PegaWorld 2026: Pega 101: From AI Innovation to Business Outcomes

All right, we'll kick it off and we'll get started. Hello and welcome. If you're in this room, you already know that AI is everywhere and what you probably already know because you're living it, is that in many cases, the value that AI has promised hasn't yet shown up in your organizations. And that's what Phil and I are here to talk about today. So in the next 45 minutes, we're going to show you how Pega turns AI innovation into real, measurable business outcomes. You'll see it designed, you'll see it built, and you'll see it running. But let me start by introducing the two of us. I'm Anu Shah, global director of specialist Solutions consulting. It's a long title, and I'm joined by Phil Kusnetzky, senior manager in the same practice. And between the two of us, we've spent the last several years working with some of the world's largest regulated businesses, helping them operationalize their AI.

And these organizations are seeing tremendous value with AI working for real in production. But before we go into how they're achieving this, let's talk about why this matters and why it matters now, because I believe that the window for getting AI right is indeed closing fast. The AI and digital leaders are growing 2 to 6 times faster than those organizations that are just waiting to adopt. And the gap isn't closing. In fact, it's widening. Every time you don't, every quarter, you're not figuring this out. Your competition is indeed getting further and further ahead. And it wasn't always this clear. I remember standing on this stage a year ago, and we were talking about how the majority of enterprise AI projects were failing to deliver any return. In fact, last year, MIT had a study they released, and it found that 95% of enterprise AI initiatives had stalled before they reached production.

But as we know and we heard in the keynotes, this has almost flipped overnight. And leaders are starting to figure out what works. So the question isn't if AI delivers value, it's when it delivers value and just how fast you can get there. Because what's coming is, I think, bigger than what most teams are planning for. What we've seen is that with AI, anyone can build anything in days. I've built apps with cloud code in hours, but by 2030, CIOs expect that 0% of IT work will be done by humans. Without any AI, 75% will be humans augmented by AI, and 25% will actually be done by AI alone. And your businesses will increasingly be run end to end by AI agents. Autonomous agents, as predicted, could automate up to 57% of the US work hours today. And both of these are happening, and they're happening right now, and they're happening in parallel, and it's happening whether you're ready or not.

And this is where it gets a little harder. Enterprise AI has challenges that the consumer AI we all use on our phones doesn't have to deal with. On the development side, you have to figure out your legacy systems and that complexity. You have business and IT teams that really need to collaborate, and you have to think of all of the security, the compliance and the performance. And there's also this new need. And I think it was always there, but there's a real need for agility, right? And it's not just shipping an MVP fast or an agent fast. It's really evolving, but evolving fast in production. And on the operations side, as you're executing, you need your consistency across your regions and channels, both for your customers and your employees, and you have regulations and compliance that varies by jurisdiction. You also have to simplify and orchestrate, not just add more tools and more code.

And you have to do all of this with cost and efficiency in mind, because none of these costs can balloon out of control, as we're seeing with tokens right now. And that's the difference between doing your MVP or a chatbot demo and enterprise level AI deployment. You can't just copy that agent in a sandbox into production. There's different pressures and different risks. So the answer has to be different. And we're already seeing, hearing and seeing stories about agents acting without the right sequencing or the right approvals. Like last year, I read this article about a team set up setting up two AI agents to collaborate, and the agents ended up calling each other in an endless loop for 11 days, and the bill was $47,000 and there was no work done at the end of it. And that's really what AI looks like without orchestration and production.

And that's what Pega is here to help you with. Our clients are using Pega. As you heard and AI to really redesign. And that's the key redesign how their operations run and then run them reliably and predictably at scale. And that's a little different than what you're hearing in the market with the pure AI agentic players. And in terms of results, we have four areas of critical transformation and with real results in terms of AI driven personalization, this is leveraging our Customer Decision Hub product. I'm sure you'll hear about it tomorrow in the keynotes. And with that, Wells Fargo generated over $200 million in incremental revenue every year using AI powered decisioning. And this AI powered decisioning became available in our platform several years ago. That was a ten x conversion improvement for Wells in terms of service and sales, leveraging our customer service product.

