PegaWorld 2025: Smart Mobility Plans, Smarter Networks: TM Forum Inspired AI Catalysts for CSP Evolution

PegaWorld 2025: Smart Mobility Plans, Smarter Networks: TM Forum Inspired AI Catalysts for CSP Evolution

All right. Well, thank you for joining us for smart mobility plans and smarter networks. This is a conversation giving a preview of some of the hard work that our teams have been doing for the two catalyst programs at the TM forum, coming up at DTW in two weeks, which is in Copenhagen, and then we'll also carry it through to some of the other regions, like the Americas in Dallas in later this year.

So first, I want to say a big thank you to Hakan Ekmen, who is the global networks lead for Accenture. Uh oh. I thought Ben was next, I can't see. So, Sanjeev Kumar, um, who is a VP for HCL technologies in the communications industry? Ben Cuthbert. By now you all know who Ben is. He's the local celebrity from. Yeah, he's. Quite famous now with his performance today. Yeah, exactly. From Vodafone Networks. And then we have Axel Wells who is an industry principal for Pega. So thank you for joining today to give an overview of these catalysts.

So let's just start real quickly with what is the Tmforum. For those of you that may not know. Tmforum is a industry consortium. And you know, last year when we were in PT, I think the message was very loud and clear. It's really about collaboration across the industry to open up new ways for the industry to grow. And its focus is what are the data components, what are the APIs, the processes that we can standardize as an industry and create interoperability between companies so that we can grow in the future. And so that's what the Tmforum really is across the globe.

Um, we have two catalysts this year that we will be doing in Copenhagen, and one is the neural net orchestrator. And I'm not going to go into all the details because we're going to talk about each of those here in a minute. Um, and the second one is a dynamic plan builder with Gen I, what I think is most important in this slide, we just talked about collaboration and the goal of Tmforum. And you can see down at the bottom the different players that we've had involved working with Pega to put together these, these industry innovative solutions that we're going to walk through in a minute.

And that is something that for the last, what, 4 or 5 months our teams have been working together day in and day out to get ready for the competition in June, where we will all be judged on the innovativeness, if you will, of our projects, and they will, you know, have winners that win in certain categories based on what we've created.

So let's let's start first talking with, you know, Tmforum is all around collaboration and innovation. What does that mean to you from a tmforum perspective? And why is that so important for the industry?

Yeah, as we kicked off the catalyst project, the target was network transformation and the transformation, the customer support. This was the initial idea. And uh, beside, uh, working with data, with talent, innovation was an important aspect to bring in why we took the autonomous network framework from the Tmforum, which is becoming de facto the standard guiding standard, how you can evolve your environment.

And when you go from level 0 to 1, you have to consolidate the data. Then you need innovation to do that. When you have 18 inventory systems. How to get the common ground. It is with innovation, with new techniques. Then the next level was um, bringing uh, initial automation across, uh, the domains in our catalyst projects. We are touching almost every domain. Therefore, uh, it was not an easy task to find the innovation to, to get the automation in place.

The next target was then radical automation, not initial doing radical automation at scale. Where you can derive conclusions out of the data. And we use many, many data sets. Uh, then innovation kicks in. In the next level you need, uh, you bring in AI, then you need a digital twin. Another innovation uh, to, to simulate to to qualify. Is the AI really working well and is the AI Antigua. And then in the in the next levels we use then uh techniques to scale the AI. Again, innovation was always present and uh, was key in the success of the catalyst.

Can you walk us through a little bit about exactly what is neural net?

The idea. Idea of with the catalyst was, um, using the orchestration in the network, um, looking into the network, uh, into the access network, into the transmission core, and, uh, and finding abnormal behavior. Uh, a copy site is down, packet loss, interference in the network, um, outages. And then, uh, resolving this fully automated closed loop. Uh, this was one part of the, um, catalyst in the second part was deriving conclusions for the end customer in the B2C, uh, segment and, uh, working with the end customer, notifying him that there's maybe a problem solved or, or, um, engaging with him on sales because his home router maybe is not working anymore.

And then doing proactively a call with the agent and, uh, with that, uh, two domains, we have been able to, um, to have a closed loop, uh, interaction from, from the network working with the network, solving issues and, and, uh, impacting the user experience positively and mitigating, for example, churn. This was one of the use cases we have been working and many, many agents, we have developed many, many routines. And, uh, I'm confident in two weeks time it will be exciting to to see the simulation in two weeks. We still have two weeks. Yeah, yeah.

And what, you know, if you look at the slide, there's all these different components that have come together. Um, what are some of the really value metrics that you see coming out of this for the client?

