PegaWorld 2025: Unlock Agent Experiences: Orchestrate Agents and Workflows With Pega Agentic Process Fabric

PegaWorld 2025: Unlock Agent Experiences: Orchestrate Agents and Workflows with Pega

Thanks, Carla.

Welcome, everyone.

My name is Krishna Potluri, senior Director of Product Management.

Pretty much excited about the session today and really looking forward to share all the new and amazing capabilities that we have with us today.

So yeah, over to you, Carla.

Let's get started.

If we look out there on what people are saying about this, Agent Aki is going to improve operational efficiency.

It's going to augment employee productivity.

It's going to automate every corner of the business.

It's going to accelerate innovation with new products, new services, new business models.

It's going to reduce costs.

massive opportunities out there.

Massive opportunities.

And the truth we know this and that's why we are here.

That's why we are here.

We are day one.

You saw the keynotes.

You went to Innovation Hub.

If not, go to Innovation Hub.

Amazing things in many of the booths including Agentic, Process Fabric and agents.

So go there anyway.

We know this and we have all this excitement.

And I'm going to propose something different because like we saw in the keynotes, there are too much fuss out there. And I don't want to unleash the Kraken.

So instead of unleashing the cracking, what I'm proposing to all of us is that we pause.

We pause to integrate what we've seen all day, and we'll continue to see.

So I will propose to you a small experiment.

Bear with me.

Trust me.

Okay.

Put your feet firmly in the ground.

Please, folks.

Sit comfortably.

Okay.

No laptops, no phones.

Please.

And start to calm down.

And close your eyes.

Close your eyes.

And breathe.

Deep breath.

Ah! And let go.

All the fuss is out there.

But here we'll calm down.

Because we need to think clearly.

We need to integrate what we've seen.

We need to think on how to transform all these promises into reality.

So let's take a deep breath.

Feel all the air in your chest. In your belly.

Hold for a bit and let it go gently feeling your body relax and we'll do it again.

Let the air in.

Deep breath.

Hold.

Feel the energy flowing.

And when you let go.

You'll see.

Feel your body relax.

And this energy going from your chest to your legs.

To your feet.

And deeply into the ground.

One more time.

Deep breath.

Hold.

And slowly let the air go.

And feel this energy flowing down to your legs, to your feet and into the ground.

Deeper and deeper.

Now with your eyes closed.

Keep breathing.

Please just visualize in front of you an AI agent.

It doesn't matter.

The shape, the color, the edges doesn't matter.

It's your brain bringing to you what an AI agent is for you.

Now bring to your mind all the amazing things this agent can do. It can generate documents, images now, even videos it can generate.

It can gather data, analyze things for you, generate reports.

It can understand images, documents, automate things for you.

Feel how your body.

How do you feel about this when you have this agent doing all these things for you? Now bring to your mind every time the agent makes a mistake.

Just like we saw this morning.

You ask it to do something.

And the answer is not what you expected.

You ask it to create an image or maybe a presentation, and it missed the mark.

You ask it to.

Summarize things or do research for you.

And when you look at the result.

At first it seems right, but as soon as you look up a little closer, there are some misleading information, some things that seems weird doesn't seem real.

How do you feel about that? Now bring to your mind this agent interacting with your customers, answering their questions, loan requests, making up with plans on the fly, disputes, just making another plan on the fly.

Maybe Healthcare claims when someone is really sick, a loved one making up plans on the fly.

How do you feel about that? Now, with your eyes still closed? And visualize that you are turning left and you are walking.

And as you are walking, you are walking in the corridor of the company.

You work and you are walking and you turn right.

You see desks with your colleagues there and they are all surrounded by agents.

And you look left.

Empty spaces and more agents.

And then you walk to the center of the room.

You look at your hands and all the control that you have on your operations, the execution of your processes, the way you want them to happen with each and every one of your customers. Visualize this in your hands.

And visualize that you are just letting this go.

How do you feel about that? Now open your eyes.

I hope you could picture the image Because we all know that AI agents are fantastic and they can do amazing things for us.

And basically it's a genie that's out of the box.

We'll have them, we'll have them all over the place, but we need to have them in the places where they are the best candidates for the best AI, for the best use case.

As we saw this morning, and when we are dealing with customers with this kind of interactions where you need to have this sort of predictability, we need, it's okay when the agent makes some mistakes and it's it's us employees working at that.

