Pegasystems Inc. (PEGA) Earnings Call Transcript & Summary

Great. Hello, everyone. My name is Alexei Gogolev, and welcome to JPMorgan Boston CMC Conference. Today, we're delighted to be hosting Pegasystems management team. First of all, Alan Trefler, welcome Founder and CEO of the company; as well as Ken Stillwell, company CFO. Alan Happy to have you here. And first of all, if we could maybe begin with a short overview of the business, what investors were new to Pega story. Maybe talk about what Pega does and provide maybe a few use cases to better understand the business.

Well, I'll let Ken kick that off. If you talk to the investors all the time.

Okay. All right. So Pega has been around quite a while helping traditionally or historically enterprises with scale transactions. So that is exclusively companies that work in the B2C industry, but there's a lot of connection there between B2C business models and a significant amount of of volume of work that needs to be automated. And that volume of work also tends to be very structured, deterministic workflows is the that's more commonly used these days. We've always thought about the power of Pega being to be able to build something once and run it millions of times the same and predictably and then allow the change to be manageable in terms of the evolution and how we innovate and how we change the -- either the the work process or the workflow because things do evolve. We've been on a journey over the last 10 years or so where we had historically a user-based perpetual license model in earlier part of our business model. And that has dramatically changed over the last 5 to 10 years, where now we have more SaaS or Pega Cloud business were all recurring. And we have a volume based or a usage-based metric as our primary licensing metric. Sometimes we have users and cases, case is a unit of measure that we think of at Pega. So the businesses went through a pretty significant transformation over the last few years, we went from a business that was $500 million and largely perpetual to a business that's approaching $2 million and generating a significant amount of $2 billion -- excuse me, $2 billion and generating a significant amount of free cash flow as we went through that transition. And more recently, over the last couple of years, we've -- we launched something called Pega Blueprint, which if you haven't seen it, you should go to pega.com and take a look at it, you can test it for yourself. It's really our way of leveraging AI and how we can help our clients reimagine or envision the future of what they want to transform in terms of their technology platforms.

And maybe just to touch on AI a little because it's such a wild and crazy part of Lexicon these days at all moments. We've been heavily involved in it with AI since 2010. In 2010, we went out and acquired a company that were specialists in statistical AI. That's machine learning. It's the part of AI, which I think are people are overlooking now some, but is still incredibly valuable. And the reason we did that is we had since our inception been experts in rules engines and process automation. How do you have processes that makes sense according to the various either legislative or business policy or other types of rules. And you could really see that being able to do machine learning off of that and pull that sort of knowledge and learning into would make a lot of sense. Of course, since 2022, we've done massive changes candidly to our business model as well as our products, which are reflected in this Blueprint AI technology, which we're really excited about.

Well, it's a great segue. Maybe we could discuss, Alan. So customers, they either buy, build or they configure solutions. Where does your local Pega solution fit into that landscape? And you talked about Gen AI, how does that change customers' decisions?

Well, I think Gene has massive implications, and I hope not to contribute to the crazy hype that's going on around the agentics and AI these days. But we're pretty neck deep in it and it's having a pretty substantial shift in our business. If you think about one of our customers, they typically have hundreds, thousands of workflows that run and describe their business. Their -- think of that as being the way they want to have standard operating procedure -- and if they have to pass an ISO regulation or they have to pass something, you actually have to even document those and show those people sometimes. And that's candidly a good thing. What we've been able to do with the Blueprint AI technology is use the full power of these frontier models to be able to apply AI in anger as it were. At the time that a customer is reenvisioning or reimagining how they want to do part of their business. This might be moving and older system to a newer environment. This might be taking 6 or 7 systems that come together as a result of a merger or acquisition. And binding them together. And what our Blueprint AI does is it lets you do that and lets you rethink it in ways that leverage are literally 4 decades of best practices. We know a lot about workflows. We've been able to incorporate those in the language model technology that is part of our Blueprint AI that really can bring that to our customers when they want to think about a new way to either engage with their customers or engage in their back office or do those sorts of operations. So what it really is, is instead of the traditional model, where people would either buy an off-the-shelf software product, figure out how to wire it up to our various back ends figure out how to hook it into how their users or their customers might use it. This lets you instead literally create something that's yours, that is specific to your back-end systems as a customer and specific to how you want to engage it. That's fully agentic in architecture, which means that there are things that can be automated, they just naturally automated. But by using all this tremendous AI power at design time. And then at run time when people are actually using it, being very selective about the use of AI. We can do a couple of really interesting and unique things. One, the systems that creates a remarkably, remarkably better and really allow customers to do -- you can take an old system and convert it and you get something that doesn't look like the old system. It looks like something that is the way you would want it to work. But more interestingly, by applying AI at run time selectively, we don't actually charge our customers for tokens. They're all this stuff that's finally caught up in the last 4 weeks where people are suddenly worried about token expense. Of course, they should be worried about token expense, $1.5 trillion is being spent on data centers and somebody's got to pay for it. We are so, I would say, smart in the way that this uses the AI and uses tokens that we really get the best of all the worlds. And I think this is going to be pretty exciting as the confusion abates in this market, it's a lot of confusion. But as it debates, -- and it will. I think the companies that have done the right things structurally are going to be the ones we're going to be able to take the lead for themselves and for their customers. .

