Ray Le Maistre (00:00:10):
Okay, welcome back everybody. Time for our second debate today, everybody feeding back from being nourished. Our expert witnesses, as you can see, are already on the stage, so let's crack on. We have a large online audience checking out our debates today. So our next session is titled The Practical Applications for Telco ai. I'm Ray La Matra, the editorial director of Telecom tv. I'm going to be moderating this session. Yesterday we held our second annual telcos and AI event. You can watch the sessions online on demand from next week and building on that day of debates. We now want to address the pressing issues facing the telcos with regards to AI adoption and support. What do telcos want from AI and what new possibilities does AI offer? What are the challenges to deployment and how will this impact their organization's, networks and services? So as with the first session, we're going to learn first from our expert witnesses. So could our first speaker please come to the lectern? And that's Rahul Atri. Oh, you're not going to, you're not.

Rahul Atri, Rakuten Symphony (00:01:30):
No, I'm coming.

Ray Le Maistre (00:01:31):
Oh, you're coming. Yeah. Rahul Atri president, the OSS Business Unit from Rakuten Symphony. Rahul.

Rahul Atri, Rakuten Symphony (00:01:38):
Hello everyone. Ray, I was saying, it's always me.

(00:01:42):
Okay. Practical experiences building ai. I think we all are collecting the experiences so far. It's too early if somebody's saying that it's already proving the middle in the field. But again, like everything, I believe we were talking about platforms earlier, I think platform is just a go-to-market strategy. It could be APIs exposed, could be a webpage exposing API models, but in the end it all makes to make business sense. Apple, Google, Tesla, they had a very clear objective of making platforms that an a p is exposed on the app stores how to make money and how to make it as sustainable business. Everyone will come and join the queue if there is money and there is business objectives to be sold together. And I really believe in the AI terms as well. We should also keep it measurable. AI is not cheap. The models are not cheap.

(00:02:36):
Infrastructure is not cheap. Somebody else is making tons of money out of it. If we're not making things measurable, there's no ROI. We should actually take a step back and see if this was an automation problem versus something else which could be easily sold. We don't need to ization of everything. We also learned that we need to very, very fast take the calls between the generic models and the very specific models. Let's term it as large language models versus small language models. So we were very early in the terms to understand that large language models are great to interact, great generic information. You can also play around that. You want to host it publicly or privately, but small language models are the ones you can actually train much faster. Can also set into your devices if you want to pull on the customer experience and other things.

(00:03:24):
And also can go and sit as agents into your edge and your network infrastructure. We're also talking about inferences. In case you have a router, you have an edge open ran node setting out, you can actually take local decisions. That's one of the reason SAPs are not that popular because they will run on open schedulers and open inferencing at the edge and take a lot of decisions. What do we do with the current industry experts? I think that's very important because when we talk about leadership, culture, openness and making something successful, I started my career as product manager. So one KPI, which I give to myself and my team is how adaptive the solution is. How many people are coming online on that solution every day? Adaption is key. You can make diverse platforms or the best platforms out there in the world if people are not logging in solving their usual problem with the platform, it's not successful.

(00:04:17):
So the current SMEs need to start understanding their role is changing. Their role will probably be tuning the small language models, designing rags if they may. But think about somebody trying to optimize a telco problem. It'll always come down to a causation and correlation. There is so much noise we have in our data right now that to filter out the signal is the most important key. If somebody is the SME in that field, let's say transport radio core or probes, you need to understand that it's important. One data platform and RACI to evolve is the key. Thank you.

Ray Le Maistre (00:04:51):
Okay, Nelson. Rahul. Thank you Rahul. Okay, first of seven. So next up is Colin Bannon, CTO for BT business. Colin,

Colin Bannon, BT (00:05:02):
Thank you very much. I spend most of my time in front of customers during the enterprise side of the business, so I'm going to come from a slightly different perspective, but I think there's a couple of truisms that we've learned so far and we're still learning. And first of all is I heard this from a colleague in the us, build competency, not headlines. Having your team prepared to be able to don't try to win the battle of building your own LLM. Let the LLMs fight it out. Let the hyperscalers fight it out. A hundred billion that each of them is spending this year in terms of capital and more than a hundred billion that they're going to spend next year for the next version. You're just not going to win in that area. Where we've learned is a couple of things. Understand models and scoring.

(00:05:56):
Start to understand what data is true to yourself and curate your data. Being able to host that data and keep it private. Don't allow the goose that gives the golden egg be given away and allow other hyperscalers or other technology companies then disaggregate you because you've curated that information and then handed it over to somebody else. And we see that in enterprises. It's very interesting to see how differently different industries are being disrupted currently by ai. The legal profession in particular is one example that is very worried about AI because what is AI very good at is looking at a lot of historical textual data and summarizing and their billable rates are going down very, very quickly. Other areas are looking at it as a godsend, but for the telcos hosting in controlling your own data, but also having a team that gets good at being able to become an expert in tuning as well.