And we understand and we see and we can identify when these mistakes happen.

And that's okay.

It's amazing when we use agents and to do innovative things like hypothesize things, do research, role play.

It's amazing what they can do and we definitely want to use them on these situations.

But when we need predictability, these AI agents, they are just black boxes.

They have their big LLM bring you do a request, you ask something they understand reason, make plans, execute, and you basically have no control about it.

It's unpredictable.

It can be unpredictable, it can be inconsistent, and it can be disconnected.

How do you feel about that? So let's just let me just tell you a story.

And this is a story about U+ comms.

As you know at Pega we love U+ group okay.

So U+ comes very innovative comms service provider.

They modernize their their networks.

They really invest to modernize their networks, invest in new services, invest in reliability.

They want to have awesome customer experiences In digital first models. The are interested in agents of course.

So let's see how it goes.

This is John.

John switched to to Uplus comms three months ago, and when he did that, he was offered a nice credit for his phone trading that he quickly accepted and was already imagining how he would spend this money in his upcoming trip to Vegas.

What happened is that three months later, he still haven't seen his credit.

So.

And besides that, he started to see some weird roaming charges in his account balance.

So he does what we would all do.

He engages with contacts, customer service, and a nice AI agent jumps in.

It's a great conversationalist, quickly understands what's going on that John is annoyed that he has a pending credit, that there are roaming requests and that he is coming.

He has a trip in in the coming days.

Index.

The agent makes up his plans on the fly.

Okay, I have all these things to handle.

Start to connect with the data, the tools, the knowledge that it has available and figure out what are the root causes for this.

Yes, there is a trading credit that was approved, but the synchrony, there was a synchronization error between the billing and the trading systems about the roaming charges.

I couldn't figure out the, the, the, the root causes because I don't have access to these tools.

So sorry, I can't, but this is an agent with a mission.

And what's the mission? Keep average handle time low, maybe, or keep customer satisfaction high.

So this is the mission.

The agent comes up with a plan.

Okay, I will do a manual credit adjustment for you.

It will be in your account in two hours. I will revert the roaming charges.

I will send you a notification in 24 hours.

Can I do something else for you? John is happy.

John is happy.

The agent fulfilled the mission.

But in reality, what I see is a lot of disconnection systems that are disconnected.

Agents that are disconnected from other systems, and agents that are disconnected even from best practices, policies, regulations that the company has.

Next day, the result is what we could expect.

John has not seen the credit nor the roaming charges reverted engages again and it goes all over again just to figure out that there are more errors.

Until we reach to a point where we need someone to solve to sort this out.

Them.

And that's where Beautiful Sara comes in, the customer service representative.

That has to do that, is there to serve customers, to offer beautiful things and amazing things that will solve client problems, but instead has to deal with customers frustrations, these disconnected systems and agents.

Does this seem like science fiction to any of us? Does this seem like science fiction to you, Krishna? No.

So does it sound like science fiction to anyone? No.

Like, I'm sure like the same question.

Maybe if it was asked many years back, like six, seven years back, somebody would have said, like saying, you know what, we're going to deploy these agents that will really coordinate themselves and get job done and run the enterprise.

Maybe it would have sounded like definitely a science fiction for the future, but given where we stand now, it doesn't look like a science fiction anymore.

It's a reality that we see agents getting deployed and getting the job done.

But coming back into this specific situation where we are in it does feel like an uncomfortable truth, isn't it, where we have a lot of agents deployed and each of these agents have a mission to accomplish, and each of these agents are trying their best to really coordinate, but it looks like there's a lot more things that really needed to be coordinated across. What is that we are seeing here in this U+ comms example, and a use case that we that we had looked into.

We are seeing a lot of agents all reasoning, planning and somehow disconnected with each other.

They together even could not solve the issue that John has really reached out to, asking for a simple claim related dispute issue that really could not be solved with all these different agents in there, and we are left with having this sort of a customer journey that is disconnected, and then there is no consistency in which each of these agents were responding back.

And we have a situation to deal here.

You know, like like like the frustrated John, I empathize, you know, sympathize with him.

But I also empathize with Sarah, like wherein the situation is with Sarah to really solve this issue for John.

Let's again taking a step back.

What is that we are seeing in this case? In this case, we are really seeing agents all disconnected from each other, putting an outcome at risk.

Each of these agents planning on its own, reasoning on its own, and coming back with independent responses with each iteration.