I think initially we started with efficiency. Being an efficiency, um, should result in, um, usually, uh, doing things faster, things you did in, in weeks, you can do in minutes or the process is skipped at all because it is done automatically. It was about the agility. That was one of the first things. Second thing was, um, improving the user experience when the customer is not expecting this, like a, like an event where they think, oh, it's changed, it's about perception.

Then, um, we said we should not touch the system anymore. Should run self-organized. No, the term is usually zero. Touch we said, is self improvement, uh, as a continuous process. The next point was compliance and auditability. It is about integrity of of what you build. Now the the traditional level one, level two worker or the leadership has to be convinced that whatever we do is somehow integer and they can rely on this.

Then everything what we did was tmforum aligned, which is important. We want to um, we want to, um, make that framework more relevant. Uh, if everyone is using this, this is the common ground. Everyone is speaking the same term, the same controls are in place. And of course, the last point it was about, uh, opex efficiency savings, but also then using the savings for, um, for maybe, Additional growth. Yeah.

So. Ben, you live and breathe in the network org all the time. And I know you're really focused on the outcomes that you want to achieve, which we love at Pega because we're outcome focused as well. Where do you see the most value being derived and achieved by getting to network autonomy?

Sure. I think there's a couple of key areas. So I can mention some of them around efficiency. And you know, how much does it actually cost us to run our network and provide the optimum service to the customer that they expect today? Our customers expect a lot more from network today than they did probably five years ago, ten years ago. You know, you need to be autonomous in order to provide the level of service, quality, efficiency when there's a problem, resolve, you know, all of those things. So the customer experience angle would be absolutely paramount to what we're doing with our catalysts and effectively autonomous networks.

Um, but also there's there's a whole bunch of processes that will benefit from a refocus to value streams in autonomous networks. You know what's most important to our customers? What's most important to our business, but equally recognizing what is an overhead and where we would want to make that more efficient and more cost effective in order to operate. And just because we have to operate that.

So, um, we know we have to innovate, um, we know we should innovate. We know we want to innovate. And I think around the autonomous networks moving to level five, there's a huge amount of work to get there. But with things like the Tmforum, it's an opportunity to sort of co collaborate around problems and ideas and stuff like that, and also standardize like just talking about which is only goodness really, if you can get standard, API, standard, um, compliance, all that kind of stuff really, really helps actually, because when you go to actually deliver these things, you're like, well, this is a recognized standard that we're adopting.

And actually it becomes much more deliverable in a much better time period as well. So the investment we require to actually deliver these things is much better. Yeah. So it's it's possible we'll get there. It's just takes time for sure. It takes time. But the more of these kind of catalysts and tmforum events and so on, it will only help really enable that faster for everyone. So it's not just Vodafone benefits, it's everybody. And our customers ultimately will will be the beneficiaries of that.

Which for those that aren't in the communications industry, connecting the customer experience and the network has always been a huge challenge. And so this is starting to paint that path of how do you actually get to a full loop autonomous network, not just the network piece, but all the way through the life cycle?

Yeah, exactly. You know, you don't really think or care about your network until you have a problem, usually, right? Right, exactly. Nobody wants to call up and complain and send somebody out with a truck roll to get something fixed. You know, it's bad for us. It's bad for the customer. It takes forever. You know, we need to be able to solve these problems fast with as little effort as possible. Yeah. So how come? Why is it so important to have a partnership with someone like Accenture to be able to deliver these outcomes?

I think as we had the initial talks over here with Pega, the, uh, the idea was how we can complement to each other. And we found the track easily, the, the network orchestration and the, um, client engagement. What I mentioned in the beginning, I think here we can benefit from each other. Um, we did, uh, recently, just five acquisitions in the network space to have, uh, all what you usually get from from the OEMs, from Ericsson, Nokia, Huawei or from, from juniper Cisco that we have all this expertise in-house.

Then we started, um, thinking about doing things differently. Uh, and uh, have developed over the past 12 months, I think many, many things where you can benefit from. And of course, we are benefiting then to the other direction from your client engagement. Um, and there was one common sense, uh, why this was, I think, also beneficial for you. We think that you have to start with the processes, the process with the workflows, discussing with everyone with something they understand.

And the standard was the we took then I think in the first session, the level two and level three processes based on item. And then we selected two handful five in the network area, five in the customer support area. And when you usually run these processes, the outcome can be replicated. And this is then the moment when the catalyst is ready. The same concept can be expanded. It can go to um a network design. So to enhance, um, the deployment and where the customer will benefit from, they will have fiber or 5G faster and or a network optimization that can be then enhanced user experience and mitigate churn.

