Airtel reduces customer frustration to near-zero levels with AI-driven automated operations

We are going to hear about AIOps Instrumental Role for Next Generation Customer Experience. I am delighted to introduce our first speakers for this session. Yesi Josetia, the Chief Strategic Transformation IT Officer of PTXL Axiata and Samit Banerjee, the Divisional President of Amdocs Cloud Operations. Welcome. Come and join me. And thank you. We love the enthusiasm. Let's keep it up for the day. Well, so you heard from Michael and myself that we really talk a lot about AI and it's been a big theme of the work that we've got here. Yessi, if you don't mind I'll start with you. What does AIOps actually mean to you? Because I think we talked before and said it means many different things. All right, okay. So, before I start, let me just introduce what Exalaxiata is. So Exalaxiata is a telco company in Indonesia. We've got around fifty seven million subscriber. And we've just embarking our third transformation. And in this transformation, what we are saying is that, okay, we need to have the highest NPS and that is the one that is driving the customer experience. Now the business problem to solve is that we need to create a very intentional customer experience. And that means there is a need to change the way that we measure customer journey. So previously, we have been looking at operations by boxes. So how much is the success rate for each boxes, but that doesn't mean that we know the customer journey or customer experience is addressed. So the challenge that I give to our technology partner is that, okay, then how we are changing the way we look at operations from customer lens. And in order for us to do that, we need technology. And this is where AI comes in because if we are doing stuff still using the old way of doing things, we won't be able to achieve what we're trying to solve here. Absolutely. And it's about looking what those new the new problems are. I mean Samit, for you, how can AMLOCs and how can you really help operators drive those new ways of working? And that's so important because AIOps is a whole different way of looking at your operations. Yeah, so I think first of all it's a mindset, okay, that things can be done automatically. I think as Michael was saying that there is sometimes a feeling that once you get into AIOps as if human being is not needed, as if you do not need any more people in operations. So first, it's a change of mindset that we can deliver things by utilizing AI ops, by utilizing machine learning, by utilizing these capabilities and taking things to a level whereby it is no more IT related KPIs, but it is business related KPIs that we want to meet. So for YESE, what is important is how she is serving the business of Excel across Indonesia, which is for more than seventy million subscribers at the end of the day. For us, it's important that we are enabling that aspect of it by utilizing AI and ML in a way such that it can be transferred to business KPIs. That's important for us. And Yossi, what was that mindset shift like for you guys at Axiart because we often think about KPIs in technology terms and we look at how many API calls we've had, all these types of things. But was that different mind shifts to go to get the technology team to see these are actually business KPIs. We're driving business growth and value. That is a very pointed question, Aaron. I think I actually have a slide, just one slide to actually show how we change the way we look at things. And this is basically a dashboard that is seen by our operations. So before you're looking at the all the acronyms, right? The OCS, RPL, Centimeters and so on and so forth in terms of success rate, latency and so on and so forth. But in here, what we are showing now in the dashboard is really a particular journey. So in here, so for example, how to activate SIM card from customer perspective. And when we measure success rate, it is the success rate of activating the SIM card. It doesn't matter about which boxes that contribute to that. So in that, we need the ability to do correlation. And that is where the machine learning and AI comes in because without that capability, we won't be able to correlate and create the journey. So that's so this is I think the dashboard that we are now seeing as a mindset shift for our operations. I give I have actually two example, but so the first one is that I think this is also something that probably resonates with a lot of telco in the room. So because ninety nine percent of our base is actually prepaid and a lot of them forget to reload and get the card expired. And before we start looking at the customer journey, when the customer complain and saying that, okay, their cart is expired before the recycle number kicks in, they can actually reactivate it. Before it takes for us like eight hours to solve a particular subscriber and they won't wait. They will just look at other operators to get it. And changing number is very easy for prepaid market like us. So with AI ops, now from eight hours of resolution time, we can actually solve it within a minute. So that's actually a massive change when we are moving from looking at things from boxes to customer journey. I mean that's a huge change. Mean really Samit, how do you drive that change? Because that's a big thing, right? It can go down into minutes. Is the tooling important to realize that? I mean again there's a skills gap culture. How do you address that with the work that you're doing? What we are doing right now is we have created a platform which we call an Amdocs Cloud Management Platform. And we treat operations like a product. It's no more like putting people and running systems. It's like more like a product. So what we do is right from the product development that we are doing, which we are implementing, we bring in operations and plug the both of them together. And in that, we bring in aspects, let's say, if it is monitoring. So we will have a single pane of glass where you can see everything together, and then you start to detect the anomalies. And the machine language starts to read that, and you feed that AI engine to start to take actions on that. So the actionable intelligence is picked up and then you start working on it. Now it takes a little more time depending on what are the different ways it should perform the actions, but you can start to apply it whether it is on incidents that are coming in from the customers or whether it is all the multitude of boxes that are being implemented in the ecosystem, the IT ecosystem, and you know that too many people are needed, too many decision points are needed. You don't need it. You can put all of that into the system and let the system start to learn on its own