It is really it's impossible to have predictability on such sort of an agents.

Additionally, having these agents disconnected with each other is bringing in these inefficiencies and also inability to really execute a process end to end.

That's the reason why we have a situation here to deal with the situation, that we have a fragmented customer journey in this case.

In general, like Carla said, like agents are great conversationalist.

They're really great, but they sort of always provide a fragmented view of the organization.

Technically speaking, we all understand why each agent most likely is deployed with the context of an application.

So which means it is defined and confined to the scope of that specific application and not fully aware about the enterprise landscape, wherein there's a lot more applications, a lot more agents that could even support and coordinate in that sense.

So this is what I see here in this up+ comms, the way their agents are deployed and how things are really working out for, for in this case, for John.

Exactly, Krishna.

And that's why we think of a different approach. We know and we value all the power that agents can provide.

But agents are unpredictable.

So what if we could have all these powers of agents, all this flexibility, this dynamic behavior, all this power to think about all the different alternatives and role play with you, and also have the predictability of workflows, the governance, the orchestration that workflows provide.

And have this across all the applications, across all the enterprise.

And that's what we are Pega is bringing with Infinity 25, with Infinity 25, Pega is embedding Agentic AI deeply into the platform, and it's coming with a unique approach because it's instead of letting all the control to the agents to reason and to make up plans on the fly.

What Pega is doing is that it's using AI reasoning at design time to create and design workflows, always in collaboration with humans, so that you have all the control about what's going to be done.

And then at runtime, deploy agents that are used, the AI power, the semantic power to identify and execute on these workflows whenever you need.

It predictability.

So predictability when you need predictability you have predictable reagents.

And then you can expand on these agents to have also other autonomous scenarios, as Krishna will mention shortly.

And once we have this, then we can provide our customers with full self-service 24/7 to answer any requests and go through to predictable flows with them.

We can empower our employees with a single unified AI platform that they have an always on AI system to help them do their job, their work assignments faster and with more accuracy.

And we can arm our leaders with full operational visibility and know what work is being done, how it's being done across agents and humans alike.

But how does this Don Krishna.

Yeah, yeah, that's really great.

It's really I completely agree on what Carla is saying.

It's it's.

There are two, two aspects here.

On one end, we have the power of AI agents and Other.

And we really are looking at predictability coming into play. So so we really need to think about how is the best option using which we can really embrace the power of AI agents that they really bring on to the table and see how we can also provide predictability in achieving the outcomes.

So this is extremely key because most often we see people saying that there's an LM model on which we can really create an agent and then get it deployed.

But it's really not being consistent.

I think I really like the demo that Ellen showed this morning.

The way he played through the chess use case and then showed how it sort of evolved over a period of time.

So that's that's how we see agents when they're really deployed without great predictability around them.

But coming back to the topic of what Carla said, we really need to think about on one end, leveraging the power of agents and the other end bringing in the concept of predictability.

So here is where the power of workflows come in.

So across these many, many years we have our clients deploying the workflows implemented on Pega to automate various of their business critical processes, really end to end.

This is some of the aspects about workflow.

Before we really go deeper into agentic things, I really want to reset and also explain about some of the things about workflow that makes them so very key and fundamental to the predictability aspect.

One of the thing is about these workflows.

They capture the enterprise's intellectual property.

They capture within them a deep thinking and detailed analysis on how what rules need to be executed, how these rules to be executed, and as well, more importantly, the rules and regulations to abide by the SLAs to be adhered to, the escalations that need to be dealt, and so on and so forth.

It is for these reasons enterprises sort of deploy these workflows so that they can achieve consistent automation at scale.

The other aspect about workflows is as well that they are extremely visual in terms of the definition.

So anyone looking at that sort of workflow will be able to clearly understand the stages, the processes, the steps within the sequential coordinations, the exceptional paths that need to be executed, where we have an automation and where we have a manual task.

So all of this is greatly graphical in that sense. So what does it really result in? It results in having a greater, deeper trust in looking at an automation and having that understanding.

I really understand what this is really going to do.

The other aspect is as well that it brings in predictability.

So and moving on the workflows, the third aspect that comes to me is about the weaving pretty well as a framework, bringing in the automation aspects together along with the tasks that need human intervention.

And these tasks that need human intervention are aptly routed to the right individual, available in the organization with the right skills, and then get it to be ensured that the psl's defined are as well met.