Virgin media consolidated seven systems into a single system, realizing NPS gains of over 30%. And then Pega is really good at operations and leveraging our core Pega Platform. Citibank went from processing service requests in a day to under three seconds and then Legacy modernization. It's a hot topic right now. Well, Vodafone went live in 40 hours, not 40 days, not 40 weeks, but 40 hours. And much of this was due to that business and IT collaboration and design with Pega. GenAI. Blueprint. So if you think about the areas of transformation, what's common across all of them is that there's speed, there's efficiency. And I think most importantly, there's compliance. And I think you need all three in a successful Platform. And it's not just our clients seeing value with Pega. I think this is the most important quadrant that we have, especially right now. I highlighted it last year, and we were just named the number one orchestration platform in Gartner's Magic Quadrant for Business Orchestration and Automation Technologies.

And this is more commonly referred to as boat. And this was both for the completeness of our vision and our strategy, as well as the actual ability to execute for clients. And but what's common is the ability to orchestrate and orchestrate with control. And I think that's actually really, really important as you're developing agents is use your agents, but have some control and structure in place. And that's the foundation of everything is the orchestration. But the foundation starts with getting the architecture right. So Phil will walk you through a deeper dive into the architecture. Thank you very much. Anu. Uh, yes. And as you mentioned, I'm going to take you through something that most vendors probably skip in a, in a one on one. And that's architecture. Um, but I'm going to keep it very light. You saw a bit of this already on the keynote stage with Alan introducing Center-out and things like that.

Uh, but why cover architecture? Well, simply put, get it wrong and AI just doesn't scale at an enterprise level. But what's the problem that we're trying to fix here? Well, if we look at traditional apps, typically you have your business rules and your business logic all embedded in these channels and embedded in these back end systems. Okay. And what that means, that's okay. Before the era of AI. But in this era that we're in now, and with AI changing and changing fast models change, agents change, channels change. You now have your logic living in four different places. Now Center-out, as you'll see, as you heard Alan say, is a little bit different here. Your logic lives in the center. Your AI decisions, your Workflow Automation, your case management. And then it flows left to all your channels, whether those are customer facing channels, employee facing channels, agents.

And it can connect right to any and all of your data and systems. But if we put this into context, maybe of a real example, let's say a refund policy, we want to go and change some of the rules around that. In a traditional app, what did we have to do? Well, we go and update the CRM system. We go and update call scripts. We're going to update some mobile apps. We're going to update some backend systems. That is for changes that you've just made. Four ways to get things wrong in a Center-out world. You make that change once. And once you make that change, every channel, every agent, every system automatically inherits that change. So you have one source of truth. And at the bottom there you see this great Marketing term, the Situational Layer Cake, right? This is where you can have the same logic live in the center, but now they can behave differently depending on region, depending on product or business line, without duplicating any processes.

So you could have your Germany office and maybe your Singapore office on different rules, but they share the same core processes, right? And this is what we call application architecture is for the agentic era. And it is what's going to make everything that you see next possible. So what do we do actually on top of that architecture? Well, three things and you heard it introduced earlier with Kerim as well. And the order matters because most vendors will offer you one of these three things. We do all three and we connect them. So you have reimagine and this is where every project should start. But most don't, right? This is where you use AI to go and redesign how you process work right before you build anything. You're not bolting AI onto broken processes. And once you've designed what you want to build, then you need to go and build at the speed of AI using a platform built for enterprise scale on day one.

You're not prototyping AI, and then six months later you have to go and replatform. Okay. And then lastly, once you've built what you have, you need to go and run it with AI agents that you trust. They need to be predictable. They need to be governed and they need to be auditable. And they need to be working alongside your people and your systems. So the key here is these are not separate tools that you have to go and stitch together. What you re-imagine and design is what you build and what gets executed. You have the same golden thread or artifact going all the way through these three. So what makes this possible is the Pega Platform. And what you see on screen is the autonomous enterprise architecture. Now it's a lot to take in on one slide. It was kind of intentional because we wanted you to see the kind of entire picture first before we start breaking it down a little bit.

But since this is a one on one, I will take it slow and I will start to introduce some products and capabilities that brings these three pillars to life. But to orient you bottom left, reimagine, as you all know by now, that is Pega Blueprint. We have build on the far right that is Infinity studio. And at the top you have predictable AI Agentic, Process Fabric. And of course, that most important thing, case management. That is the one element And underneath it all you have any AI, any data, any Cloud a completely open architecture, you do not have to go and rip and replace what you already have in your estate if you don't want to. So that's the map. Now let's go and dig a little bit deeper. And we start at the design. And as we see this is of course reimagined. And this is Blueprint.

