Guy Daniels, TelecomTV (00:24):
Hello, you are watching the AI-Native Telco Summit part of our year-Round DSP Leaders coverage. And it's time now for our live Q and a show. I'm Guy Daniels and this is the first of two q and a shows. We have another one at the same time. Tomorrow it's your chance to ask questions on AI and its use within networks by telcos. Now as part of today's summit, we featured a panel discussion that looked at the benefits of becoming AI-Native. And if you miss the panel, don't worry because we will rebroadcast it straight after this live q and a program or you can watch it anytime on demand. We've already received a lot of great questions from you, but if you haven't yet sent in one, then please do so now and use the q and a form that's right there on the website.

(01:24):
I'm delighted to say that joining me live on the program today are Manish Singh, CTO telecom systems business at Dell Technologies. Rahul Atri, president OSS business Unit for Rakuten Symphony, Danielle Rios, acting CEO of Totogi, Patrick Kelly, founder and principal analyst of Appledore Research, and Scott Cadzow, chair of the TC on securing artificial intelligence at ETSI. Hello everyone. Thanks so much for returning to answer viewer questions and let's get straight to our first one. And the question asks, what are the key attributes of an AI-Native telco? What defines an AI-Native telco? Danielle, let's maybe come across to you to start us off. Is that okay?

Danielle Rios, Totogi (02:26):
Yeah, that's totally fine. I always like going first and breaking the ice for everyone, but yeah, I think it's very short and sweet here on becoming an AI-Native telco. It means that both internally and externally and everything that you do, you try to solve it with AI first, right? And so that could be creating a board deck or a presentation and email solving a problem, doing analysis, attacking that problem with AI. First is part of being an AI-Native telco. It is permeating everything that you do all through the culture, all the way down in every single job. And so that's my secret to becoming AI-Native.

Guy Daniels, TelecomTV (03:10):
Thanks very much, Danielle. Nice to reveal secrets on the show. We want more secrets please. Patrick, what's your take on what it actually means and what it constitutes? What defines an AI-Native Telco

Patrick Kelly, Appledore Research (03:23):
Clearly reusability, so reusing core AI capabilities. I think the other key attribute is to reposition your technology talent. So bring together your domain experts together with the data scientist. And so when you're going out and you're defining the high value use cases, you've got those two groups kind of teamed up. I think the other thing that's important is you got to treat AI as a true product set. So you'll need to have product managers for AI subsets where you're solving specific problems. And then finally I would say establish ai AI labs for quick experimentation. This allows you to essentially fast track anything that's successful if it's not and it doesn't prove the business value out and you just move on. So those are sort of the characteristics that I see guy.

Guy Daniels, TelecomTV (04:30):
Great. Thanks very much Patrick. It is always good to go back and retest the basics here, the AI-Native telco, but what does it actually mean? And we do get a lot of questions at that level of what it actually constitutes, but some of the other questions we've received from you going to more detail in more different areas. So let's move on to our next viewer question then. And the question asks, until now, telcos have been adapting the technology developed by others like Kubernetes, et cetera, and trying to adjust telco solutions on top of these adapted and adopted tech stacks, which is why says our viewer, which is why telcos have problems in being cloud native. What do you suggest to the telcos in the field of AI-Native? Should they build their own technology stacks or should they adapt them from what's out there in the market? Interesting question. I'm sure we're going to get a lot of responses on this one, but Scott, am I able to come across to you first and your thoughts and advice?

Scott Cadzow, ETSI (05:39):
My short answer is you cannot be AI-Native, borrow other people's kit. I mean it's like wearing the wrong clothes. If it doesn't fit, it doesn't tailor to fit you, it doesn't look right, it doesn't fit right and you've got to move on. You've got to start thinking about what is our entry about what is our entry for? So taking an off the shelf product, an off the shelf service, an off shelf capability and trying to make it fit means you're always going to be a hybrid, you're always going to be a Mong girl. You're never going to get the ideal fit to your situation. So I think yes, you can start to play with off the shelf and other piece post kit. Ultimately you're going to have to go your own large part of that will come down to my domain, which is standardization, bring it together, collaborate, work together, get industry native capabilities, and then build those into your AI-Native organization.

Guy Daniels, TelecomTV (06:35):
Great. Thanks very much Scott. So that's our first response, tailor it to the telco's needs there. Rahul, let's come across to you next please.