So these are the various capabilities that I think when I, when I really think about workflow and why they're really key in bringing in repeatable, consistent and predictable automation and at scale.

Now when we think about, again, the outcomes that I mentioned in the past, power of AI agents need to be brought together and the predictability of workflows.

So this is where where we think greatly that we should be really heading towards.

With Pega Infinity 25, we are launching a predictable AI components coming in as part of our release.

So let me make a quick mention about some of these capabilities that are going to be made available soon.

So the first thing that I'll talk about is the design agent.

The design agent helps users create and compose processes end to end without really needing to start from scratch.

The design agents like the Blueprint and the Autopilot help create powerful applications.

And not just applications, but workflows and every detail within them so that it's very easy to understand what this workflow is going to do, and is that abiding by the regulations within the enterprise, and also ensuring that sort of definitions comply with the industry standards that are out there.

So this is what design agents are really going to do.

Most important, I think even the morning the demo that Kerim has given, it's definitely not the case that we should just use the GenAI based Blueprint outcome and say that's exactly what we should be deploying.

So we should own it in the sense we always have this discussion saying it's not the outcome that that we should really think as and then have a discussion whether is that really 100% correct or are there any issues with it? Every organization is unique, every workflow is unique. And then GenAI Blueprint the way it generates it has a start.

It gives you a platform where there is a great collaboration that could happen between the business users and the technical users, and have this understanding on how they would like to standardize this sort of a process and then take it to the implementation.

So when we hear stories about having clients getting started with Blueprint and then going live with this application, it is it is this aspect that really are they're leveraging to their best.

Now moving on.

The second agent that I would like to mention is called conversational agents.

This is completely new coming in Infinity 25.

Conversational agents allow customers and employees to really have a consistent replies.

Each and every time that they interact with conversational agents are going to have a semantic layer which sorts of receives the request, understands the intent, identifies the workflow within the application that need to be executed.

Get started executing the workflow and ensures that there's a handshake between the communication coming from the user and in completing the workflow execution.

This is exactly what we saw in the morning.

Kerim doing a conversation with the agent bot.

That's exactly an example of a conversation agent wherein wherein the most important aspect here is it is powered by a workflow, powered by a workflow.

So that's what a conversational agent is going to be.

And then subsequently in the demo, I'm also going to talk and show a bit more on these conversational agents.

And next comes the automation agents.

Automation agents handle more, more increasingly complex tasks so that we can bring in advanced automation into the workflows themselves.

Some of the some of the advanced tasks would be some research that need to be done beyond the case, maybe document classification, maybe an attachment that is coming into the application which need to be analyzed, details extracted, populated within the case and then provided to the case worker.

So all of these tasks need advanced computation and then advanced logic.

So automation agents are the ones that are going to make the existing workflows more richer, meaning which a workflow can now be added. An additional step using which we can configure these advanced automation agents so that they can perform these logics and then associate and enrich the existing workflows.

So as I say, it's going to extend the current case type.

So which also means that every step that it is going to execute is completely audited and traced and tracked.

So so that's that's what automation agents are going to be.

The next is about knowledge agents.

Knowledge agents are powered by Pega Knowledge Buddy.

They sort of help surface the best practices available within the organization.

They sort of help the knowledge worker or the case worker make decisions more quicker.

Last but not the least is the coaching agents.

The coach agents is so very unique.

It always resides in a context like for example, a case execution, a case having a context of a coach.

A coach, while executing a case has the complete details about the context of the case, the history that it all went through, the user who is working upon the operator who is acting on the past history of the user, and all the details in that context.

With all this in consideration, the coach is going to give in the moment guidance for the case worker so that they can really up level their game and make decisions more with confidence and then move on.

So the couple of the agents here, I talked the design agents are more from the design time perspective.

Like Carla said like we have the agents that can help build applications faster, including the Autopilot in that case, and the Coach agent and the conversational agents are all the ones that we are really introducing as part of the runtime execution.

So which means the case worker who is working on the case, they're really going to see these interact with these conversational agents and Coach agents and get the guidance and then execute them as well.

So when we now think about all these new capabilities, we also see that each workflow is becoming more and more agentic in that sense.

And when we think about a Pega application having all these workflows becoming more and more agentic, we now start to see that each Pega application is becoming a lot more powerful than before, a lot more agentic than before. And when I say conversational agents and agents and all that, we're also introducing agent X API.