and it will take actions. You can put auto healing in place. So it will start to take actions even on the infrastructure side of the house as well as all the multiple systems that are talking to each other. What actions do they need to take? You have to treat it like a product. You don't have to treat it like an operation. And you can take each one of the sections like whether it is observability, whether it is monitoring, whether it is infrastructure assessment, whether it is cloud management, all of these areas you can start to implement the AIML component of it. By the way, this platform also works not only for the on premise applications, it is also working for the cloud applications and in a hybrid environment as well. So it could be multiple clouds, on premise, all together. I mean Michael, that's interesting, isn't it? If we think about operations as a product, don't think we've really heard that before. I'll take a little bit of poetic license here. We know we've got SAS and we've had infrastructure as a service. We're looking at networks as a service. I'm hearing we're have operations as a service now. You can take it a little bit further but I think the consumption of services through AI is going to become more and more important to actually get that customer experience to move to the customer journey as Jesse pointed out. Absolutely. And one of the things that we often get asked and please if anyone has any questions do raise your hand and Stephen will look and go next year. That's the best way for me to look at this. But how do you measure results? This is the thing I think I get asked every time with CSPs. Can we start? Okay, so I've got another slide for this. So for every, I would call it like innovations or initiative that we're doing, we need to have a metrics on what success look like. And when we are seeing from customer experience lens, we can actually measure it from different perspective. First is that customer lifetime value, so they are not churning. And then in terms of the men operation themselves, right? Because if we were to do this and we need to dimension number of person that is doing this, it's going to be very, very different. So with automation, then there are some numbers here that is showing things are resolved quicker. So that means in terms of operations, number of resource is not does not require that many. Number two, then we can also looking at it from customer lifetime value which is improving the revenue as well. So I think there are a number Some impressive results there, three percent revenue growth, two percent EBIT growth, seventy three percent end user tickets resolved automatically. Correct, because now with machine learning and AI, when a ticket comes up, it will actually go through that learning process and apply the next action that needs to be done for that subscriber. Yeah and Sami, it's very interesting because this is like service assurance use cases. And that's where I see the biggest opportunities at moment for AI ops. Are you seeing the same thing with Amdocs? Are people kind of choosing start with service assurance because people say, Arun, where do we get started? Yeah. No, it's for sure the service assurance. As I mentioned before, that operators and CIOs of the operators, they are always looking at the business KPIs. So business is in terms of services. Okay, what kind of services we are delivering? And you can look at these numbers that we are talking of over here, which we are not measuring for ourselves, we are measuring it for our customers. So if YESE is defining that these would be the KPIs that we want to target, show us how we are going to get there. So our technology stack that we are talking of over here, the operations as a product or as a service, we need to entune that with those business KPIs. But Samit has another KPI, right? So if he's not meeting that, I am actually giving him penalty and surface credit, right? Right. But that's a partnership, right? And how is that changing? I mean, is it going to be things like this that there are these types of penalties and really driving both of you to get the better results? Right, no but that's always good because you need to have a carrot and a stick. You need to have both of them and I think it works. I never give you a carrot. Do we have any questions from the audience? We have one there. Thank you very much. Hold on, the microphone is coming. We can't hear you all the way through here. Please say your name and what company you're from as well. My name is Yaniv. I'm from Netcraker. My question is to Samit by you talk about AI not replacing labor, but still by can you give an indication by percentage wise how much labor you are able to save when you apply those products that Amdocs described in terms of the cost of providing the service to, in this case, Axiata? Yeah. So so what I first meant by AI not replacing labor, it means that we are not making we're not getting into a scenario where people are losing their jobs. It's just that people need to create more products, more solutions related to AI and ML because the field is very, very wide. So you can still have work. It's like in the old days of seventies when people thought computers are going to come and take my job away. That didn't happen for that matter. As far as replacement of the real labor is concerned, there are different use cases that we have, but it ranges anywhere between thirty percent to forty percent in the initial stages and it can go up to fifty percent to seventy percent in the later maturity stages. Great. Do we have any other questions? Over here. Thank you. Mikael Maraval from Orange. I just would like to know how you actually measured that what you've improved at operation level. Did improve also your revenue growth and EBIT growth across other, let's say, initiatives like from the business. How have you been able to prove that? Okay. So before we embark to the initiative, we sort of need to understand first, right? Okay, so this is the business problem. There must be benefit out of this. I mentioned a couple related to operations using AI. One is resource. That means that the direct cost either flat or actually going down. Number two is customer lifetime value because with the market like us where the churn rate is fairly high, so if we can actually keep them longer by giving them an exceptional service, then there is a dollar counting behind that initiative as well. And one thing I want to ask you about this and just following on with, what about trust? I mean Michael you worked in security as well for a long time. Is trust, did it make you nervous with AIOps and trust? Look absolutely. We've got to make sure we've got the right wrappers in place with whatever we do because at the end of the