Cofounder, Pegasystems
Live from PegaWorld! @pegasystems
PegaWorld 2026: Why Enterprise AI Just Grew Up 🎙️
Live from Las Vegas — Pega makes the case for a more pragmatic kind of AI: predictable outcomes, predictable costs, and no more burning tokens to re-reason the same problem a thousand times. $40B has gone into enterprise AI and most CEOs say they got nothing back — here's the counter-programming.
In this one:
Why "token maxing" is over — and what outcome-based pricing changes
Reimagine, don't bolt on: redesigning the work before you build
Deterministic where it counts, agentic where it helps
Blueprint, the Agentic Process Fabric, and AI that'...
About Alan Trefler
Alan Trefler, founder and CEO of Pegasystems, spoke at the company's PegaWorld 2026 conference in June, where he criticized the current state of enterprise AI. He argued that much of the discussion around AI is "mostly gibberish" and described the industry's approach as one where companies were given free tokens to get "hooked" before being charged. Trefler announced that Pegasystems' Infinity 26 release would not charge customers for token usage, stating that "using Pega AI will not accrue a token charge." He advocated for a "workflow recipe mindset" over what he called the "madness" of having AI "re-reason every single time you do something," and emphasized the importance of predictable outcomes and costs. In a separate interview in May, Trefler discussed his 43-year tenure leading the company he founded at age 26. He described being a public company CEO as "hard" with "a lot of distractions," and said he believes there is "way too much of a short-term focus in America." Trefler characterized his personal wealth as a "number on paper" resulting from his ownership stake, and said he has "one home" and is "not running around." He praised Nvidia CEO Jensen Huang for "doing an amazing job" pivoting the company to take advantage of AI.
Transcript (26 segments)
✨ AI-enhanced transcript with speaker attribution
A
Alan Trefler0:00
I actually can't see most of you here, the lights are sort of in the eyes, but I can hear the sound, I can feel the energy, and I know that you would agree with me that this is an amazing time. And it is wonderful to be joined by these brands, to be able to contemplate where we are, what we need to do, and some of the important choices that have to be made.
You know, I'm told that a lot of these conferences, including many I've been to, really don't delve very deeply into things that are controversial or different ideas. I want to set the challenge for myself with you to completely change that today. To give you a way of thinking about this world that we're at that would influence how you chose to apply and evaluate AI in your organization.
And you know, this is based on our long history of understanding technology change, working with AI, and I would tell you this is an enormously fraught, enormously confusing, and enormously important time.
You know, we've all for the last couple of years been exposed and every one of these organizations have been playing with AI. I said these little tokens by the drug dealers to get us sort of hooked, you know, and that's hey, don't have to pay anything for this token or these tokens, $20 a month, all you want. And all of this creating the sense that there was, because there is, something magical here. And that somehow this magic would convert itself into just a utopia. A utopia that would take so many of our operational, administrative, customer problems, make them better. Really boost organizations in ways that they might not have conceived of before.
And you know, it sounded good. It is magical. I mean, look. Everybody knows we all want to vibe our way to whole new coding systems. Everybody knows we want to create thousands of agents with English language prompts that are going to just figure out how to do all the things we need to do. And hey, if one agent is good, then having 400 agents calling each other, that must be better. Yeah, we'll put a control tower right in the middle of it and then it's a mess.
But you know, I think as we've gone through the last two and a half years, in particular there's some things that developed in literally the last three to four weeks, there's been some concern that maybe there's a little darkening here on the horizon. Maybe we should worry a little bit about this utopia here.
You know, for instance, we discovered that these new models that people have, they actually don't always come up with the same answers. I mean, look, we like them because they were creative and probabilistic and that can be extremely powerful. But they seem to almost form opinions of their own sometimes, do sometimes unusual things leading to unusual headlines. And then every week, or couple weeks at least, there are new models that fly in. And these new model versions that fly in, they change the agents you've been building. They change the outcomes you may have come to depend on. And boy, that can really add to the confusion. And there's more by the minute.
And if I use these agents to go generate code, it turns out that AI can generate massive, massive, massive amounts of code. Not really surprising if you think about what it does and how it is. But, you know, what I always learned when I was growing up as an engineer, it's not about the quality of the code. It's whether you can navigate it and actually understand what it does.