I'm sure you would have heard some of you in terms of the demos, and also I've seen them working and all that.

So agent X APIs are going to help execute these agents and then ensure that there is a seamless way in which these agents can be integrated across multiple channels.

So with all this coming in with each Pega application becoming a lot more richer agent, we started to wonder what next, what next? So workflow is becoming agentic design time runtime in between.

And then we also each Pega application is becoming agentic and we start to see our clients saying, you know what the use cases that they really are thinking of and deploying and so on.

When we start to think about it, we started to wonder like, what next? And the next thought that comes is we should also think about a situation where there are a lot more agents across a lot more applications in the system.

So that brings us to the concept of Pega Agent Trainer Process Fabric when we have this process.

Processes defined across applications.

When we have these sort of agents across applications, we started to realize that we really need to ensure that there is a great deal of coordination that should happen across the agents available in the enterprise.

So these agents could be on Pega, or sometimes these agents could be also be defined on non Pega applications.

So Pega agent Process Fabric is an service that sort of orchestrates agents systems in an open agent network, so that every interaction from the client is greatly reliable and consistent.

That's what agent Process Fabric is all about.

One other aspect.

Some of you might have heard the term Process Fabric in the past.

So agent Process Fabric extends the architecture of Pega Process Fabric.

So by doing so, we now have an easier leverage in terms of connecting with all the applications that are registered.

These applications could be Pega or non Pega.

I'll keep repeating that phrase.

So a Pega agent Process Fabric leverages this sort of existing architectural technology of Process Fabric, and so it can seamlessly connect with Pega non Pega applications and bring in the data.

Apart from the data, also be able to register all the agents available in these applications and then provide a seamless way in which we can coordinate and execute the workflows available in the applications.

Provide an access for all the users working on the open assignments and cases on these applications.

Be able to orchestrate also on the agents available across these applications.

So all the capabilities of these um Pega agents, Process Fabric are classified into broad three categories one, I call it as intelligent automation.

The second orchestration capabilities.

And the third is about operational visibility.

Let me let me start about Intelligent Registry.

So Intelligent Registry is all about being able to have a universal discovery and definition phase wherein you have all the workflows available across applications being commonly seen.

Let me take a quick recap on Sarah's job.

Sarah are customer service executive having to deal with the situation with John.

Suppose if she had to really deal with the situation, that there are so many applications out there and each application has agents out there, she could leverage all these agents and to get the job done.

How easy will it be for her to know on a daily basis, what agents are getting created in each of these applications, how to really reach out to these agents, and what Platform on each of these applications is defined on, and then trying to keep a pace on one side.

But at the same time, her job is not about keeping pace, about the technology that is evolving on a day to day basis.

Her job is mainly to ensure that she supports clients like John, who is frustrated and ensure that these customers are really taken care well, so that in the end of the day, the customer retention is held and then her job is to ensure that we she she increases the satisfaction of these customers.

Right.

So so given the situation that we have a role and the deal the employees working with agents and applications across platforms, we started to bring in this intelligent registry using which we have a track about each of the application, the related agents, the related workflows, the related tools that are available across these applications at a commonly available place.

It is completely possible to configure some of the workflows or agents as non discoverable in the sense they are.

They are available right now somewhere in the system, but maybe they're not supposed to be used, maybe they are going for maintenance or whatever be the reason. We can always think about those use cases.

So so these are the Intelligent registry is all about a common place where we could see a common registry of all the agents available across each of the application, and as well the workflows available across the registered applications.

And subsequently let me talk about the Agentic orchestration.

I talked about the conversational agent.

All I said was the conversational agents, has the semantic layer, takes in a request, interprets the request, and decides what workflow to execute in a given application.

The same logic we apply at the Agentic fast Process Fabric level.

Agentic Process Fabric level.

The same conversational acts very differently in the sense the request is considered.

It sort of interprets there is a semantic layer that sorts of interprets the request and identifies and concludes, what are the agents to be executed across the applications that are available in the system? Or maybe what are the workflows that need to be created across the registered applications in the system? Or maybe what is the data that I have within my own instance that can be looked up and then responded back? So it has these three sorts of resources that it interprets, decides, and coordinates to execute, retrieves the use the response, consolidate, consolidates and then provides back to the user.