And Blueprint is where you take that initial idea that you have, and you turn it into that working application, whether that's a new product, whether that's a legacy system that you're looking to modernize, or whether that's that broken process that's been broken for many, many years that you want to go and start to reimagine. And if we look at how we do that today, I'm sure many of you in the room will agree. Business goes and writes a requirements document, throws it over the wall to IT. It goes and delivers. It reads. It delivers it in, say, three months time. It's not quite what business wanted. And now that AI that you bolted onto that broken process doesn't work anymore. And Blueprint tries to fix that on day one. Your business and your IT teams collaborate in the same workspace. They see the same journeys, the same case types, the same decisions.

The AI agent goes and analyze your legacy systems. It goes on research industry best practices, and it generates this optimized starting point for you in minutes. But as Kerim pointed out on stage, what a lot of people new to Pega finds quite surprising is Blueprint is completely free, right? You can go on to Pega.com use it. There's nothing to install. It's browser based. There's no license requirements. But let's dig a little bit deeper under the hood of what's happening when you use Blueprint. And three things power every Blueprint. The first design time AI reasoning. So when you describe what you want, as I said earlier, Blueprint doesn't start from a blank page. It reasons up to four against up to four different resources. That can be Pega's own best practices, which we've built up over 40 years of implementing mission critical applications and workflows. We could allow our partners and you saw some of those examples on stage.

They can upload their own IP to Blueprint and it can use that. You can use your own best practices if you choose to upload it, or we have access to the broader internet, which we can access safely through large language models. The second bit is, and this really matters if you're a regulated business in the room, your data stays yours. The documents that you upload are stored privately and securely. We don't use any of it to train any models, and it remains private and stays on your cloud or your tenant. And the third thing, when you're ready, Blueprint then unlocks that design and that application into your choice of cloud. So whether you are on a private cloud, whether you're using PegaCloud or you are on premise, what you build is what gets executed. So you get AI reasoning, reasoning with best practices baked in.

It is private to your business and it is open to your cloud. Now there was a lot of show and tell of the Workflow Automation Blueprint is the thing that you saw maybe up until now, but there is one more one, one more thing there. And this is how you actually engage your customers in the first place from a marketing perspective. So there is a second Blueprint. It is called the Customer Engagement Blueprint and the idea is the same. It's free. It's browser based. It's available on pega.com. But here you upload your business objectives. Let's say I want to improve my customer retention time. Okay. So you update your brand guidelines, you update your marketing assets, and then GenAI goes and analyzes that and it generates an action and experience library full of next best actions, uh, full of customer journeys full of even a simulated ROI model.

And that's before you write any code. Okay. And when you're ready, you take that customer engagement Blueprint and then you import it into the Customer Decision Hub that is. Pega. AI brain for personalized 1 to 1 customer engagement. It's the same engine that Wells Fargo used for that revenue uplift and the example that Anu gave earlier. So you design a blueprint and you build in CDH and that's reimagine. We have this great starting point now. But now what do we do with that starting point? And this is where build comes in. And this is where our new brand new paint is still wet on it. Infinity studio comes in and Infinity studio is, as you saw, our AI first studio. Okay. And the key word here is AI first. We didn't just take AI and bolt it on to our old existing studios. We reimagined the whole experience with that AI assistant at the heart of it.

And that AI assistant that you can see on the left there, and you saw on the keynote stage, that's your build partner. It goes and generates components. It can tell you what your application is doing. It can run test suites. It can troubleshoot all using natural language and what this means. Even though you might be a senior Pega developer or someone that's just getting started, you can have a working conversation with your application. And then you see there on the bottom right, those AI providers, you have complete model choice. You can switch AI providers at any time, whether that's copilot open AI, anthropic. So no matter what your organization is standardized on, you can go and use it. There is absolutely zero login login at least. But here's what actually makes it super exciting for the first time in Pega history. You do not have to be using Pega's tools to go and build Pega applications.

I'm sure many of you in the room have seen Claude code by now, but if you haven't, this is that third party. AI coding tool that you that you can use. And this is connected to Peggy's new MCP toolkit. The developer importantly is in their own terminal. They have not logged into Pega, right. But they can go and interrogate the Pega application. They can go and generate and run tests. They can go and make changes across multiple code bases if they want. And the point I'm trying to really make here is quite simple. Your developers can now go and build Pega apps with the same AI tools that they already are using today. Cloud code, cursor codecs, GitHub copilot, whatever your organization has standardized on. It really is the most open Pega has been for many years, and we're just getting started. But enough slides.