(00:07:01):
These are some of the key things, but some of the things, the scars that we have already around competency rather than headlines is most of the practical applications that we've seen so far. First of all, we've been doing things along ML for many, many years. This is just an evolution of that and it's a means to an end. And what we found is when you try to apply AI hype to certain business processes that you realize very quickly that if you haven't sought out the business processes in the first place, you're not going to actually going to be able to get the benefit of AI out as well. So that's really important. But really in terms of competency and the building, it's still early days, but building a team to do that. But the other thing I'll say on the final thing is for telcos, actually the real thing that I think is particularly interesting right now is the implication on the network itself and being a trusted partner to have an AI ready network. There's been very, very little modeling and understanding of the implications of what AI is going to do on the network. And I think that's a really big opportunity for us as trusted partners to deliver the value and experience of AI as well. And I'll stop there.

Ray Le Maistre (00:08:19):
Okay, thank you very much Colin. Next in line is Warren Bayek, VP of technology at Wind River, Warren.

Warren Bayek, Wind River (00:08:32):
Good morning. Those who know me would tell you that I'm prone to hyperbole, so I'm going to start there, but I promise to bring it back into current day practicality really quickly. So there's no doubt in my mind that what we're doing in telecom today where we're virtualizing the far edge and we're creating this environment where we can infuse ai. And by enabling that with this virtualization, eventually the world and your day-to-day lives will be impacted in a profound and positive way in ways that even the smart people in this room can't begin to predict. That said, we have to deal with the present. We have to do things now to enable that possible future. And we've heard it said, and it was said a lot yesterday, that culture eat strategy every day. And what that means to me in turning the telco industry back into an innovative industry is that US leaders need to push down through our organization, the DNA of an AI first mentality in our workforce and in our people, and I'm not just mentioning this for data scientists and AI engineers, the entire organization has to think of themselves as AI first and AI enablers.

(00:09:41):
And the way I see that playing out from a practical perspective is as leaders, we need to make that education part of their normal and typical job description, not something that people do nights and weekends as an adjunct to their careers, but as a critical part of their day-to-day careers. The other thing, I heard something yesterday that Neil said that really resonated me with current and short-term AI and telco. It's really a game of inches, which means we have to find critical business problems and critical business pain points that the telcos are having and solve them through ai. And we heard yesterday some really good examples of that where we're tackling energy efficiency and we're tackling good use of spectrum to alleviate congestion ideas. And there's some other areas, and I just want to mention another place that telcos in my world at Wind River, we're in the infrastructure world and as we virtualize through virtualized, ran and o ran the far edge, we're creating an insanely complicated architecture, right?

(00:10:43):
You're bringing in disparate architectures, disparate companies to handle the services in a virtual way. So it becomes very difficult to manage and operate that environment. And in one particular case, for instance, an event or an alarm that affects customers happens out here. But the real root cause is something that happened over here and it's rolled up that caused what you see as an impact to a customer. And it can be very difficult to figure out what the real problem is and solve it quickly and efficiently to give the customer back the service or to give them the optimal service. And in a large scale deployment, at one of our virtual telco environments, we used AI to solve just such a problem by pulling in disparate data from all these different vendors and putting all that data imperfect as the data may be into an AI engine, which let them get to the root cause very quickly and solve the customer problem. So that's just one practical place that AI is used by the telcos to create operational efficiency.

Ray Le Maistre (00:11:45):
Okay, great. Thank you Warren. So we're getting lots of different perspectives here on these practical applications for telco ai. Next up is Aaron Partouche, who's innovation director at Colt. Aaron.

Aaron Partouche, Colt Technology Services (00:12:08):
AI isn't just about transforming networks, it's about redefining possibilities. The critical question for our industry is how can telco deliver true value to VA ecosystem? And maybe more importantly, can the AI solution succeed without telcos at court? Back in 2017, court led one of the first industry SA MPLS, greenfield deployment aiming for cloud like consumption model up 200 gig. Across our networks we segment routing. Combined with ai, customer can customize network path based on lat sea resiliency or even carbon emissions while tracking performance in real time and adjusting dynamically based on intent. AI driven applications will demand closer integration with network. Unlike traditional bandwidth usage based on OTT, driven by OTT, relying on CDN and buffer mechanisms, future network must dynamically adapt in real time to meet the future AI application needs through intent-based network. We all are enterprise user or applications to optimize performance dynamically, but that's not it. That's not enough because tech companies driving AI adoption are overwhelmed with requests from telco. So we telco operators need to collaborate to develop application at scale. We need to provide simplified solution that offer access to specific SLA for short duration and stimulus global coverage, eliminating the need to negotiate with each local or regional operator. Coal is contributing to this effort by actively working with the method on a P standard like Sonata to streamline intercar interconnectivity simultaneously. Initiative like G SMAs open Gateway are becoming essential, enabling developer to interact seamlessly with telecon infrastructure.

(00:14:50):
Together. This effort establish a solid foundation for automation and collaboration within VA evolving AI ecosystem. Thank you.

Ray Le Maistre (00:15:05):
Okay, there's that collaboration word again. Okay, next up we have Mabel Pous-Fenollar, who's global head of Digital and zero touch operations at Vodafone. Mabel, the floor is yours.

Mabel Pous-Fenollar, Vodafone (00:15:21):
Thank you. Morning everyone. Let me put the perspective of an operator. So at Vodafone we see AI technologies are and will help us both on the business operations as well as the customers. So let me bring it to the business operations on the field of operations and networks. So in network operations, our vision is based on a transform operating model and capabilities that combine autonomous networks, but also intelligence operations that are using AI and generative AI for those technologies. It is super important that we as an industry use the technology and our partnerships that we are building to move towards those new technologies in automation and ai. But not only that, it is super essential that we do strategic shifts on the operating model, the way that we do things. We've seen and we learned that the new technologies needs to us to operate them in a different way.