So Agentic orchestration is all about ensuring a seamless connectivity across agents and ensuring receiving a response back across the agents and then providing a seamless and consistent response back to the users.

In this regard, I'll make a quick mention on a couple of points here.

So I kept saying that there are applications could be Pega and non Pega.

So as long as it is Pega, we try to use the agent API and APIs to coordinate with the agents that are available in there.

But if it is a non Pega application and if there are agents out there, we support the MCP protocol and also agent to agent protocol so that it's completely possible to register any of the agents which are compliant to any of the standards, but not not just that there would be agents who would say, no, it's not about MCP protocol or agent to agent protocol, but if they have any specific APIs that need to be implemented, it's completely possible to have a third party component implementing the coordination, using the APIs to reach out to that agent and then integrate into agent Process Fabric.

So these are the various capabilities that agent Process Fabric is going to give.

One important aspect here is about the context of the user who is acting on agent Process Fabric level is always retained when the coordination is happening across the registered applications. So which means even when there is a coordination happening across agents, across applications, the user context is never let go.

So it is always the user context that is retained.

So meaning every fine grained security configuration that is set into the respective applications will be completely adhered to.

Now moving on.

The third aspect is about operational visibility.

Like I said, Agent Process Fabric extends the architecture of Pega Process Fabric.

The Process Fabric had a schema using which it had the ability to retrieve and regular basis synchronize the data of all the open cases and open assignments from the applications which are registered both Pega and non Pega.

And with agents coming in, the agents now have the access to such a data.

And like always said, like data is the new gold like digital gold looks like it's more costly than the actual gold, but overall so Agentic Process Fabric has this sort of data coming in from across applications and across the workflows, which are ongoing and executing.

So any query that is that a user asks to agent Process Fabric can be greatly leveraged from the data that is available within its own self.

But there could be situations when the data within the agent Process Fabric is just not enough.

Two ways to really deal that the the agent orchestration also will identify saying this data is not available locally and it can reach out to the knowledge agents available in respect to applications and also retrieve back and consolidates and then give it back.

It could also also be a situation where an application developer using this agent, Process Fabric, can extend the robust schema extension capabilities and then extend the schema in such a way that they would also retrieve more additional data than the ones that are available, and then populate that within the agent Process Fabric.

So that will enrich the data at a common place about the metadata more in specific coming in across applications.

So all these capabilities are what makes agent Process Fabric a lot more, lot more, um, wonderful.

Now we'll see a little bit of this in action.

There's a lot of things.

So just recapping agent Process Fabric is that layer on top of that goes over all applications providing intelligent registry, providing end to end orchestration and operational visibility. We'll see here a couple of demos about our use case.

And let's just remember Sarah, our customer service representative.

That has to deal with John's frustration as well as many other customers frustration as the company was rolling out the agents and the number of agents is small.

The number of systems things are manageable for her.

She can interact, understand where to look for.

But as things scale and you have dozens of systems, hundreds of systems and agents, it becomes just unmanageable for her to see where to go.

And instead of improving her productivity, it decreases her productivity just to figure out what the what she has to do.

So that's where Pega agent Process Fabric comes in.

Correct.

Completely.

Completely.

So we have a situation wherein we have the applications in this case, again, going back to the same situation of John and Sarah in there.

Um, so what was the situation? The John sort of raised a billing dispute, and then there were a lot of agents going around the enterprise, which are trying to solve the problem.

And that was the situation they really could not solve.

And so Sarah got in and she really needed to deal that situation.

I'm going to do a quick reset on that a little bit off of that use case, because you Pulse comms is a Pega Pega customer in that sense.

They deployed some of their applications on Pega now.

And so some of the agents now are powered by workflows.

So let's let's start that.

So in this use case the way you see it there is Agentic Process Fabric in the middle.

But also there are three applications that I'm focusing on right now.

One the billing agent.

It's built on Pega. And also there is an KYC application built on Pega.

And third a roaming package related records management and related application built on Salesforce.

Let's consider that as a use case about how we have more than one platform in consideration.

And the request coming in from John is still an open issue.

And now now it is Sarah who really need to deal that.

Okay, so that's the situation where we are in right now.

So when Sarah logs in to the agent, it Process Fabric she would see something amazing.

She would start to see all the work items coming in across applications in a common consolidated portal.

So she would start to see that one of the top most important requests on her plate is coming from John and coming from the billing application, and it is of the highest of the priority that needs our immediate attention.