Makes sense, Alan. And could we maybe talk more about Pega's competitors? And how is it changing? Obviously, all these hyperscalers and SaaS platforms, they're expanding their agent strategies? How has this changed recently?

Well, we're very involved with both AWS and with the Google GCP lines because we run we don't maintain our own data centers anywhere. When somebody uses Pega Cloud, which the majority of our business now is and the significant majority of any new business runs on Pega Cloud, it's actually something that hyperscalers are really quite happy about. And we have -- for example, AWS is going to be with us at PegaWorld. And I think talking about how they were -- in doing things. They have agents that actually know how to translate legacy systems, and we integrate with those so that we're in a position where, for example, AWS transform which is an agenda capability they have to read like old COBOL code. We will take that as one of our inputs when we are reimagining how one of these systems could work. Of course, we take lots of other things. as inputs, too, will take user manuals, will take what's on the customer's website, we'll take all sorts -- so pretty much anything you throw at it, Blueprint will digest as part of making a new solution. So I think we're pretty well aligned with the hyperscalers there. The noise that's in the market where people worry that just all software is dead. I think rumors of the death of software are greatly exaggerated, particularly if you pay attention to the different types of software that are out there. There's just a lot of -- certainly, there are companies that are dead, but there's a lot of different types of software. I think we're ultimately positioned to give the customers the ability to manage the workflows they need to manage to run their business, but to do it in a way that is both innovative and deterministic. And I think that's a big advantage over a lot of the other things I see out there.

And I think I'd just add one thing. I think leaving say a genetic engineering or code writing out and you think about the competitive landscape, I think there's a pretty deep misunderstanding around companies that don't really do workflow, but essentially custom build process connections to be able to try to replicate that activity that don't really have the structure, very hard to manage change -- it's very hard to get repeatability. It's really just another version of custom code. And then there are workflow providers that are purpose-built for really straightforward, simple use cases like a ticket management system. And I think that there -- I think many times, all of these companies get thrown into the same bucket in terms of workflow. And what we pride ourselves is that our clients get value in building enterprise scale. When I say enterprise, I mean repeatability at scale, millions, hundreds of millions of repeatability things that happen that need to be done in a very deterministic way. And I think lots of companies say that they will help that workflow, but they really are just another version of custom code or more simple use cases.

And deterministic does not mean not varying. I mean it's really important, and this is where the AI helps a lot and making it so you can have decisions and subtlety and other capabilities that will do the right thing for every customer, but they'll do it in a way that first, you don't reimagine it literally every time a customer shows up and they'll do it in a way, but it ensures that you're treating to customers the same way, which in lots of businesses is considered a good thing as opposed to reinventing something for each one of them.

That's a very important point, Alan. So with generative AI disruption point solutions and kind of obviously, low-end workflow companies how does Pega architecture and product suite position you against very big competitors like Salesforce?

Well, it's interesting. I think Salesforce, Microsoft, ServiceNow they've all done something, I think, is wonderful. They've all created something called a prompt studio, and they want people to go in and create agents, which are pretty easy to create, though, whether to do one that really does exactly what you want to do is so easy, not so much. But they create agents by putting English language prompts in -- and you listen to -- ServiceNow is a great company. You listened to Bill McDermott, talk about he's going to have an AI control tower to let the thousands or tens of thousands of prompt driven agents that you create to magically be able to operate, find each other, call each other and do something wonderful. I personally think that approach, which is the general approach for some of those other companies I mentioned. And candidly, is the preferred approach for people like Cloud and Open AI because it generates staggering numbers of tokens to have this happen, that preferred approach, I think, is madness. So in Pega, if you want to have an agetic process, you create a workflow -- and that workflow is doable by a person only, we'll go to the person. If a step is doable by a person, and we'll go to that person. But our super agent is able to read any workflow in any Pegasystem and execute it. So what do you get? You get the power of the AI creativity, but a design time. At run time, you're using the very narrowly for language translation and for selecting the correct workflow, but you're executing a workflow, well, you'd actually tell somebody what it was going to do before we did it. And we happen to think it's not candidly whether we're 10% ahead or 20% ahead or whatever, we are structurally doing this the right way. And I think that, that will over time come out.