(00:16:29):
So it's super, super essential that on top of the technologies and the partnership, we focus in the operating model and the shift. But not only that, the third big pillar that we see as strategic shifts is in the culture, the people and the skills. We really, really see that this is not only about technology, it is about the change in our operating model as well as up skilling and transforming our employees together on this. So we see, as I said in Vodafone that we can use AI applications for the business to help our customers for the technology on support centers, more inner in the network operations. And all of these is a technology but also transformation on how we do things in the industry. So thank you very much.

Ray Le Maistre (00:17:27):
Wow, Mabel's clawing back time there. You came in well under, that's very rare, but again, we're talking here about the development I guess of the AI native telco here and we've heard that a few times already. So our next expert witness is Manish Singh, CTO Telecom systems business at Dell Technologies. Manish.

Manish Singh, Dell Technologies (00:17:50):
All right, thank you Ray. Good morning everyone for next three minutes. Attention is all you need as I take you on a journey to explore the rich landscape of opportunities that AI present to the telcos, and I'll just simplify them in probably four key pillars. Let's start with productivity. I think not only telcos broadly across the enterprises, it's actually already well understood and the results have started to come in the amount of productivity gains that you can have. Let me give you a quick example. In the world of telecom, some of the work we are doing, not very interesting, not very sexy, but network troubleshooting, alright? You can unlock take large PA files, message flows, logs, et cetera, unlock the value of anomaly detection that AI does very well and you can actually boost the productivity of the network engineer to quickly start to troubleshoot the use cases.

(00:18:46):
Let's go next one. In terms of fault prediction, and I must note at this one you have large language models which are not necessarily very good with the time series data. And we've been working on another use case with a different model looking at the time series data, pairing it with again, logs and flows and doing fault prediction at higher accuracy. Now think about if you're running an operator, what does that mean to you? If you can start to predict faults ahead of time when the faults are going to happen, starts to change how you do your workflow management, improve your network KPIs and more. Let me move on from these examples to the next pillar, which is more around customer support, customer service, customer care. I think this is the most obvious one Number of telcos have already getting started on this one. And the value that you can start to unlock with the large language model starts to become much more obvious in this case, especially move from the world of chatbots to chat agents, start to do more integration and automation with your existing O-S-S-B-S-S stacks and more in that domain.

(00:19:57):
The third one, which is not so clear and in certain markets is definitely taking shape, is the opportunity with sovereign ai. And I would especially take you in the world of in Asia Pacific, maybe parts of Europe where number of telcos who are looking at what they can offer in the local markets with local languages, local context in terms of building software and AI capabilities in those markets and provide services to the enterprises, to the governments in those domains. And I closed last but not the least with the one that is most unclear in my view, but is a very, very, very interesting opportunity, especially for the telcos, is what can be done with edge that priceless real estate. Where can you bring AI to where the data is data has gravity, most of that data is going to get created in the real world and can telcos participate in that world by bringing AI on the edge? Thank you very much.

Ray Le Maistre (00:21:08):
Okay, thank you. Thank you Manish. And last but definitely not least, Vishal Singh, senior VP of Global Business Center for cloud and network services at Nokia. Vishal, you're going last. It's the prime spot.

Vishal Singh, Nokia (00:21:23):
Thanks a lot Ray, and well almost good morning, good afternoon to everyone. We have been experimenting with AI ML in telecom industry for actually quite some time. We've been playing with that and recently with generative AI also, we've been doing quite a bit of work in the industry. Actually. The challenge is not about what AI can do. The challenge is how can we enable AI as a technology to deliver business value. The use cases are consistent, obvious customer experience, churn prediction, anomaly detection. I think our panel has talked about it. Our ability to actually look at the network and the network data because so much of data gets produced in the network is mind boggling. And our ability to really take that data and maximize the value is where the challenge is. Unless we actually look at a framework which actually helps us normalize this data and stitch this data in a way in which it can be consumed by applications, it is very hard to create value.

(00:22:39):
Therefore, our challenge really is to look at a system which actually helps us to take this fragmented data, normalize it, and bring it together in such a way that not bound applications can consume it and deliver the value wherever it is needed to the businesses, to the consumers, to the enterprises, to the developers, as we talked about earlier together the day. Now once we do that, the telcos can actually drive decision making and business models, which can then really help them solve certain fundamental problems. Things like simple things like reducing the operating cost, improving customer experience. How do you make network more efficient? These are the problems we'll be able to solve a lot more in a predictable manner and therefore create tremendous value for everybody across the businesses. And that's where our focus is. That's where we are looking to really create more value as a business, as Nokia and work with our ecosystem and our partners. Thank you.

Ray Le Maistre (00:23:46):
Okay, thank you very much Vishal. So we've heard from our expert witnesses, but what we need now is a great telco debate motion for you debate on so over to Chris and Graham who will introduce the motion and then present their arguments for and against. Thank you Ray Graham.