Therefore I think, uh, to just to three use cases. I think there's a benefit to work together. And that was the reason why we did this together. Yeah. Yeah.

So we're going to pause on the neural net for a minute, and we're going to jump to Sanjeev and Sanjeev. If you could touch a little bit on the second catalyst and the highlight of some of the challenges and outcomes that that you are taking on with the dynamic plan builder.

Sure. I think the the way industry is evolving, the way I would say, for example, in the US, right penetration, consumer side is saturated, which means it leads to a lot of churn for operators to build their business and grow their businesses, which is a significant challenge for them. We have interviewed talked to multiple operators in Europe, Americas. We find 3,035% of the churn rate, which is a significant erosion of value, You both in business terms, in losses, profitability and so on and so forth. Apart from customer experience impact, it creates.

So it is a huge, I would say, focus area for operators. CSPs and the need to respond with speed at all levels is just very different from what it was like he was saying, right? Five, ten years back. Expectations were different. Everything is nowadays real time. Hundreds of plans, hundreds of options. Hundreds of promotions have to be done by operators. And thereby I think the need to run with speed is super critical.

And that's where we bring this whole AI centric approach to find out. And I think in the morning, your CEO was talking about using the right AI for the right purpose. I think we are. While I think your team were presenting Blueprint, we are actually living it. We are doing it and we are seeing the benefits. We are seeing the challenges, and we are discussing with you guys very transparently how we can improve this whole collaboration.

So there were, I think, the retention challenge of telcos, the speed of response, challenge of telcos and the need to retain customers are the challenges that we are trying to address through this program. And.

Axel, you've been in the industry for quite a while. You were at AT&T with us. And what what do you see as why this is so important from an industry perspective?

Yeah, one of the things I did at AT&T, I actually managed a team that would build offers, but that process would take, you know, 6 to 8 weeks or longer sometimes. And during that time, the competition was taking advantage of the fact that they had rolled out this new promotion, and our ability to respond was limited by our ability to actually bring a new offer to bear. And during that time, it impacted the value that we were able to give to our customers. It impacted margins because sometimes there would be credits given out or worst case scenario, we would lose the customers from a churn perspective.

And what this catalyst really starts to do is it shrinks that time frame from that very long tail that you, you experience when you're building a plan traditionally, or creating an equipment offer traditionally or even today, you know, all these new capabilities we have through value added services, streaming and and other services that are being brought to bear. This really starts to shorten that time frame. Instead of it being weeks and months, it now becomes days. And that's really what is at the heart of this catalyst and why it's so important.

And just to add to what he said, right. I think I also want to bring a perspective through this catalyst. Interestingly, we are not trying to reduce the cycle time of it, but we are trying to reduce the cycle time for business when it's about creating the concept. The idea to get the approval, the, you know, the margins done, the legal workflows completed. That's where I think the AI gives us unique ability today. And the Platform Pega brings the capability that it can actually help squeeze that life cycle. And then obviously it creates the value overall life cycle. But I think that's where this focus of this particular catalyst is to more look at how do we enable business to move quickly rather than helping it to move quickly. I just wanted to.

Which is important because we know a big part of the cycle time is the business coming up with what is the plan and the offer going to be and what are all the approvals and everything that needs to be done to get there?

Yeah, that's that's exactly right. I think one of the the key pieces I'll extend on that is it's also helping the business, not race to the bottom in terms of the the offer. It is really allowing the business to focus in on what is the right approach to drive the right type of margins, the right type of value for the customer so that the customer remains engaged and doesn't leave? Yeah.

So so, Sanjiv, I wanted to ask you a question on this point. Um, as our team has been working, you know, with the champions, uh, those are the telcos that are involved in the project. Um, one. We have one in each geography. Yeah, exactly. Yes. Uh, and, um, they they step forward. They identified this need for this dynamic plan builder, um, and to make it so they could autonomously adapt to the changing customer needs and, um, to do it without the human intervention that it really takes today to facilitate that.

We see that what we've built is a key next step in moving forward towards that agentic AI and full autonomy that we're talking about today. Can you provide some insight into any challenges you see as this matures and what can ensure you know our success?

I think it's an interesting question, a very important one. The way we have started exploring, and we are also learning, by the way, that a lot of telcos look at customer management, churn management as a segmentation issue, which means they deal with a segment of customers, premium customers, mid income customers, low income customers. It kind of shields the granularity of the customers under that segment. Because look at your customer base of 240 million. 343. 340 million. Look at the customer bases of Verizon and T-Mobile's and China Telecom and so on and so forth. Right.