Let me show you a short video that will bring both this reimagine and build to life. I'm going to start by opening Blueprint, a non developer tool where I can type in plain English what I want, a retail bank product complaint workflow Blueprint generates the entire process, stages, steps, personas, channels, and the data model. I then make a live human edit, adding an overdraft fee dispute path to show that the user remains firmly in control. Here's the empty Blueprint canvas. I type in my prompt in the background and add supporting assets. A screen recording of the existing process. A process specification document and a code analysis report of a legacy complaint system from AWS. Transform. The canvas reads this process IP and generates a complete case Life cycle in seconds. Complaint, intake, triage, investigation, and resolution. It auto populates the steps needed to get work done, and a red amber green view highlights where action is required to improve workflow fidelity.

I add the overdraft fee dispute path using the assistant, which suggests a new branch in real time while the business owner stays in control. I can navigate to the data model driving the cases, including customer account, complaint document and resolution. Complex rules and decisions can be defined here to control how and when the workflow proceeds. Standard reports for operational data can also be configured. The next section allows configuration of user authentication and inbound interactions from file drops to Kafka messages. The central pane shows where application data is sourced, whilst on the right, I can define integrations with enterprise systems and data sources, either manually or via configuration files. After confirming personas, I move to the feature selection defining engagement channels. GenAI capabilities. Knowledge. widgets, next best actions and interaction behaviors such as customer verification and dialog scripts. Using the preview feature, I can now see in real time how the application will appear to back office users, mobile fieldworkers, and how it integrates into other enterprise applications.

I can also view the experience for contact center teams, customers on web and mobile, and even a fully functional chat interface that automatically understands the semantics of a request without the traditional need of training and maintaining NLP models. Let's think about this. Not only have I just defined a workflow, it's immediately agentic and I'm able to experience it, test it, modify it. I can even call in and talk to it live. I can also emulate how the application behaves as an Mcpp server, enabling agents to initiate and interact with Pega cases directly. The assistant can be used here too, allowing UI and data changes through plain language requests. Thanks to our Center-out approach, these changes are instantly reflected across multiple channels. Blueprint brings business and technical requirements together in one place and lets me experience it, reducing discovery time and helping teams reach early consensus on the desired outcome before export.

I receive guardrail guidance on best practices. When exported into Infinity studio. The Blueprint then becomes a working application. No translation or rework required to complete the build. I use the AI assistant to add an Agentic investigation step, generate test scripts for the overdraft path, and integrate a document summary agent. The assistant builds these components, reports back, and gives me a clear path on what needs to happen to trigger testing and move to deployment. Importantly, it doesn't generate prompt code. It creates rules, properties, and data transforms based on Pega's model driven architecture, ensuring everything is explainable and auditable from day one. Within a single studio, I can move through design, build, test, and deploy. I can also ask Claude to log in directly and summarize a Pega application, including case types and technical details. I then instruct it to create a last audit date property in the policy compliance workflow.

Claude initiates a change request, identifies the need for intake and schema creation, then implements the property and provides a clear summary. The change is logged under an approval workflow, ensuring the human remains in the loop. This combination delivers the flexibility, speed, and scale of AI in a controlled, predictable way, orchestrated and guided by the power of the Pega Platform. All right. Let's see. All right. So I love that video. But what you just saw is the design, the reimagined, the build and the run. And this, this slide is a picture, a visual of what you just saw in the demo. So there's three columns and three different ways to build in Pega. On the left we have Blueprint, as you've heard about by now. At pega.com/blueprint. And that's powered by the Blueprint AI engine. It's online, it's free. And what I love about it is Blueprint is version agnostic.

So you're not stuck designing in a particular version of Pega, right? And it's AI driving your application design. In the middle, you have Infinity studio with that AI assistant for teams who just really want to build inside Pega. And then on the right, you have your AI coding agents, as Phil said, Claude Cursor Copilot connected through the Infinity Agent development toolkit. So there's three entry points into Pega but one Platform underneath. And what matters here is the the arrows between the various columns. And as Phil said, we think of this as the golden thread starting at design and flowing through to runtime and execution. And it's the same application that's accessible from all three. So imagine a business lead sketches a workflow in Blueprint. A developer could extend it in Infinity studio, and then an AI coding agent could generate test scripts that run overnight.