So usually what happens would be Sarah, in this case I might click on the assignment or use the get next, pick the same thing and then and then start to work upon it.

But Sarah does something different.

Sarah says, okay, instead of me trying to open this specific assignment and complete it for John, how about how about I use a conversational agent and see how the experience is going to be looking like she would proceed to the conversational agent? And the conversational agent is extremely smart.

What it did is about saying, I know if it is Sarah who is coming in.

Her most important task in hand is coming from John.

But wait.

I also should be able to do a lot more support to Sarah than just saying, you know what? You have the most important task coming from John, what it is, what it did is that it sort of did 360 degree analysis about John and then came back with an answer saying, you know what? There's a lot more data here that need to be dealt with.

John first, definitely.

It's about the billing dispute.

So it said like this is the bill that got generated.

And also it also said what is the dispute that is raised and also the reason for the dispute. And also it also said, like, it's easy for you to click on any of the links and start to open them and then view those sort of tickets directly.

Like I said, these two applications are powered by Pega, and so it's possible to correctly open the respective IDs.

Next, it also said like you know what, Sarah, you didn't know so far, but there is one urgent attention item that needs an urgent attention that is that John need to get started doing his update about the KYC as well.

And he's saying, you know what, there's no ticket right now, but do you want to really raise it? That's a good question for Sarah to know the information that was missing so far in her radar.

The third aspect, because John also reached out about the travel plan to Vegas and then said, like, yeah, I really need an, you know, best, best international roaming package.

The system said, like, I think the request is pending somewhere, but here is the best package that you can start to suggest back.

So so Sarah in this case has the information across three different applications getting consolidated at her end.

And then there is a complete 360 degree view through which she understands about not just one ticket from John, but about everything that need to be done for John.

And in this case, it's the agents that sort of use the job from the respective applications.

Pega non Pega consolidated everything and then give it give it back to Sarah.

Sarah proceeds with this, um, conversation agent and says, yeah, this is really great, but let me have a double check.

She opens directly the link and proceeds to the billing application and sees the dispute case indeed feels confident that yes, this is definitely a dispute case.

And definitely this is the $350 value that exists as a claim dispute.

And she comes back and feels yeah, I think that is right.

She could even close the case as a dispute.

But she decides to continue with the conversation agent, and she proceeds to the conversation agent and says, you know what? Among the three things that you told, let's resolve the dispute.

Let's raise a ticket for KYC.

And the system comes back and says, job done.

The dispute that was raised as a case within the billing application is now resolved, meaning which from the next iteration the issue will be completely taken care. There's no disconnected aspect.

It's a complete workflow, completely predictable.

The second aspect is more about here is an idea that I am creating so that the John can completely resolve the KYC request.

I'll make a quick note of that idea.

I think it is i-459.

I'm going to make a quick mention about it later on, but so the system created a workflow using the conversation that Sarah is doing.

That's the most important aspect here.

The Sarah proceeds by saying, you know what? Let me also create an email and communicate back to John.

The system does it so well this time.

It also includes the international roaming package information.

And Sara reviews the request and says, you know what? Send the email and then the system says job taken care.

The email is being sent to John on your behalf.

So here is an example where we have Sara using a common Agentic interface, not logging into multiple applications.

She just has one chat interface and then the system is smart enough.

It's not about Sara asking questions, it's about the system doing the analysis.

Hard work.

360 degree analysis and then bringing back all the data, Sara, and saying, this is how you should be really solving the issue for John.

It's just not about one ticket.

But like every other aspect that came in from multiple applications is completely taken care.

And then the story doesn't end here.

So like like mourning the way, um, Kerim showed the demo, there's always a sneak peak, you know, because that's what builds in more trust.

How can we really trust what's happening here? The what? We we have a greater trust when we can really see what is happening behind.

So let's look the persona here as a manager who can really see what's going behind this sort of an agentic orchestration.

So what we see here is the Agent Tracker UI, wherein agent is informing about saying, you know what, there was a request.

The request semantically was analyzed, three different agents was reached out, and then the responses from these three different agents were consolidated and then given back to Sara.

And then the discussion continued.

Sara continued the conversation by saying, yeah, please get two tickets of these three resolved, and the system is saying that was a text given by Sara.

And this is the analysis, semantic analysis that was done.

And then the two tools that got selected.

And then what are the tools that got executed, and what is the overall response that even came by in.