If you build just a segment model, there is so much granularity of the individualities under that and the way society is evolving. You have a YouTuber, you have a gamer, you have a work from home, and their needs are different. But if you segment them at a different level, you might think that I have applied my right level of NB, but results are not coming because you're not slicing it the right way. So that's the challenge that we're trying to deal with on the way.

And the second challenge and this I have seen more for European operators where there are group models. Right. Because you have multitudes of entities or large telcos, Course, there will be siloed orgs. Everybody wants to break it. Attempts have been made to break it. I think in the morning the CEO was talking about building integrations, you know, modernizing. But we know it does not work to the level that we want to.

And that's where the authentic, you know, agents that we are building, along with the Pega team that is seamlessly stitch together a multitude of these orgs, which is product marketing Insurance care. And I will even ride into, you know, the work Accenture is doing. It can even extend to that larger group, thereby build the right level of articulation and right level of efficiency that we need to solve these problems together. Yeah, absolutely.

So, Axel, one last question on this before we kind of go to just roundtable questions and take questions. How does the Pega technology enable the success with the dynamic plan builder.

Yeah. So within the catalyst we're providing the the connective tissue, the workflow to tie everything together. There's the models at the start that, um, you know, come from uh, work that then are tied into the product catalog where we get the data that is necessary to do the calculations and drive the process forward. Uh, so that we can make the suggestions to the, the business.

And then that's really where Pega plays a part. We've also brought some of our genetic capability into it with the Coach and knowledge, uh, agents, to facilitate communication with the individuals that are on the business side, working on the process of pushing that particular offer through the flow and then being able to facilitate the the final completion of that work so that it can move on to the next phase of this catalyst, which you'll have to stay tuned for in 2026. Keeping us on the edge of our seats as usual. Absolutely. Um, so we'll go. Are there any questions in the room? I do have one. Go ahead. Could you possibly either. Yeah. Could you go to the mic over there?

Yeah. So my question is for for the catalyst that you are doing. Uh, I understand there are a lot of I GenAI models that you are using to do the network orchestration, make it more autonomous way. What is the level of accuracy you are targeting as part of the entire process and how you are ensuring that all these models are effectively solving the problems. Mhm.

I think the there are a couple of things, uh, I indicated that, uh, we also put one track in is the integrity of what we are doing and um, uh, not ending up with perfection, but what we are going to do is, um, make sure that the catalyst is successful. And we defined one metric, uh, it's called the PTEN. It is the 10% lowest, lowest, uh, annoying areas we will find in the, in the network. And then making sure that they are translated to real actions means when we find areas with low throughput and we find areas, uh, having flapping in the network where always network elements are out, Reliability is concerned, um, starting with those 10% and solving them and then seeing the, the impact, uh, with the real customers.

And this is then the calibration. And then we want to go incremental. The P20 we go the ones in a better position but still weak. And we want to really incrementally work through this to make sure that, um, when we take action we see an impact. And then, uh, I think when you usually are above P50, the rest is fun. Therefore, I think the initial exercise with the 10% areas or regions or uh, where, um, the situation is serious, where you don't have 5G coverage, where you have the high interference outages, where we will see mute calls when customers are calling each other. We will find these things, uh, solving them. This will calibrate the system. Therefore, and when the accuracy is not there, we have to improve the performance of the AI.

And you still need to depend on lot of the network data, right? I believe this use case is more on the left side of the network that you are getting.

Indeed. We will um, we will train, uh, the catalyst with data from the mobile network, with usually data from Ericsson, Nokia, Huawei. Um, we have already aspiration to go across. We have already, uh, the enterprise in mind. Uh, when you have Cisco, juniper, Nook and other nodes connected and we will train. The ambition is even with poor data, even with poor data, we should be able to run this. That's the idea.

Okay. Thanks, Sagar. Yeah, thanks. Are there any other questions?

Just let me one comment towards what Sanjay said. I also agree that both catalysts would pluck them together, will resonate to really go the whole journey. Therefore. I think they've been two work streams and this is the first time. We didn't. Know. When we came here. We figured out. This could. Work. Exactly. I just learned 20 minutes ago.

Running quickly down two paths, and now they're coming together, which is pretty awesome. I'll just add one question, because that question gets asked to me as well. You know, when we go to the customers, you know, while we are talking about catalyst programs, you know, how do you measure the success, how do you predict, how are you confident because you are still experimenting, right?