And this is really how we view Agentic enterprise application development. And so you've just seen how we've spent a lot of time talking about how you can design and build with AI. Now we're going to move to the part of the architecture where you're going to run your systems. And that's all about having AI that's predictable. Agentic Process Fabric. As you heard about in the keynotes and case management, which provides the context for all of these agents. And this is where AI projects either deliver value or you'll see them quietly disappear. And let's talk about why running AI is really hard in an enterprise, because it's not the same as spinning an agent up in isolation. And think about an analogy. And that's a new employee joining a complex organization like yours. They have a degree. They have skills. They may even be really, really brilliant.

But before they can actually do anything useful, they need some structure, right? They need to know your rules, they need your policies and they need some data. They also most importantly need some supervision and AI agents, I feel like are exactly the same thing. They're very, very capable, but they're also brand new to your business. And without some structure that you can dial up or dial back, they could freeze. Or actually scarier is improvise, which is why the leaders we talked to, Phil and I are wrestling with two key questions. The first, how do I build these agents that I can really trust that follow the rules of my organization, the regulations, the standard at scale, every single time, not just in the demo I saw the second. How do I deploy them effectively into my organization and my operations? Because the trust isn't enough.

I need these agents to get into the work. I need them in the channels right, that my customers use. I need them in the systems my employees use, and I need them in the cases that my back office processes, right? So you have to get them deployed. And this is really the same challenge we talked about years ago when we saw data science teams develop these brilliant, brilliant predictive models, but they really struggled operationalizing them within organizations. And these two questions of build and deploy are the center of how you run. Right. And of the two, I think the trust is the harder one, because if you don't trust, you can't deploy, and none of that matters. So you can develop the beautiful workflow across multiple systems. But if the agent at the center of it does something wrong, 1 in 100 times, things are worse, not better.

So let's talk about what makes AI agents trustworthy. And that starts with understanding how enterprises operate. And this is what enterprises typically look like, right? You have dozens of systems, lots of channels, people, agents, tools, partners, and they don't talk. And obviously AI can help, right? But it's not whether it can help. It can. But how do you ensure it's all done the same way every time with the right context, the right data and the right authority at the right time? Because we feel like an AI agent without context isn't intelligent. It's, it's just really, really confident. And being confident without that context provided in the case at enterprise scale is how things can go wrong. So let's make this concrete. So you have a customer filing a complaint with their bank, and the complaint may be 2,000 pounds goes through a mobile app on a Tuesday.

By the time that complaint is resolved, it's hit six systems, multiple teams, a few AI agents. And you need regulatory disclosure frameworks that vary by jurisdiction. And this is the type of complexity Pega is really, really good at. And Phil, why don't you walk us through how we do this? Yes, indeed. Thank you very much. Anu. So let's take that 2,000 pound complaint that Anu just mentioned there. That lands at the top left there, the intake stage OK Pega goes and captures all the details via the mobile app. That could be the original claim form that is attached to the case. And then we can use AI to start analyzing it. And this is where something like our document agents can help or AI for short. Okay, so you can go and analyze that original claim, maybe the original rejection letter, all the supporting evidence that was attached to the case, right?

And then it can go and generate a notification. And maybe in that notification, there's a reference number and maybe an indicative timeline on how long it will take to resolve. And all of that happens in seconds. And really what you're looking at here is 40 years of case management, right? We have the intake evaluation investigation and resolution stages. Okay. And really this matters a lot because if you listen to the industry these days, there's this concept of an agent harness. They're saying AI needs an agent harness to be effective. And really, if you look at case management, it is the perfect foundation to give your AI agent that structure, that context and that governance, which is the three things it needs the most. Now, if we look at how this complaint moves to the next stage, this is where Pega starts to operate a little bit differently to the other vendors in the market.

So a complaint moves from the intake stage to the evaluation and triage stage. Now the previous stage followed very structured flows during the evaluation stage. Here we can now start getting a little bit more dynamic, a little bit more AI guided. Maybe we can now go for risk and fraud analysis. Let's go and call an AI agent. Right. It can do its own reasoning, right. It can go and do a lookup on the claim history claim size. Look at similar patterns across thousands of similar disputes. And we connect to that agent maybe via MCP. And when that agent comes back with a response, we assign that risk category. And once that risk category is assigned, we now back into a structured flow. So we've gone from that dynamic UI AI guided flow to a structured flow. And now your compliance checks and your, um, you can't see up there anymore.