Rahul Atri, Rakuten Symphony (06:45):
I love the word Scott just mentioned industry native rather than AI-Native to the first question as well. I think the industrial still finding out ways to define your native telecos one way is to say that you use AI into your daily routine and culture, but I strongly believe that it's not about who build the technology first, but how efficiently you or we are using that. I'm a product manager by profession and open source does help a lot. I mean who server developed Kubernetes, it's Google, but I think it took a lot of us going forward, the application stacks or the telecom networks going forward, and I don't think technology coming from anyone is the entry barrier or probably an outlook to look at. We flourish faster, we adapt faster if we go open source and harden it. I think that's how the community builds and a lot of the products incon out there actually started with an open source enterprise model.

(07:46):
Let's pick up the best if you are not the builder. And it also come back to the organization structures and what you want to be. You want to be build versus buy. And if you are an organization love to make products internally, it's not a bad move to build products because whatever you build as an asset in AI and the complete stack from infrastructure to the models to use cases, you'll be able to resell and reconfigure and have multiple go-to-market opportunities. But in case you get the best of the breed or the customized stack together with partners, not a bad strategy.

Guy Daniels, TelecomTV (08:21):
Great, thanks Rahul. It's a good question. This one, Manish, let's come across to you as well. It comes down to build or buy. Should tokos build their own technology stacks or should they adapt them from what's been developed elsewhere?

Manish Singh, Dell Technologies (08:34):
Yeah, guy, the classic build, buy or partner question, isn't it? I'll say this. First of all, the pace at which the technology is developing and moving in AI is phenomenal. So I think anyone, I would say not only telco, anyone who's thinking of trying to build by themselves is going to run a lot of risk of actually being left behind as technology continues to leave fraud. Second thing I'll say is there are areas especially for the telcos who need to look at things that you need to do and there are certain first principles that come to mind. First and foremost, data is the differentiator when it comes to AI quality data and quality outcomes out. And so clearly telcos need to, first of all in my view, need to focus on how they break the silos, what do they do to get their data organized?

(09:37):
And they are the ones who have the data, nobody else can do that for them. So they absolutely have to do the data and get the data piece right. Second thing, common infrastructure, think about how you build a common infrastructure does not mean you've got to go build everything on yourself. As I said, the pace at which the technology is moving, the choices of GPUs are going to expand. The choices of AI accelerators are going to expand on and on, so there is no reason to even go and build that infrastructure on your own. That said, have figuring out what infrastructure and where, again the first principle comes in, bring AI to your data, where is the data? So I sent to the conversation more around data. In the world of ai, if you're thinking about a network use case, most of the data is on premise, it's sitting in the network.

(10:33):
And so think about how do you build your infrastructure and where you build your infrastructure to bring AI closer to your data. Third thing, edge, edge is a price real state for the telcos. It creates the opportunity to further bring AI to the data not only for themselves but also for their customers, for lines of businesses that they want to serve. And again, in there the question comes in where you bring, the other thing I'll say is think about models and there are all kinds of models and there are foundational models that are continuing to be developed grown. Some of them have been open source think lama. So my view again there is there's no particular reason to go build an entire foundation model. There are absolutely great reasons to go fine tune the model or depending on the use case, whether you put a rag architecture or you fine tune a model that depends, but again, depends on the use case and goes back to the data. And then lastly, I'll just sum it up. I mean I think I would encourage everyone who's thinking about AI is start with the business outcomes. What are the business outcomes you're looking at? Whether that's in the area of customer care, whether that's in the area of network automation, achieving that vision of full autonomy of your network, stop there and then figure out what things you need to build, what things you can partner, what things you can leverage from open source. So it's going to be a mix of all in my view.

Guy Daniels, TelecomTV (12:08):
Great. Thanks very much Manish. We've got several more views to get through on this question, but Danielle, I'm going to come across to you first to introduce you to this question.

Danielle Rios, Totogi (12:19):
Yeah, what's really interesting about the viewer that submitted this question is that they're kind of hearkening it to the move to be coming cloud native in telco. And remember we were late to that transformation, right? All the other industries had already moved to cloud and it was pretty well established and there were tools already available. Kubernetes had already won out versus Docker for example. And so we still did that transformation wrong. We didn't do the hard work to become cloud first and build it the right way. A lot of telcos lifted and shifted into the cloud and learned the hard way that that was the right approach to take. And clearly from the viewer's question, they've realized that the difference with AI is that there's no clear winner no IBM out there. That's like the safe bet tool to bet on that you won't get fired for picking.