We used to have contests when I was in college in which we would try to write a computer program in the fewest number of lines. I guarantee you that none of the Claude AI went through that education. There's a real desire to be very expensive. And don't get me wrong. We write a lot of code with AI. Everybody should write code with AI. But should you trust AI to build whole comprehensive systems that are somehow going to have to be fit together and going to change? Well, I don't know. I think there's some credible dog hunts there.
And then, four weeks ago, this amazing thing happened. Now, we knew this was going to happen several years ago because we've seen this movie. But a lot of people seemed stunned when suddenly the dealers began wanting you to pay for the tokens. So, all of these $20 a month plans or $100 a month plans suddenly became oh my god no maybe it'll be as many as it'll be. And you know we have the CIO of Airbnb saying he ran out of tokens in his first quarter. And by the way he's just beginning. And so this idea that these things will cost something should not surprise any of us. Because we've been reading about the trillion dollars worth of data centers that are getting built. I mean something's got to pay for these. You know and SoftBank's IPO.
So we're in a really interesting situation where the decisions we're making have a lot of influence. Can we actually let these things run loose or can we find a way to organize the AI, organize the code, and organize the agents so they operate in a harmony? And they'll actually give us the city that we want. But a city that ultimately we need to be the ones to control, to understand, and be able to afford.
And what Pega believes firmly is that if you use Pega to reimagine your AI that you will be able to achieve predictable outcomes. That is the AI will give you the answers you want to give your customers on a consistent basis. And you'll be able to do it with predictable costs. And being able to do that we think is absolutely critical.
Now to do this, you're going to have to change the way you think about applications. Because the word intelligent, which means kind of all things in all places, does have some concepts that we can tease out. And there is going to be a huge change in the way that business applications operate. Historically, business applications have operated as screens sitting in front of databases and going through and humans typically being able to type stuff in and getting stuff posted. And in a world that's intelligent, you're not going to have humans and you certainly don't want to do what we make people do today. We make people sign onto a system and then sign onto an application and then navigate that application to get to a particular point of it. And then having found that, get trained to do something on that page of the application.
You know, the way applications of the future want to work is you want to be able to have the application be a collection of processes, data, security, but have it be able to come to you so that you can tell it what you need and the right application will reach out to give you the right workflow to get things done. And this is the vision that Pega's actually had for a long time around something we call Center Out.
Center Out is the idea that the heart of your applications and the heart of your businesses should be the combination of the decisions you make, the workflows you execute, and the context, the case data you collect, that you can describe, show an auditor, do machine learning from. And it needs to be independent of the channels on the left, whether this is an individual or a customer or an agent, and also independent of the data sources. Because customers never have a data fabric that's as good as they want and it's always changing, particularly if folks merge. But, this idea of applications and building applications that are these collections of workflows that you can study and you can understand gives an organization both power to execute but also that reliability and predictability.
Let me draw an analogy. Let's say you were in Vegas and you wanted to open a restaurant like many of the very fine ones that are here. What would you do? Would you get a space and put the restaurant in? And you wouldn't run out and start hiring a bunch of chefs and putting signs on the wall for people to come in and tell you what they wanted. No, what you would do is you'd create a test kitchen. And you'd exercise how you wanted to create that menu that would be both exciting to your customers and something you could operationalize, you could do, you could build consistently. And then you would build that test kitchen and select the recipes that reflected what it is you wanted to do as a business and how you wanted to build your reputation here. And then you have a basis for being able to scale up how you want that recipe set to work. And don't get me wrong, you can tweak the recipes for an individual patron, but you should have a core architecture that describes the business processes that make your business work, the service catalog, as it were, for your business.
This is what we invented when we used AI to create Pega Customer and the thing about it is this is a terrific use of AI to be able to use AI to build creativity, to challenge, to pull information together from different sources, and really grind it. And this is as hungry for tokens as anything you'll ever see. However, we thought this was so important to our customers that we gave it away for free. Put it on pega.com. Because this allows customers to really rethink what they do. And you know what the secret is? For every blueprint of a recipe, you're going to produce it hundreds, thousands, tens of thousands of times. So, the aggregate cost is really very modest because of the way you do it.
Contrast this to the way that the agentic folks talk about reasoning and wanting to re-reason every single time you do something. We think this is madness. We think that workflow recipe mindset is an architectural mindset for your businesses and may not be 100% of what you do. But let's face it, a tremendous amount of every business has similarity. And you should look there before you go build burn tens of thousands of tokens to make something up from scratch that might not give you a reliable answer.