Graham Wilde, BWCS (00:24:05):
We do have a motion. Thank goodness for that. The motion is, is it on the screen? Yes. The motion is Telcos will only benefit financially from AI when they deploy their own customer focus models. And this time Chris is speaking for the motion and I am speaking against,

Chris Lewis, Lewis Insight (00:24:24):
Thank you Graham. I have no food analogy to introduce to the discussion. Many, many years ago, actually it was 1980, you missed in Manchester, my degree was in computational linguistics modern languages and we were trying to translate from Russian to English, English to Russian in the opening saying we put in, the spirit is willing, but the flesh is weak. What came out was there's plenty of whiskey, but the meat's off now in order to get there, that took four years of work, right? That took four years of studying, of trying to understand programming languages. It was time was prologue and c plus plus. And trying to get on the old batch card system to push that through, there was a hell of a lot of work. I'm delighted to say that many of the models today actually do deliver that process very efficiently. But the point with AI is that, and it always worries me when we talk about an AI native telco or everything's ai, AI is a tool that we will use at various stages within the telco organization, within any organization, within our homes, in our own lives in the future as well.

(00:25:30):
What I think is important is that if we focus back inwardly at the network, we miss the whole function of what the telco is supporting in that ecosystem discussion we had in the first debate by moving it along towards the customer facing side of it and making sure that all links work all the way through, that's when we are really going to create value and deal with the customer. As Manish just said, the potential of AI at the edge. And if the edge means the edge, and I think the way you are defining it, manishh, but also out onto the devices that we're going to be using, the telco needs to understand where all of those pieces fit. The models, the language that we use has to be very focused on the way we as users want to use it. Whether it's dean's, developers, or whether it's me as an individual or whether it's telecom TV as a business or as a society as a whole.

(00:26:20):
What we do have in the large language models have created a fantastic platform. Can we use that word opportunity to actually be able to give us access where technology as one former CIO once said is for the first time starting to speak our language. And I think that's a really important thing to keep in mind that we're not basing always back towards the technology language of the past. We are moving to the language of the user, of the future, of the customer, of the future, and only adapting by taking the scale of the LLM, moving it down the SLM into the more focused business. And then in a telco context, understanding all that data that was just being talked about as well and using the language of the customer to be able to interpret the data that goes on within the network to deliver the service that the customer actually wants.

(00:27:09):
The negotiation between you and I as customers with the telco or whatever the telco is in the future actually is going to be in the language that we want to use, will give us the best option for us and the way our business or our households want to work. And it will be done in the language of the customer. So we're gradually narrowing it down. Let's say we're moving from the LLMs to the S SLMs using the rag and all the stuff that we've talked about from a technology point of view, it begins to talk the language even down to producing the brand of a telco and the way Colt wants to talk to the business or BT wants to talk to a business or to its customers and then actually focusing it down and delivering it in the way the customer understands gets the best deal and actually supports the whole ecosystem. So I would encourage you ladies and gentlemen to go for the model because by delivering that financially, the telco will benefit Graham.

Graham Wilde, BWCS (00:28:01):
Oh Chris. Okay. Well ladies and gentlemen, this motion is clearly hogwash and I urge you to vote against it when you get the opportunity. Lemme tell you why. So Chris told you a story about his time at university. I'm going to tell you a story about a meeting I was in about 30 years ago in Hong Kong where the CEO of a company, I was doing a sales pitch for some telecom technology or whatever and the CEO EO of the company at the end of the meeting said, oh, one other thing he said, the internet. And I said yes. He said, well what's the business case for it? And I was like, oh, that's a really hard question actually. And when I thought about it, I said, I probably didn't answer the question very well at the time, but on retrospect what I should have said was, well actually it's an umbrella term for a suite of technologies and those technologies will play in virtually every part of your business.

(00:28:55):
And to the extent that that's true, we don't talk about an internet native business, do we anymore? All businesses have the internet and they use it in different ways. And the same way they're talking about an AI native telco, we'll just become completely obsolete at some point in the future. So AI is a suite of technologies like the internet. Some of those technologies are large language models and they have a great role to play in customer focused interactions as we all know. But many of those technologies are already in assistance. Many AI technologies already exist. They're not large language models. They're being deployed in your networks right now and you are benefiting from them right now. So in that sense, this motion is wrong. Okay, just to give you one example, just one tiny example. So I work for Hutchison. We own 11 mobile operators around the world. We also deploy private mobile networks. We partnered with Nokia. Nokia has AI built into its technology for private networks in the radios and so on. So we are benefiting from that. It's not customer focused. Noia doesn't know who our customer is really, but it's optimizing that network and it's making it more efficient without us doing anything, we're already using it. So this motion is complete crap. You should vote against it. Thank you very much.

Chris Lewis, Lewis Insight (00:30:27):
He always resorts to abuse when he doesn't order food.

Ray Le Maistre (00:30:32):
And I noticed Chris, you said you weren't going to talk about food and then you instantly mentioned I did, but the meat was off. The meat was off. So there we have it. We've got lots of expert witnesses on the stage with lots of different perspectives about how AI can be practically used by the telcos, but we've also got the motion, let's keep the motion in consideration here. And it's all about benefiting financially from ai. So you've heard the arguments Fawn against. Do we have any questions from the audience to kick off for our expert witnesses? There are seven on the stage here. So we have a question here. The microphone is just coming over. If you can say who you are and address your question.