(13:10):
So I think just by the nature of the timing on ai, there's not the single gold standard that everyone has decided is the winner. So yeah, I think telcos are going to be building a lot of this themselves and like we said in the panel, there's going to be a lot of experimentation going on the building blocks from the startups, maybe not from the big old companies that you're used to relying on, but the startups that kind of really get AI that really got cloud from the beginning, use those as building blocks to accelerate. But again, I think enterprise software is going to be highly commoditized. It's be lots of throwaway code as we experiment our way to what is the gold standard answer on AI and let's learn from being late to the party on cloud and not doing it right the first time and kind of building on Kubernetes and not getting all those cost savings. Let's be AI first, let's be cloud first. And I think that will be the way through the jagged edge of the AI revolution, which will be probably over the next five to 10 years. This is a journey for everyone.

Guy Daniels, TelecomTV (14:15):
Yeah, absolutely. Thanks very much Danielle. And as you quite rightly pointed out, this question's prefaced by the challenges for telcos to adopt cloud native over the past five or six years, right? We've got some more responses. So going back for round two on this question, it's a really important question as brief as you can please, but I'm going to come round for some more responses from you. Scott, let me start by coming back to you.

Scott Cadzow, ETSI (14:41):
Yeah, I think there's also a degree of risk. Whatever you do, you've got to manage the risk and the danger is that AI is still a very much developing technology. The cloud is quite a well understood technology, even if the telcos were late, AI is still in development, still not as mature as we'd like to be. And the danger is that if you jump in too early, you're going to lose huge amounts of money. If you jump too late, you're similarly going to be behind the curve. Finding that sweet spot of when to join and when to decide to become more like an AI-Native is going to be critical to your business. So getting the right partners, getting the right relationships, deciding whether you develop your own or buy in or adopt, those are all critical decisions and you've got to do the risk work first before you start making that leap. But ultimately, I think we're all confident that AI is going to be a major part of future business. Telcos cannot avoid being part of that business, they just have to manage that risk because ultimately they have businesses but everyone else depends on them. A telco failing fails much more than just a telco.

Guy Daniels, TelecomTV (15:53):
Yeah, absolutely. Thank you Scott. Rahul, let's come back to you for in additional comment.

Rahul Atri, Rakuten Symphony (15:59):
In short, I think the telcos the strongest point is the only customers we are the connectivity towards anything they do. Every customer today is a telco customer, but how do you make that customer into an enterprise Data plus AI plus connectivity or whatever could make money and save money is what we should be aiming for. That's all.

Guy Daniels, TelecomTV (16:21):
Thanks very much Rahul and Manish, you wanted to come back briefly with the comment as well?

Manish Singh, Dell Technologies (16:25):
Yeah, very brief. I think clearly the telcos were late to the cloud and I think that still many of them are still getting there. There are a couple of things. Number one with ai, I think it'll be prudent to say that telcos really need to think ahead on this to ensure that they don't get OTD on ai. Number two, I think there's a very key lesson that we can all learn and the industry can learn from the whole cloud native transformation, which is it was done very much in silos and that's why telcos have struggled to achieve, to realize the benefits of the economies of cloud, to have horizontal cloud infrastructure on which they could bring in different network workloads. We OSS core, right, radio, access edge, what have you. We've seen that in many other industries, but we have not seen that in telcos because the implementations ended up being siloed. And I think the implications of that is you don't realize the benefits. That's a lesson learned that needs to be brought very much into the realm of ai. That's why you've got to think about your AI infrastructure, your models, your use cases that you're trying to achieve and do them in an open disaggregated manner so that you're not getting locked in and you are not creating silos. Not to forget data is the differentiator. Got to break the silos as well.

Guy Daniels, TelecomTV (17:49):
Great. Thanks very much Manish and thank you everyone for comments on that question. We're going to move on though and we have a different type of question coming up now. This one refers to telco business models. Really, let me read this one out for you. Does becoming AI-Native require the creation of new AI-Native entities such as Verizon's visible or Vodafones spun off IOT platform which represent a step towards differentiated value propositions? Interesting question. Patrick, can I come across to you for your thoughts on whether or not this is an approach to consider?