day whatever we do in this new world, security is paramount. We really need to make sure that wrapper is there. So we've got to do all these good things and we tend to talk about the things we're going to do but at the back of your mind, you've always got to make sure you've got to have a security wrapper around So it is paramount and it is really important because, you know, the bad bad people out there are always trying to break in and take the good people's value away. So we've got to put the right security wrapper around all this stuff, otherwise there's trust factor's not there. And if the trust factor's not there, people won't buy into it. Absolutely. So if we can come back to you, Sumit, thinking of that. When you're starting to work with operators on this, is that reluctancy, is it that we're about trust, we really hand over our operations, which is crucial to their business to something like AIOps? True. See, that's why you need to go through it like a journey. So you don't switch it off all of a sudden on one day. You have to see how the system is maturing. You need to see the use cases developing. You start to look at the data and you also go back and manually check-in the initial stages of the implementation that whether this is really giving the right results or not. So yes, are talking about the anomalies that are getting detected in the system and the AI system is taking action on that, you validate. And only when you run through a proper set of validations, you say, Ah, you know what, now I can start to rely on it and now I can trust it. But you don't leave it over there as well. You have to come back once in a while, within a month, to see whether that's performing exactly like the way you thought it or not. So this allows you to create more trust in the system, number one. Number two, it also gives you a lot more inputs to decide going forward where else you want to apply the AI. So that also starts to trigger the thought process. And that's where where we have the AI or the ML related scientists, the data scientists, they can take actions. So they can understand the system, how it is behaving and where all we can put it in place. That's actually a key what Samit is saying, right? So he is actually distinguishing between operations using AI and AI ops So so when I'm posing the business challenge to to Samit, so what I have in mind is actually operations using AI. But when he start to embed AI in operations, then he needs to start thinking AI ops. Yeah. Because whatever changes or whatever data source that is that is coming in as a new set of data, he needs to retune the AI algorithm to make sure that we can actually get the result as we expected in the original design. Yeah. Mean these are really, really impressive results and this is the carat but we can't give Summit all the carat. We need to hit him with the stick again, right? So, I think the big thing that people want to learn, what didn't go right? What have you learned together on this journey? Because it's not easy doing AI operations. I think everyone is aware of that. So what's your biggest learning and takeaway? So that cycle of monitoring, right, because that is crucial. We didn't get this right at first time. I think this is a journey right now is probably already well oiled run more than twelve months, but in the beginning, right? So we need a frequent iteration to make sure that it's running. Because at first you might actually want to design like one minute resolution, it actually not go well, right? I mean the initial cases where you would expect something like what Yossi mentioned, eight hours it used to take or two hours it used to take, and you are expecting results of that to come down, you will not get to that one minute resolution in one day. It's going to take time, number one. Number two, there is also an expectation setting that happens. Like when we are going to cloud, there's a first expectation, Ah, you know what, my cost will reduce dramatically. When we are going to five gs, you know what, I'm expecting fabulous use cases that will come up. It doesn't happen immediately. So there's always that expectation management which needs to take place to know that what's the real time that you are going to take before you can start to see the results. And yes, these are the ultimate results that I am wanting to see. Because as YESE is always challenging us, is very important because only when she keeps challenging us, we'll keep finding the answers for the next and the next and the next. So there's a journey in the whole process. I think this stick keeps coming out, hear it keeps showing up but this is important our IOPs when we think about what it is we're going to achieve. I think managing expectation is critical as we do that. So yes, you've achieved great things, three percent revenue growth, two percent EBIT, seventy three percent tickets resolved automatically. So, now I've got to ask you about the future and why we're going to push forward and push Amdocs forward. What do you want to achieve? What does the future look like for you in the next two years with AIOps? So I think to the operation team what I have been asking them, okay give me zero touch operation, give me an unmanned operation, right? So it might actually be far fetched right now, but I believe with the advancement of technology we can actually have a very light touch operations, right? It's to this product. Back to this product. Samit, for you, what does the future look like for Amdocs in AIOps? Is it something you're committed to? Yes, I mean, I know that it is a big challenge when we have this expectation, but it is not only for ASE, it's even for us. We would also like to get to a stage where it is zero touch operation ZTO, that's already a term that is starting to float around in the industry as you know. But even if it is not zero touch, it is minimal touch kind of an operation, light touch kind of an operation. We will have to get See his time. Will there. Will So so we would probably like to get there. And maybe it will be like those Boeing 787s where you do not need the pilot to do anything because the machine can run on its own, but you need to have the two pilots just to keep a watch whether things are working or not. So maybe that's the future that we are aiming for. Absolutely and that becomes really important managing at scale, what are those processes and protocols are to do. And I'm sure if anybody wants to hear more about this, I know Amdocs have got to stand, you can go and visit and Yessi, know you're going to be around and I've got plenty to say on this topic. It's been my absolute pleasure to have you both here. Summit, thank you very much. Yessi, thanks for your participation. Round of applause please. Thank you.