Robert Curran, Appledore Research (00:31:23):
Hi Robert Curran from Appledore Research. To what extent do you think that the upside is limited to the data that telcos already have access to and what's the extra information? I'm thinking in terms of for example, the nature applications that S may not currently have access to what users are actually using. I think end user customers are not just thinking about connectivity that's connect taken for granted. They're thinking about the applications they use. So to what extent is, as I say, the upside for the telco is dependent on getting new access to new information that they don't currently have. Certainly in terms of creating value as opposed to just taking network data you happen to already have access to already.

Ray Le Maistre (00:32:03):
Okay. Yeah. Okay, we'll come to the end there.

Vishal Singh, Nokia (00:32:07):
So simply speaking today, yes there is a lot of data in the network but it's not really accessible to or for that matter, there's not enough intelligence we can get out of that data because it's all fragmented, it's all vendor specific. It lies into the data lakes and we all know that there's not much we do with it actually speaking from the network data as we speak. So there is a lot of upside by actually really making that data available for applications. And applications is where the innovation would be, where you'll actually use a lot of AI or for that matter, the large language models and other capabilities. So the upside is at this point in time is massive. And if I may also say that we know we as an industry actually kind of missed the bus on cloud and security. Artificial intelligence is one area, especially generative AI is where the entire ecosystem of all different industries are at the same starting point. So it's our opportunity actually to step up the game because we have the access to data but we just don't have the way to consume the data. The data is still not the best form and factor to consume. And we talked about in the previous panel about APIs and developers and everybody else. There is huge potential if we make it available, accessible and apply the language models that we need to and especially also the industry specific language models. So there is a tremendous a Sorry, go ahead.

Ray Le Maistre (00:33:37):
Okay, I think we've got a few people, but I just want to follow up there. Vishal, do you think that, because you talked there about the existing data, but I think Robert was also asking additional, do we need even more data to make the best use? I know Mabel wants to come in on this and so does Colin. In fact, I think probably everybody does, but do you think there's enough data there already for the telcos to be able to develop profitable revenues from that and they just need to tap that or do they need more?

Vishal Singh, Nokia (00:34:05):
See the data is getting generated in humongous data gets generated in the network massive. And if you put that, so if there's a network data, there is interaction data through CRM and other channels, the digital channels and there is transaction data. So if you triangulate the interaction transaction and the network data and this data is continuously getting refreshed, new data is getting generated. So I don't think it's about more data. The data will always get generated. The question is what are we doing with it? The opportunity lies there.

Ray Le Maistre (00:34:33):
Okay, I going to come to Mabel next then Colin and then Rahul, but I'm going to get to everybody in the end.

Mabel Pous-Fenollar, Vodafone (00:34:39):
So in my view, when we are already exploring with some scenario, so we have a lot of internal data, but we're already exploring crowdsourcing data. So I dunno how many of you know about the down detector public data that tell us when customers see that there is a problem in the network. So we're combining internal data with down detector where customers say, look, I have a problem in here to really see because sometimes the systems see a thing and the customers experience another thing. So we really, really see that there is enhancements of data outside our field that can help us to bring more value to our customers. And we use non detectors where there is a lot of data in some of the countries that use the detector a lot to understand where in our mobile or in our fixed networks we have problems and we create alarms and check on the system. So my answer is yes, we need outside our field and we're exploring some. But yeah, it is key that we work with certain partners to see which other crowdsourcing data or data we will use.

Ray Le Maistre (00:35:51):
Okay. Colin,

Colin Bannon, BT (00:35:53):
Have you read the terms and conditions for any AI apps recently on any of your mobile phones?

Chris Lewis, Lewis Insight (00:35:59):
I read no bts. Yeah,

Colin Bannon, BT (00:36:00):
Look, we take this really seriously right now, trust particularly for enterpris. Half of our enterprise customers have put their AI teams in their risk organizations. For one, I think I am going to be a bit of a cynic here, the role of the telcos, one of the monetization aspects of this is protecting our customer's unique data that is imperative to their differentiation of their business. And ultimately, we'll probably the ones with the kill switch, if it all goes bad, right?

(00:36:38):
If this stuff gets away from us, the telcos will be part of that role to protect. But in the short term trust and protection of that data is key from that perspective. And we see so much data though more and more of it is obs ated on the network and it's not our role. We have to be really safe for our customers about privacy concerns on the network. And as a partner for our customers, we got to be really careful around what data we monetize versus what we protect for our customers and what they're willing to pay value from us. So I turned it around the way as who are they going to trust more and what is the role of trust in this area of ethics, sovereignty, security. And I would argue that the telcos need to find to parlay their role to continue to be the trusted partner on privacy more than ever.

(00:37:38):
Because if you think about ai, it's still scaling, right? And you look at the trillions that are going to be poured into this over the next couple of years and the thesis is the more processing, more power, more time and more data you throw at it, these models continue to scale on the generational AI side and that data is gold, but it could be incredibly disruptive to our customers. And if you final word on this, if you think about law of unintended consequences right now, half I would say of customers that I've been recently speaking about are thinking about getting rid of BYOD and you would think, well why is that? And it would surprised me as well, and this is a non-scientific number, it's just a set of enterprise customers. It's because of the next generation have been watching social media and they're watching videos on how to get ahead and summarization and note taking is really important. And because the enterprises haven't rolled out a summarization tool yet, they're actually installing an app on their phones that will listen to the conference calls and actually then summarize those notes and take them away. And you've got CIOs around the world going, hang on a sec, where is this data going from our internal conversations on that perspective. So there's unintended consequences landing right now and what we need to do is to help our customers monetize it but in a safe and trusted way.