Patrick Kelly, Appledore Research (18:36):
My view on it is I think when you look at how the market's evolving, any idea to separate out or spin out is probably a mistake for operators at this point in time. I think what you want to do is focus on where you can get some high value, whether it be in revenue optimization, which is dynamic pricing models, things of that nature, or if it's in the network operations area where you're looking to optimize some of the workflows, build those high value use cases, but keep 'em. I think it would be a mistake to look at as a separate entity and potentially spin out something that I believe is critical. And when we talk about AI-Native telco, again coming back to the domain expertise that's of significant value as well as the data that's owned by the operator and increasingly we're seeing a lot of the operators wanting to maintain control of the data even from the suppliers they have. So opening up APIs and such. So I think that's where the focus should be in the near term.

Guy Daniels, TelecomTV (19:56):
Great. Thanks for that Patrick, very clear. Thank you very much. And Rahul, let's come across to you next for your thoughts on this question and this approach.

Rahul Atri, Rakuten Symphony (20:06):
I personally believe it depends on the go-to market strategy. A, your brand is everything these days. So how do you define yourself if you're known as AI-Native, cloud native, modern, not a bad strategy to collapse everything into the same brand, but in case you want to create more business lines, it's always a easier approach for us to go different strategies. For example, if Telco wants to serve healthcare or telco wants to serve manufacturing, telco wants to serve enterprises in case the brand is strong enough that they're modern enough looked at as a technology firms, why not keep it together? But in case we look at the SERVCO net model, the service co can be differentiated and the AI, telco or AI offerings can be separated into a go-to market strategy. Not taking away that AI-Native telco still needs to hold on the network operations lifecycle all into the air bucket and they need to still have the culture of being the air first, but go-to market approaches could be anything and depends on how you want to take aspects and strategies and also products to market even though the fabric has to be there.

Guy Daniels, TelecomTV (21:15):
Great, thanks Rahul. So it very much depends on the individual and what their strategies are. Manish, lemme come across to you next please.

Manish Singh, Dell Technologies (21:24):
Yeah, I think the way I look at it, AI is a technology and it's going to get applied broadly. And then there's the question of go to market. So when we think about applying AI as a technology and where this is going to get a simple construct, I would like to give the audiences think about AI in the network, AI on the network. And what I mean by that is you're going to take ai, you're going to take generative ai, apply across the network life cycle, drive a lot more automation from planning, forecasting optimization, energy efficiency, on and on, we can go. So there's a lot to do of AI in the network. Then there's the question of AI on the network and what are you offering to your customers, different industry verticals, et cetera. I think in there comes the question, telcos are a trusted brand and in many, many markets they would be very trusted to even be trusted with data.

(22:17):
And when we talk about ai, we've got to talk about data. So I think we got to keep that in mind. Second part is sovereign ai, the rise of sovereign ai. Again, remember we are talking in this context, generative ai, large language models when we talk about language, language captures culture, language captures context, there are local languages that need to be enabled. So if we take a broad view and we see that in many markets, telcos taking that step to build these sovereign AI capabilities to serve the customers. And then comes the question of go to market. In those cases, do you take your existing brand and extend that and leverage the benefits of your brand, your trust, security aspect that telcos have done very well and do you want to leverage that or do you want to create a new one? And there are examples of both. I'll just give the other ones like newer brands getting created. Example being in Indonesia, Linta, that's a good example. Or in Europe you see Fastweb, again telcos creating new brands to offer those sovereign AI capabilities in market. So it kind of depends on what your go-to-market strategies and what's the end offering that you're bringing to the market.

Guy Daniels, TelecomTV (23:32):
Yeah, thanks very much. I think this will develop over the next 12 months. I think if we ask this question again in 12 months time, I'd be very interested to see what's been happening within global telco community. Right? We have more questions coming up from our viewers, but before we do so it is time now to check in on our audience poll for the AI-Native Telco summit. And the question we have been asking this week is how can telco's best leverage AI innovation to improve operational efficiency and develop profitable new services? And you can see the results appearing to my right here in real time. Perhaps it's no surprise to see a strong showing for working with vendors. I think we've got a lot of vendors watching, but it's also interesting to see the interest around the edge, the network edge. If you have yet to vote, then please do so because we'll take a final look at the voting during tomorrow's live Q&A show. So back to the questions. Next one we have received from our viewers. Let me read this one out because well, we've had several similar questions on this topic, but this is one of them, very representative of this one. How do operators make money on AI? We hear a lot about AI for cost savings, but what about revenue generation? Danielle, it's ultimately about revenue generation. What are your thoughts on applying AI for revenue generation?