Your evaluations happens the same way every time. So most platforms make you choose one or the other. You either have to choose rigid workflows where AI has no room to breathe, or you choose an authentic AI process where AI is kind of freelancing and making things up as it goes along. Pega gives you the best of both worlds in the same case. So you get predictability where regulation demands, and you get flexibility where AI adds the most value. All right, let's move on. The next question you look at is, well, who handles what and when. So this is where something like Process AI can come in. Now, Process AI is different to GenAI, which we've been looking at up until now. Generative AI, as the name indicates, is good at creating things, summarizing things, extracting text. Process AI is really, really good at predicting things.

It watches millions of cases and learns the patterns. Which cases miss SLA, which advisors solve, which disputes the fastest. And if we look at the example on screen, Process AI is figured out that there's a 70% chance that if this piece of work goes into the generic queue, we're going to breach SLA. So it doesn't put it in the generic queue. Instead, it routes it to a senior travel advisor or specialist who has handled this kind of dispute before and is able to do that within the SLA timeline. So this is not keyword routing we're doing yet. This is intelligent routing using Process AI. And when that specialist or maybe another AI agent comes and works on this case further down the line, they get to see the whole picture, the claim history, that risk score that was assigned from triage, the recommendations from the AI agent that pre analyzed this before.

They see every previous interaction, they see every document that was attached to this case, and they are ready to go right from the start. And this is exactly what makes AI agents and people effective in practice. They never operating without context. They're never asking the customer to go and repeat themselves, and they're never making decisions with only half the data available. But as we all know, context on its own is not enough. This complaint will also need the right people to work on this at the right moment. And three things help with that. The first thing is SLA tracking. So if an underwriter or the advisor doesn't do this piece of work or reviews this work within eight hours, the system escalates it automatically. Nobody has to keep track of this work. Nobody has to chase it. The work keeps itself on track. We could also go and use something like Agentic assignments.

So maybe we need to get a senior policy expert involved in this case further down the line, so an agent can go and reach out over email or maybe over SMS, send it a one paragraph summary, maybe some recommendations, ask a simple question, and then the senior policy advisor doesn't have to go and log into Pega. They don't have to go and read 40 pages of documents. They can respond to that question in the channel that they're already using. And third, we have a concept called get network. So this advisor runs logs into Pega on a Tuesday morning. They see the most important work that they need to work on assigned to them, and that the system is determined they need to do at that point in time. So really, the net effect of this is AI is not necessarily replacing your people, but it is making your people way, way more effective and making sure that they're always working on the right thing at the right time.

Now, what if that same complaint came in through a different business line? So imagine that same complaint maybe came through the Singapore office instead of the UK office. It's the same complaint type. You still need the same outcome. But now maybe the regulatory disclosure is different. The complaint window is different. The escalation path is different and certainly the language would be different. Most platforms will let you or will ask you to go and build a process for each of these variations, and before long you have 15 different variations all drifting apart. And this is where Pega Situational Layer Cake handles all of that. You have one core process. It's differentiated by region, by product, line by business, but only where the business demands you change the core and everyone else inherits it. You change the regional rule. Only that region is affected. And it's also not just regions.

You can apply the same thing to those channels at the top. So our complaint originally came through the mobile app. It could have easily come through voice chat email or through the contact center OK whichever channel it came in on. The logic, the data, the decisions that you make is always the same. And this is what we mean by Center-out, which Alan introduced on the keynote stages. You define the logic. Once in the center, it flows out to every channel. Automatically. You make one change and it's instantly reflected everywhere. What that means for our customer, whether they called in, whether they emailed, or whether they opened up the mobile app, they get the same answer every single time. And six months from now, when the regulator calls in and believe me, they will. You can show them everything as everything is audited in the Pega Platform in real time.

That's every decision, every agent action, every human approval, every escalation, every document version, every channel touchpoint, every regional rule, every model that you've applied. This is what we mean by end to end traceability, not we can audit it if you want. Now, every step is editable by default in the same system in real time. This is exactly the difference between an AI capability you can demo, and an AI capability you can deploy in a regulated industry in production. So let's summarize that and pull it all together. What what did we achieve for that customer? Well we acknowledged in second using document agents. We triaged it intelligently using dynamic AI guided workflows. We routed it to the right specialist using, say, Process AI. We resolved it against the right regional rules using our Situational Layer Cake. It was audited end to end, and all along the way we kept the human in the loop.