So what we did is we took the data from one of the champions, you know, 5 or 6 champions, historical data of last six months. We created It using our AI models. You know, the outcome, the action. And we predicted based on this what number would have retained. And we compared it with the actual numbers that they have seen. And we see about 70% accuracy as of now. And to that point, you know, I think he was also referring that we need to also now fine tune, train it more and more to bring accuracy higher. That is where the stage is. But I think it is being done with historical data, with the outcome that we want to predict.

In fact, if you go to my chart, we have also tried to take a target number of at the KPI from a retention perspective, from a cycle time reduction perspective, what we want to target, that is what we are seeing the outcomes for.

Awesome. I'll have one more piece on that last question that you asked me, Ryan. And it's really for Sanjeev and Hawken. It's a comment that I'll make around how, um, the SES, our partners, how you really are bringing to the table your knowledge and expertise and using the tools to really drive both of these catalysts forward. With the champions. You play a critical role in the success of both, um, you know, the customers as well as the industry and being able to bring these capabilities to bear.

Like, we could not do this without the expertise that you bring, especially around the modeling. Ben, in his last session, even had a comment about partners. And you have to look at who's specialized in what areas and how do you rely on those partners. And I think this is a great example of, you know, bringing both complements together of skilled resources that understand these spaces. Yeah. Definitely agree. So Ben, everything is Agentic Agentic Agentic does that mean workflow and automation just goes out the window and everything's just an agent? What's your point of view?

I don't. Think so. Maybe one day. You never know. Skynet really lives. On. Um, so. No, I think what we're going to need orchestration, workflow, and all of the technologies we've got today. Really to work together with the agents. Um, you need something to orchestrate. Even the agents. You know, I kind of think about it still, as they're not much different to an API from my perspective. You know, you're calling them as a service to performing a job, um, at various points in a journey, probably a value stream going forward rather than maybe point processes.

Um, but they still need orchestrating. Otherwise they're just randomly doing jobs, and you still need that view of the journey that they're going on. Where is it at in that journey? Who's asked them to do something as something? What's prompted it? And you need that auditability as well around your process. Properly end to end, not just very specific pieces. Yeah. If you can get to a point where you've got an agent doing a whole value stream or something like that in the future, maybe. Maybe somebody. Yeah. But. You know, it's not like the reality is there's going to be lots of these things. Yeah. And you, you need to be able to see for your, your processes and journeys or what's happened along that timeline. Yeah. And the only way you can really do that is through something that's orchestrating that whole thing. Yeah.

As I've been thinking about it, I can remember at AT&T, you know, in some of the more complex areas, it would take us 2 to 3 years to get resources up to speed on actually how to process some of those orders. Yeah for sure. Giving it to just an agent is like giving it to a new hire, right? And it's just not feasible. Yeah. And when the agent fails because it will one day. What happens? Right. It needs to go somewhere else, right? Let's go to a person. And where's it going to go? Yeah. Yeah.

So I think we're coming up on time, but are there any other final comments that any of you would like to make?

I think, you know, to what you're saying. I was going through like, you know, they talk about MCP nowadays, right? MCP is nothing but abstraction, creating some layers of abstraction on top of APIs. End of the day. So there is underlying investment infrastructure and, you know, resources that exist which can be leveraged more effectively using these kind of workflows that we talk about. So when it comes to agents, I think as agents we can scale those automation in AI faster. And uh, it is uh, it is not a threat. It is a it is helping.

And and when we go away from that topic here, we would just look into, um, the education field. We have 5 or 6 billion people and we would be able to educate everyone. The people have never had education versus the ones always been educated. When you would have the capability to educate the entire population on Earth and an agent can be helpful because he can adapt himself to according to who's educated. Right? And you can do it at scale. You could go and educate 2 or 3 million people easier. It will help everyone.

Therefore, I think the is this concept which is also here, um, supported by the TM forum, uh, will help. And and this conceptual architecture where you have an orchestrator agent, you have a super agent, you have an agent. This is so simple. I could explain this to my grandchild in two years. And then then. And then the and then, uh, and then implementing all the ecosystem partners would work on the same standard. It's really helping everyone, um, uh, to, to, um, put the investments effectively and then. Helping, helping especially here of course, the industry we have to help the industry to transform faster. Yeah.

I'm with you. I think, you know, we shouldn't hate agents. Obviously, they're they're moving us towards the scale of autonomy, and there's a lot of value. Going back to what Alan said earlier today is it's the right AI to do the right job. And that's really what we have to figure out is where is it? The right eye, where is it overkill, and how do we apply it to move us forward? Yeah. Yeah, absolutely.

All right. Well thank you. I appreciate all of your time today and have a great PegaWorld. Thank you, thank. You, thank you, thank you, thank you. Thank you.