Danielle Rios, Totogi (25:13):
Well, as the CEO of Totogi, this is an area that we are super excited about. We've been working on it for several years and now with generative ai, it's actually advancing the idea and we've actually been deploying it to real telcos. And so I think people have been trying to do revenue enhancement and revenue optimization with customer value management (CVM). And the problem with that is that there's this long lag between the data getting to the CVM, maybe a machine learning model working there and then humans tweaking the answer and then back out to the individual. And I think the big idea that AI is going to be able to bring to the telco finally is highly personalized marketing plans tailored to the individual, which previously was thought to be practically impossible. How could you possibly support an individualized plan for every single person, millions and millions of subscribers.

(26:14):
But AI is going to do that and traditionally I think on CVM, you're looking at maybe two to 3% revenue uplift with those efforts. It's really hard to do. Maybe best case is 5% to Toki, we're getting 10%. It's pretty easy to get 10%. We see visibility to 15% revenue uplift. So we're not looking for new ideas and chasing, oh, it's all about 6G or it's all about private networks or all the hype that we're seeing around network APIs, which that's going to be really hard to actually realize the revenue if the telcos aren't owning those developer communities. I think personalized marketing is right down the main fairway of what telcos do really, really well, which is serving subscribers. And you talk to any subscriber on the street, they do not feel like the telco really knows them, really sees them for how they use the network.

(27:08):
This is a huge opportunity. Manish has mentioned several times how important the data is to get prepped for this kind of world, but with AI, you don't need to move it out of other vendor systems. You can leave it right where it is. The AI can ingest all this data and you just got to get a way to make those quick decisions. Let the machines do their magic, get it out to the subscribers, start with small use cases, build from there, and I think you're going to see your revenue actually grow. And I would stick to the core business. I wouldn't be looking for brand new fancy maneuvers, new ideas. I'd stick to what we do best and really knock it out of the park.

Guy Daniels, TelecomTV (27:48):
Great. Thanks very much Danielle. Very clear. Stick to the core business. You will make money if you apply AI in this respect. Great. We've got more views. Let's hear some more views. Let me come across to you next.

Manish Singh, Dell Technologies (28:02):
Oh sure Guy, let me give you a topic for the next session, which would be AI ready networks. How about that? We've been talking about AI-Native telcos, but one thing we got to keep in mind is AI needs data and networks need to get ready to serve that data. The need for that data is going to be everywhere. It's going to be on your personal devices. We have now new AI PCs coming in, personal computing devices that are AI-Native and then of course the edge. So AI is going to be everywhere from your mobile devices to PCs to edge, and all of them need data. So first and foremost, the networks need to get ready to serve that data and in their lies the opportunity because not all use cases, not all data is going to get created equal. What's the implication? I think the implication clearly is the opportunity in front of the telcos on serving that data is network slicing.

(29:04):
Here in the US we see one of the leading telcos who's talking about now what 16 plus slices that have been created enabling a number of use cases. So talk about from a monetization perspective to me that's right there within the network. Second thing, edge I've touched about this edge is priceless real estate creates the opportunity for the telcos to bring AI to the data and serve the industry verticals in their markets. And that means that edge starts to become really important. And in there there are a number of initiatives going on in the industry, right where even there are, I mean if you think about things like AI ran, where can you even bring it that far out time will tell whether it's near edge, far edge, irrespective edge presents a very unique opportunity for the telcos to again bring that AI inside the markets.

(30:01):
And last thing I will say number three is the rise of sovereign ai. The demand for GPUs is everywhere. The need for sovereign ai, local language, local culture, serving local markets, not to forget data. Data has gravity, it requires privacy, security and data sovereignty. And these are all the reasons that are creating this demand for sovereign ai. And telcos again, have a very unique opportunity to participate in this. We already are seeing number of telcos who are building sovereign AI capabilities leading to monetization. So there are different ways you can monetize, be it in the network, be it on the edge, or be it with sovereign ai.

Guy Daniels, TelecomTV (30:49):
Thanks very much. Manish. Hold that thought about AI in the round because we have a question on that coming up very soon, I believe. A couple more comments, I'll go to Rahul in a moment, but Patrick, I want to bring you in at this point. So what are your thoughts on how telcos can best make money from ai?