We could have easily straight through processed this as well if that was indeed a requirement. So it's a new set earlier speed efficiency, compliance. You get all three. When it comes to Pega. But let me actually again show you what this looks like. Larry Clark, a mid 20s high earner, receives a late payment charge on his account. He decides to raise a complaint. Larry messages saying he had an issue about a late fee. Pega automatically understood that we should launch a complaint case out of all possible service requests based on the semantics and language. Pega then automatically rephrased the ask from Larry to pre-fill the complaint and asked for any other information needed, for example, supporting documentation. As it is situation and context aware, it can collect information to assist processing and prevent downstream rework. He's then presented with a summary and asked up front to validate the submission.

Gone are the days of dumb chat bots. This is a genetic self-service that determines the appropriate workflow, understands what's needed to execute, and communicates its actions directly to the customer. Now comes my work. I start by landing on the work dashboard, where I can immediately see the cases assigned to me in my work list. I can clearly see what needs to be done next and the urgency of that task. I select Larry's case and begin my review. I can see that it's currently in the triage stage with key details already populated, including customer information, accounts, and sentiment. Process AI looks at how thousands of past complaints actually played out to predict this one. And it's flagging a 23% chance we missed the 15 day SLA slightly above our 14% average, and mainly because fee disputes involve extra review steps and 12 days have already elapsed.

The good news is that the clean complaint history and the late fee statement Larry already attached are both pulling this risk down. So this case is manageable if we keep it moving. We can also trigger external agentic calls, in this case pulling actionable risk data into the case. I can see that doc AI has automatically generated a document summary of the attached PDF, extracting and displaying key data into the case, payment, behavior, timelines, and any notable anomalies. Removing the need for me to read the entire document in detail. Gen AI Coach gives me case specific guidance. Firstly summarizing the chain of events, then providing me with advice on what are the next steps to be taken. This can be incredibly useful in long running cases, or where multiple experts are required to give input or make decisions. We can predefine instructions to increase efficiency for newer team members, quickly raising productivity and making every employee your best employee.

For wider questions around operational procedures, I can access Knowledge Buddy, which is trained on our enterprise knowledge stores, to return qualified answers quickly. Due to the low value of the complaint. Larry's standing and potential churn risk. Further investigation is not required, and the issuance of a goodwill payment is suggested. In certain cases, we can trigger an arbitration step for review, where the AI suggested in human opinions are weighed and overridden if required. Keeping the human in control throughout the process. I also have full visibility of the case history and audit trail. Every action, update and decision is captured automatically, ensuring transparency and compliance without additional effort. Once the investigation is complete, I move to the resolution stage. Here, the system generates a draft closure communication using Gen AI automate. The email is already written in a clear, professional, and empathetic tone tailored to the customer and the outcome of the case.

I review the generated communication and if needed, I can make minor edits. Otherwise I approve directly. This eliminates the need to write customer communications manually while ensuring consistency and quality. After approving the communication, the system sends the final notification to the customer. Finally, I can see that the case is now completed and any follow up actions, such as collecting customer feedback, are triggered as part of the process. All right. So we're going to wrap up quick because I know we're at a time and you have other sessions to go to. So what you just saw is Pega runtime AI in action, conversational agent, semantic AI worked. We ran through the four stages Phil walked you through, and then Pega could hand off to third party agents. And the reverse was true, right? You can also invoke Pega from a third party agent and execute that complaint flow.

And one of the points I'd like to make is we're not saying don't use agent, just they're brilliant. Just use them with some structure and do ensure that you're executing what your business designed, and I'll fly through this. This is really about Agentic Process Fabric, which you heard about in the keynote. Lastly, there's a couple of sessions Phil wanted to highlight that would be really good to go to. Yeah. So again, come and see us in the Innovation Hub Anu and I will be there with the engineers who built this, uh, come and talk to us, ask us the hard questions and things like that. But if I can call our two sessions, if you don't know where to go next, go to the Daimler Trucks one next. Really digging into what we just showed you today and the one following that is Lloyds Banking Group.

They built out this award winning solution using everything that we showed you today and scan the QR code. Go and use the Blueprint Workflow Automation. The one on the right is customer engagement. Choose the one you want and that's it. All right. So thank you everyone.