Ray Le Maistre (00:39:14):
Okay, thanks Colin, Rahul.

Rahul Atri, Rakuten Symphony (00:39:16):
Everyone has data. The operating system of this device has your data. The OTTs like Netflix have your data. I don't know what we are protecting on that, but telcos have the data. Robert, to answer your question, the DPI does have the endpoints the customer is trying to access. They know which application, which webpage the customer is accessing to. So I think telco has more than what data is required, have so much noise, we need to find the signals and take the right calls to make it, let's say get more revenue or monetizable. We know what customers are acquiring. The challenge is the decision makers are sitting in silos, the packages, the products we want to take to market versus how ready the network is versus what the operating systems and apps and customer endpoint applications are running. All these people are sitting in different places and as telco we want to sell is more connectivity and more data and other things.

(00:40:11):
Until we have this mindset, things will not change. Money will not come. Also, I think these monetizable use cases are more about patents and ML rather than ai. Nobody's going to switch on the button to say AI recommend the packages or products to customers. But you can actually customize to figure out the patents to say this kind of users in this belt in this regions are social media junkie or bloggers and you can create capabilities and monetizable assets to them. Then also think about what the future telco would look like. What they be a super app like e and i ET group. Now Rakuten has ecosystem company. So think about how do you become one application which can reduce all the touch points in customers every day. You'll own the data, you'll know that okay, this customer is going to my application one, use case two, use case three from my connectivity pipeline.

(00:41:03):
I think till the time we are just thinking about how to sell more connectivity, that mindset will not come. And also I don't think there's a data problem. There's a picking out the signals, you're picking out the wrong signals and you're not collectively taking decisions as let's say the sales network, product enterprise and others. And customer is smart. I mean for example, to the point earlier, zoom already launched the AI note taking. App teams have it, people are going to use it. If not they're going to download the application and make use of it because in the end it's solving my problem or do I save my meeting notes and going on and on and on. The world actually dumped everything on OpenAI and they're getting access to data. Somebody is training those LLMs and the data is out there. Meta is earning huge. How are they training LLM models, which they're rolling out with open source model. So I don't think it's a data problem, it's a signal problem. And how do you construct that mindset to say everyone in the organization, ultimately it's about monetization. If you're collecting data, tell me what are you going to do with the data? Otherwise only the infrastructure provider and cloud providers are getting richer.

Ray Le Maistre (00:42:15):
Okay, Manish then Aaron.

Manish Singh, Dell Technologies (00:42:18):
Alright, just a couple of things. Number one, I'll just start by saying good data and presence of other good data becomes great data. And there is part A is just within the world of telco, we do have a problem that we need to look at and solve is how do we break data silos because there are datas that are sitting in different parts of the organization that don't even see each other. And there is reasons for us to think about how can we break that data silo and bring those data together to really unlock the value of data and do that completely to the previous comment that was made in a trusted secure way. Because yes, there'll be i I data as well, you're talking about customer data there, regulations, GDPR and others that also come around it. So part A is that just within the telco entity, part B is bringing external data and enriching with external data.

(00:43:16):
I'll give you two quick examples of use cases that I already know today that telcos are working. Number one, in case of fiber trenching and planning using not only their own data but external geospatial data, looking at that, pairing that understanding where the bridges ravines, et cetera are and where can they go entrench the fiber, bringing all of that external data into it. Similarly, another example where some of the larger social, I mean larger over the top platforms would share their application experience data with the telcos and that data becomes very valuable because without again disclosing any PII and also you can always take care of that part, but understanding where the network congestion et cetera is and that paves the way for better site planning, better network planning, et cetera, et cetera. So yes, external opportunity definitely is there with external data sources.

Aaron Partouche, Colt Technology Services (00:44:11):
Okay,

Ray Le Maistre (00:44:11):
Aaron?

Aaron Partouche, Colt Technology Services (00:44:12):
Yeah, so agree also that we should firstly understand what we can do with our network data. There is also, and considering also the sensibility of our customer data, which for me it's special. I mean I would potentially be very, very careful about it. I agree with Colin on that. But network data and after, how we can correlate with external data on that with Manishh, I mean I can give you two examples of how we could use external data to correlate with our network data. One is we are working on the pilot to show the carbon emission on each route that we are providing to our customers or that we need to have an external data source. A second example of external data that we are working on is we can potentially, if we look at fiber, the fiber optics and the data from the fiber optics could start to notify if there is a tsunami using information from optical fibers correlate with external data, for instance temperature. So you're right correlating with external data but considering really data that we can use could help us, could definitely help us. And we need to look at that.

