mid-teens ACV growth target?

I think probably the only thing that's still -- it does concern me, but I don't know how to quantify because it's not an empirical concern is supply chain disruption from the Middle East. I still don't know how to -- I don't know how to measure that risk. I don't -- it could be nothing. It could manifest itself into something, but I think that's the one that I'm just still kind of not sure how to -- we don't -- we're not in the energy space, but I think I'm just more worried about the macro impacts that could happen. Europe has started. Europe has been under a lot of strain with the Ukraine, but I think with natural resource with having some shipping delays and certainly running low on inventories for oil and gas. Those are some areas I'm watching. That said, the consumer has held up pretty well.

Okay. Great. Yes, I think we're just about of time. So we'll wrap it up there. Thank you, Ken, very much for being here. I appreciate everyone coming in, and the breakout will be in Jenny Be upstairs on the second floor.