So we can start to see from the start to end, every interaction that Sara had with the conversation agent is completely tracked can be seen, evaluated, understood.

In case there's anything that we see as a gap, we can always improve upon them.

So this is how the agent tracker is going to look like.

There are a lot more cool, cool aspects here.

It has information about the tokens that that were used for request and response, the time that it took for each of the step, the LLM model for the semantic analysis that is used, and so on and so forth.

And I said like does does it does does.

The conversation agent is greatly remitted to the CSR? Usually, no.

Let me continue with the story with the John because John I said like received an email uh, with with saying, you know what there's an here is the link for I05049.

Uh, they can John can click on it, launch the portal, complete the KYC as well.

That's a possibility out there.

But let's see what John would do.

Let's take the scenario where John received an email, tries to do a self-service portal, logs into self-service portal, and he see a conversation agents welcoming him, saying, John, you have a ticket here.

You know, you can always launch the ticket by opening and then resolving it, but you also have a conversational way in which you can deal with it.

And it's asking about some of the information, saying a couple of more specific details in this case.

And John decides to provide, uh, by by providing those specific details in a chat experience.

And then John gives his first name, last name, email, phone number and proceeds to submit.

So what is happening here? The conversation agent.

There's a semantic layer that understands and executes the workflow.

The I5049 that's intentionally took the reference.

So now even for a client or a direct client who is using a self-service portal using a conversational agent, he's able to interact with an workflow that is getting executed.

The conversation agent went to its next step and said, you know what? There is a photo proof that need to be submitted.

Let's say in this case, a passport, and John is uploading the passport and saying, here is here is my ID proof.

Do consider it for the for the workflow to complete and the document analysis kicks in.

All the information is extracted, populated into the case.

And finally John is informed saying, John, your job is done now.

Now I sort of verified the data that initially what you gave, along with the documented proof, what you gave and submitted all the information back to one of the experts so that they can have a double check on it.

So in case you did the double check step, if not, the the process would have been completed.

So this is an experience of how a John would interface with agent Process Fabric really not knowing about all the nitty gritties that are available within the systems and so on and so forth.

So these are the maybe I think we will lack of time.

So maybe we will be available at the booth, uh, to really talk about a lot more use cases here.

I'm just conscious of the time in case if you're okay, I can continue a bit while.

Is it okay? Yeah.

Yeah, just just a couple of minutes maybe. So the third one of the aspect that I really talked about is having intelligent registry.

How does an intelligent registry in a generic Process Fabric looks like? So? So when Intelligent Registry is all about having the applications, workflows, the related agents, the related tools being available at a common place, let's proceed to see how that's going to look like.

Sorry.

No.

Okay.

This is a landing page where we have the all the applications registered.

There's a graphical way in which it is saying saying about what sort of applications from what sort of platforms are available in there.

And also if there's any admin related tasks that need to be executed and so on and so forth.

So let's start with the applications.

These are the various applications registered for this instance coming in from SAP, Salesforce, Pega and so on.

There are multiple views in which we can start to see.

Let's also categorize by by the Platform and start to open each of them, and then start to see about what an application really means here.

When we open a specific application, in this case coming from Pega, we can start to see the agents coming in.

From there, we can start to see the workflows coming in from there.

And so this is this is the metadata that we do capture when an application is registered.

Let's proceed.

Proceeding to the workflow registry we can start to see about all the workflows coming in across the applications.

Along with the discoverability aspect of each of that, we can mark some of the workflows as not non discoverable meaning, which they won't be allowed to be instantiated from a generic Process Fabric.

Lastly, the Application Agent registry displays all the agents available in the system across the applications that were that were registered and made available in this context.

Yeah.

With that, I'll pass it on to Carl again. Yeah.

So we are over time.

But just to just to summarize key takeaways.

Folks, we are all about using the right AI to the right job.

And the AI that you can trust.

So in Infinity 25, what you will see is fully governed.

Agents like Krishna showed you with the conversation agent, predictable agents, but you also have autonomy.

You also have AI agents that you can configure that help with your employees and your leaders to do all these amazing things that the agent can do for you.

And all this orchestrated and integrated across the enterprise with Pega agent Process Fabric.

I know we are over time, but we will be here for questions if you want to, and we invite you to join us at Innovation Hub in two booths.

Asiatic Process Fabric and creating AI agents.

Thank you.

Thank you, thank you all.