Now, this blueprint idea has AI built into it, and it's been enormously exciting. But what I found even more exciting is something you'll see in the Innovation Hub today, where we've got eight of our partners in booths showing their branded blueprint, where we've given them a language model vector the database, something called a vector database, where they can put their intellectual property in. We can't see it. But when one of their staff members uses Blueprint in conjunction with their vector database, not only does their name come up on the top of the Blueprint, but they get something that takes what we offer, supplements it with what the customer has added, and supplements it with their best practices. And we think this is hugely exciting, and you should go and see how some of the partners have done amazing things in terms of being able to do this. And I think this really is a great harbinger for the future of how we can engage with our partners to do tremendously better business with our clients collectively as well.
We're also doing something with Blueprint that I'm really excited about in the next release, the 26.1 release. We are taking Blueprint AI, and we are adding it into what we call Infinity Studio. So, historically, people would go through a Blueprint process, they would download into their system, and then they would use our traditional tooling to be able to complete their build. Now, Infinity Studio is available to them also at no cost, that brings AI to the helping and the fixing of their build, and from what I have seen, massively reduces the time and effort needed to finish Pega system, and will work with our customers' existing systems, our customers' 23 systems or 24 systems and 25 systems. When they upgrade to 26, their existing workflows all operate now with the power of AI to improve them, which we think is super exciting.
But, perhaps the most exciting thing I think I have to bring to you today is the picture of what we think that Utopia should look like. We start with the idea of applications. Applications being primarily collections of workflows. And you know, you could come up with lots of ways to trigger these applications. The applications are all open now, but we have created automatically something called an application control agent. So this application control agent is just there. It knows how to read the workflows of that application. It knows how to take an input and find the right workflow and connect to it. And it knows how to do that using the LLM very narrowly. So it doesn't burn off the tokens. It's super efficient. We burned our tokens in book rate. At run time, the tokens here are just semantically fine. You connect the right thing and hook it together. And then if there's a step in that agent where you want to go and do something where AI in the agent makes sense, you can call either a customer agent or a third-party agent or Pega agent for that step to do that particular thing, to do a summarization or to pull fields off of a document. So the AI here is strongly used but tightly controlled. Giving you that predictability and orchestration as well.
But what's even more exciting is customers want to talk to these things conversationally. And we want to be able to do this without having you write prompts. This whole business that people write prompts in English to figure out things, I think it's a lot less reliable than having a workflow. So what we do is we actually have created our own conversational agent out of the box that in any language will let you talk to any application, find its right workflow, execute it, and if you have security, walk you through it, or allow any application a customer might have to communicate through MCP on the front end to be able to find the right workflow and get it done.
And we know many of our customers don't just have one Pega workflow in it. But you have a multiplicity of them. And so we have created the Agentic Process Fabric. The Agentic Process Fabric uses a language model database, a vector database, to pull together and register all the workflows in all the different applications. And by either talking to this Agentic Process Fabric or calling this Agentic Fabric, the entire power of all the workflows is available to you. Without reasoning, without wondering how it's figuring this thing out, without questioning how you could explain it to your audience.
This incredibly powerful design system that leverages AI enormously at design time, selectively using AI at run time, provides the predictability of outcome and cost that we set out to do, and I will challenge you to tell me anybody else who is doing this. Because we think this is markedly different, and that's the feedback that we're getting. But being able to do this has allowed us in releasing Infinity 26, which by the way I'm using in-house and is available now for early adopters. If you reach out, we'll be available for all around the end of the month. Infinity 26 without having to go and write prompts like you would have to do in a Microsoft or a Salesforce or a ServiceNow. Be able to get all of that power with the reliability of workflows.
But something even more exciting for us to announce. Because this is so structured and specialized in the way that it uses AI. We don't have to charge for tokens. So using Pega AI will not accrue a token charge. And I think that that is just marvelous because it allows you to focus on results. On getting things done. Because ultimately that's what you want to do in your business. You don't want to have to worry about these charges. You want to be able to know that you can build a system to accomplish things, get predictable outcomes, get predictable costs, and build for change.
So with that, let me thank you all, invite you to the hub to see this for real, and tell you I think it's going to be a terrific show. Thank you, everyone.