It's interesting. There are examples of their analogies to this workflow discussion that Alan is having that we would never pause to think that doing it using agents or AI. For example, imagine if you just scrapped your ERP system and what you said was I'm just going to ask an agent what the GL transaction should be, every single time. I mean it's -- we would look at that and think like there's no possible way you wouldn't even -- how would you even audit that? What would you get as a result? How would you have so there are situations that are very parallel to this discussion that Alan's have having a round, I need to make sure that when I do a dispute on a banking transaction that, that follows a very deterministic process. largely because that process may be regulated. And there may be some variation, as Alan said, because something unique comes up or there's an exception or there will be variation with a human interacting with that workflow. That's where AI is really powerful to augment the workflow. But I do think the concept is is very well understood in other kind of analogous use cases. The workflow is done because you need to repeat it at scale.

And the thing that I think people miss because sometimes they think of workflow systems as being just kind of ticket tracking systems. When you build a workflow right and particularly when you use like Blueprint AI to design your workflows, it can put a lot of discrimination, a lot of selectivity into the branches of those workflows. So it can -- no different types of customers, no different types of risk profiles, no different types of steps, but it can show them to you. It's not that it's figuring it out every single time you go through. And I find that, and I think customers will find that increasingly comforting as they hear people, Julie this morning was talking about an AI control plane, I tell you, I think we're going to go through a phase where this stuff has to go out of control before people realize that it should be brought back into -- well, it just should never been done that way. I mean, there are other ways to do it. And so we represent the alternative way to do it.

Thanks a sense, Alan. And so as enterprises move from experimentation to ROI-driven implementations, can you talk about some of the themes you just highlighted the AI governance and explain the ability for your customers this approach that you have towards this strategy?

I think it's a difficult time for customers because the hype cycle has been in full gear for a while. There are lots of things being said the whole declaration of war with the entire software industry, where the AI models have basically said the entire software target addressable market really belongs to them. I think customers are trying to figure out some of this stuff. And we're moving into operationalizing more -- but the company is still trying to figure out the key elements of the architecture, I think, and I don't know how much longer that will go on for, but it's going to go on for a bit. The reality is the world is moving very, very fast. These things are coming at them very, very fast, and they are contradictory. And so this is an opportunity, I think, for people to make decisions that will be very consequential either good or bad. And I think some of them are being a little cautious because of that. Having said that, a month ago, we were token maxing, right? You guys know token maxing, burn as many tokens as you can gets and why and gets up and says, you're not spending $0.25 million a year per engineer on tokens, you're just wasting your life. And I think people have reconsidered that. And I think that reconsideration is going to come fast and it's perfect because candidly, we figured out in 2023, when we started that this free launch was not going to exist at some point in time. And I'm really pleased that I think between now and the end of the year, we'll be out of free lunch territory. People will realize that somebody was expecting to pay for all these data centers. .

I think it's -- to add on to what Alan was saying in the last -- inside of a month, I would say, in the last 2 to 4 weeks, almost every client conversation that I'm in, has some part of the conversation that says, hold on how many tokens are you going to charge me for, right? So I think the clients are becoming much more sensitive because of their own experience, naturally, there are the very highly publicized like I ran out of my took budget by February, you will see more and more of those use cases but clients themselves are starting to understand there's variable pricing models. I think that everyone kind of knew. I think we all knew in our stomach that like it was going to come. It just came really fast, right? I mean like the fact that the realization of like -- this is a very costly way to do repeat scalable transactions, it's not efficient at all. It is not the right way. It's the reason why we went away from writing custom code 40 years ago because it's incredibly hard to manage. By the way, right in the code is like 20% of the problem, right? The 80% of the problem is operating it, scaling it, changing it. And that problem comes much harder with the proliferation. That's why that's why we exited. That's why Pega, we existed to try to give our clients an alternative to that. So I do think the token thing is not something that's going to go, well, I just think the conversations are going to continue to escalate around -- how do we use AI in the right way. That's not a point to say don't use AI. It's a point to say, use it like anything else when it is appropriate, not use it just for everything indiscriminately.