Ray Le Maistre (00:45:45):
Okay. And Warren, did you want to complete the set? I mean there's a lot of ground being covered there. I think

Warren Bayek, Wind River (00:45:50):
We're only going to get through one question today. Yeah, so I agree with the panelists that there is a tremendous and insane amount of data locally within the telco networks today. And understanding the safety and protection of that is incredibly important Colin, and as we go to external data that becomes even more important. So I'll leave it there. But to go right to your question, which was about external data, are there other sources we can get to? I think there are, it's an interesting question of what telcos are going to do with that and how they're going to monetize it. But I'll give you one example from our experience, our parent company Optiv is if I look around this room, everyone either it's on the table in your pocket, you have a phone that's connected to the telcos. So they have some data there. Well I bet almost everyone in this audience also has a car and some of us have maybe multiple automobiles, right?

(00:46:48):
Well those are becoming connected as well. And our parent company AP is heavily in the auto industry and they're deep into this transformation where automobile companies are becoming software companies. If you listen to the Wall Street discussions of the CEOs during their quarterly calls, almost all of them mention we are going to have x billion, tens of billions of dollars of software revenue in a very short period of time, going from zero to that. So car companies understand they have to become software companies to stay successful. Why is that important? Well, your phones are connected to the internet, you know what else is connected to the internet, your cars. So in the next few years as this transformation in that industry happens and cars become connected devices, there's suddenly this another enormous source of data that telcos are going to have access to. Now protecting that and keeping it sovereign critically important, keeping it tight and controlled is important because like telco's, automobiles are mission critical, right?

(00:47:52):
You do something wrong in a car and there's catastrophic consequences. So yes, the safety is a very important part, but there's a tremendous amount of data that can be shared between say as the cars become a rolling data center, that's another source of amazing data that can be used in a way that we can't imagine today. And how telcos monetize that is a great question. I know that someone will ask that or they should if they don't in the end, I have a phone, I am not going to pay a telco money extra money week to week. We talked about this before manam, we talked about this earlier this morning. They're going to have to find other ways to monetize this interaction with external data. And it's going to be through partnerships, right? We're seeing it right now where we have discussions, the car example to continue that on. We have examples where the car companies can do things that people will pay for and the TCOs have to find a way to use their interfaces or their access to their data to monetize those external data sources.

Ray Le Maistre (00:48:51):
So there's obviously an awful lot of data internally, externally as a great opportunity to use that data. I want to bring the panel now back to the motion because, and the motion is telcos will only benefit financially from AI when they deploy their own customer focus models. I want to hear from the panel now starting at the far end and coming down this way, are the telcos already benefiting financially or do they need these customer focused models to be able to really make money from the combination of AI and all of that data that they have?

Vishal Singh, Nokia (00:49:37):
I think we'll have to really look at this slightly differently. If we keep doing things on our own, we'll continue to be in this world where telecom industry will be in the state. Where we are today, we'll have to really leverage everything that is out there. And we talked about collaboration, ecosystem partners, APIs, everything. So I don't think so that there is any need for telcos to build their own customer specific models. What telcos do have, and we talked about the panel entirely discussed today, the different opportunity in terms of data and ability to monetize that data is about actually harmonizing that data and making it available for consumption. Because datas are sitting in silos, datas actually has got tremendous amount of privacy requirements. So there is a lot to be done and the industry out there is available. If you look at the banking sector or financial services sector, they did phenomenal job with the entire FinTech industry. There is an opportunity in those areas where telcos can actually create businesses on top of their business, which can leverage what they have. And customer models will be dependent upon the businesses that they're serving. Correct? It's really how many customer models will build, how many telcos will build their own customer models. I don't think so. That's a scalable model. We'll again, get caught trapped into this so-called wall garden mindset that we've always, we need to open up, we need to explore, we need to really embrace the industry at large.

Ray Le Maistre (00:51:06):
Okay, thanks Al. Manish, what do you think? Is the industry already making money from this?

Manish Singh, Dell Technologies (00:51:13):
Alright, I'll break the emotion in two parts. Make money and build your own models. Let me start with the build your own models. No, I don't think so. The industry needs to build, and I've heard about the telco specific models, telco specific LLMs. It's a very, very, very, very expensive proposition to even start to think about it. Take the foundation models, there's a good set of that available, fine tune them with your data. You are going to find that that becomes more effective rather based on your use cases. You actually don't even have to go to large language models. You might actually find smaller language models. Fine tuned will be very effective. So that's part A in my view. As far as the model is concerned, part B make money, it is absolutely clear from a productivity gain. The opex line I would say it's very clear AI is going to impact and telcos who get left behind will actually find themselves in great danger in my view because you have to really unlock the value of all that productivity gains. And last thing I will say is this industry has talked about autonomous networks for a long, long time and I think with the rise of agents that autonomous network is now within reach. And think about this, if Waymo can get a driverless car driving around the busy streets of San Francisco and solve that, is solving an autonomous network harder?

Ray Le Maistre (00:52:43):
Okay, Mabel. So benefit financially, obviously it could be taken a couple of different ways. So do we need these customer focus models in order for companies like Vodafone to benefit financially from ai?

Mabel Pous-Fenollar, Vodafone (00:52:59):
So if I go back to the question and I think the key part of the question is only so I will say no. So first of all, this technology is not only bring in financial benefit, we heard it is all about productivity, about customer experience. So I don't think it will only bring new monetization. And then if I move to the only build our custom models, again the only part is what I'm hesitant. So I don't think it will only, so as you said, there is large language models, there is companies that didn't have a skillset on those things. They build models, they have large language like large CPUs, we don't have those capabilities so we need to leverage from what it's already built. And they, in the technology industry, there is tools about domain specific models and we might need them for certain parts and certain use cases, but it will not only be used in custom LLMs.

Ray Le Maistre (00:54:13):
So potentially supplementary but not the key to it. Okay. Aaron, what's your view

Aaron Partouche, Colt Technology Services (00:54:21):
Today? I mean AI is mainly used for productivity gain in most of industries and this is what we are benefiting also in telecom. So it was well explained by almost everybody in the panel. I mean we can troubleshoot quicker, we can provide, we can provide services, we can do some predictive maintenance. So a lot of use case. So we are already benefiting from AI because like most of the industry, how we use AI now, what kind of role we will have today in the revolution of AI is for me the really key question, not sure we have to create our own language model. It's probably not where we will be the most relevant. But understanding that these guys that are developing AI application will need telco more than before, more than the OTT, but found other ways they deployed CDN, they deployed buffer mechanism. Now they will need us. So what can we provide to them? So trying to understand how we can provide really the capabilities and simplify their life will be, and I would like to link with the previous discussion, the platform discussion is we need to be in the head of a developer and we need to provide what they need. They try to solve the problem they don't try to buy. So this is how I see our role. And so definitely no for the question, I don't see interest to develop our own model

Ray Le Maistre (00:55:58):
Warren.

Warren Bayek, Wind River (00:56:00):
So not to be too blunt, but to go right to the question. I mean demonstrably the answer's no, right? I mean it's been proof I'd make someone happy. It's not a theoretical question. The answer is no. Telcos are doing it today without customer specific models, right? We are saving money as people have mentioned, we're saving money with operational costs, we're improving service to customers. So money is being made and saved today. Now that said, I'm going to piggyback on what you mentioned earlier, which is, and Manishh touched on it briefly with his 0.4, that as we go to future use cases at the far, far, far edge where devices are involved, I do think it's going to benefit to have specific, they're going to be not even small, tiny language models that are very specific and really inferencing. And that may not be from telco specifically, but they would enable, maybe enable special use cases to be used on their devices or on their networks.

Ray Le Maistre (00:57:00):
Okay, great. Thank you. Colin.

Colin Bannon, BT (00:57:02):
Look, telcos, we feel the pressure we feel from public shareholders but also from our customers. All our customers, they don't have a plan B anymore, right? You go back five or six years, your credit card company, if the network went down, you could get the piece of paper out, put your card on it, run it, and check like that. We are so aware of the soles in the planes that are going on that if the network goes down, the power goes down. If the network goes down, the DHL truck that's moving the insulin between the two hospitals doesn't work. What we need from AI is to help us deal with the incredible complexity of the systems that we have to improve operational resilience so that the trusted services that we provide are there for greater good, for connecting for good. And if you go back to the heart of it, it's not there yet.

(00:58:00):
But if we can leverage it in a way, not just to improve our unity economics, but deliver meaningful outcomes that are uptime and the protection of the realm and the safety. Because we really haven't talked about AI in terms of the persistent, more sophisticated cyber threat that is coming from AI as well. I can promise you the people who are using AI better than anybody else is the bad guys right now. And you need responsive set of platforms that can checkmate as you go through this asymmetric warfare as we go along from that perspective. So I think there's some real meaning behind this, not just the sort of, and I agree with you by the way, it's a means to an end, but what we're looking for is more powerful tools against the incredible responsibility that we feel every day to deliver meaning for our customers and safety and trust.

Ray Le Maistre (00:58:53):
Okay, thank you Colin and Rahul. Finally, economic benefits, do we already have them? Do we need these small models to,

Rahul Atri, Rakuten Symphony (00:59:00):
I don't think so. Somebody will become the hugging face for telco to host the various models So far, I dunno how many of them have taken AI to production. So we have not technically hit the boundary where we need to produce our own AI models on the financial stuff. I think we are very poor in terms of measuring otherwise, but saving money is possible and there are measurable counters and matrices for that making money. We don't even let our closed loop systems make any changes. There's no energy saving production releases for radio nato. We're so scared of letting the software take action. So best of luck with that

Ray Le Maistre (00:59:39):
Same. So we hope there's some value being created, but we're not quite sure how to answer it or what to do. Okay, so we need to end this part of this session and move to the vote. So here's a reminder of the motion. It's on the screen. Sorry mate. Here we go. So Chris was arguing for the motion. Graham I think is feeling confident. He was arguing against,

Graham Wilde, BWCS (01:00:05):
It's in the back.

Ray Le Maistre (01:00:06):
It's in the back, right. Well, let's see. So get ready with your paddles and remember green side forward, if you want to support the motion red side towards the stage. If you're against it, please vote now look at that. Okay. I would say it's a 95, 90 8% against and a couple four. So I think there's a lot of

Chris Lewis, Lewis Insight (01:00:31):
Naive people in the audience. The motion, are we going to go through a day where every motion is going to be completed? We'll see that. If that's the case, Ray, that's fine.

Ray Le Maistre (01:00:32):
Absolutely, absolutely keep, that's right. So thank you everybody. Really interesting session there and I'm glad we got to the money point because at the end of the day, that's what keeps this industry going. It's networking time. Again, we have 20 minutes to do whatever you need to do. We'll see you back here just before 12 for the final debate session of this morning. And a big round of applause for our.