Patrick Kelly, Appledore Research (31:07):
Yeah, I think DR hit on the near term opportunities, microsegmentation obviously is an area we see a lot of activity in. I think AI can help there. She talked about dynamic pricing models, so I won't cover off on that. So I think she's spot on there. One of the other areas that we see, and there's still critical areas for many telcos is churn detection. So you can take AI models, you can do some prediction there, predicting high value customers that may churn for you. That's a big part of where we see potential activities and around revenue monetization. It costs a lot of money to acquire a subscriber or a customer. And so a lot of emphasis still is placed on making sure you can retain or keep that customer. One other big area I think is around revenue assurance. We haven't talked about that and that's the ability to actually identify discrepancies in the billing and revenue collection.

(32:10):
Again, you can use prediction models and enhance some of the capabilities beyond what service providers already have deployed as products around revenue assurance. So I think those are the high value near term areas. As far as edge, we think edge is a huge opportunity for operators going forward, particularly around the number of points of presence that they have. But it's a little bit longer term in that the go-to-market model really hasn't been sorted out. So do you bring in integrators or specialists that know a particular vertical, whether it be in healthcare or manufacturing or pharmaceutical or other areas where you're going to run those AI workloads in the edge?

(33:03):
We think that the operator has the asset, they can actually monetize that asset, but there's going to need to be some serious consideration on how they utilize AI in the edge workloads. The data center side is controlled by the hyperscalers and the cloud providers, so that market's already been predetermined. And so what remains to be seen is the sort of partnerships that we see developing between the cloud providers, the telcos themselves, and these industry vertical specialists.

Guy Daniels, TelecomTV (33:46):
Great. Thanks very much Patrick. Well we're hearing a lot of great ideas today in response to this question. Rahul, I'll come to you next please.

Rahul Atri, Rakuten Symphony (33:55):
I think it'll happen in three phases. First is internal optimization, something like customer experience, something like a reduction of L1 and L2 in NOC. Second phase would be go to market enhancements so that we can move faster to go to market. You can have more products and catalogs live, like Daniel was also saying. The third phase would be creating more avenues, more use cases to earn revenue that can help creating telcos more than just the connectivity, but a catalog to offer a complete package, your customers the data where they're spending time, where what kind of services they like to use. And then you can package it all together. It might come as a healthcare as a private network or it can come as healthcare plus iot plus various analytics on the top. And I think the first use case would be about the data monetization from telecom.

(34:45):
And I think this is how the three phases will evolve. We're already seeing first and second phase, our 1.5 phase being live right now in Rakuten. We are a consolidated 70 different companies and plus. And we have some consolidation method where we know our customers across multiple domains. We use one Rakuten idea across the group. We already see this happening. A lot of mobile customers are using multiple services or otherwise, and I think this trend will only grow. You see already a lot of telcos are going GPS service or probably offering FinTech as entities or even having a multiple larger marketplace of different catalogs. I think this will happen in three phases. And the third phase would be about monetization and revenue, starting with the data monetization, but then ultimately making sure what needs to be happening from quite that telcos would have a catalog other than data and connectivity to offer.

Guy Daniels, TelecomTV (35:43):
Great. Thanks very much Rahul. And let's go across to Scott as well. Scott, what are your thoughts on monetizing with ai?

Scott Cadzow, ETSI (35:50):
Just going to build on comments. Other panels have made, essentially telcos are largely trusted. Identifying user individuality, tailoring your data, telling your offering to single customers increases the possibility of that trust. And then

(36:09):
if you have trust, churn should reduce so you retain revenue also, if you have enough people trusting your network, trust grows. Trust is, I dunno, it spreads. It is like a smile if smile, people smile around you. If you trust an operator, other people begin to trust the operator. You build revenue by trust and if you can use the data, use AI to tailor that data, tailor that knowledge of your customer base, then I think you'll get increased revenue just because you become a native partner of your customer. And if we do that, revenue will grow.

(36:51):
There is risk, but ultimately use data responsibly, use data responsibility to build trust, build trust to build retention and retention will reduce churn means you can grow your revenue our time.

Guy Daniels, TelecomTV (37:07):
Great. Thanks very much Scott. Thanks everyone for those comments. And I want to come on back to Danielle, you started off the comments on this answer, so let's come back for some additional thoughts now that you've heard the rest of the panelists.

Danielle Rios, Totogi (37:21):
Well, I might be taking control of your live q and a show guy, but I'm kind of creating another question from this question which is I think a lot of telcos are thinking about this idea of sovereign ai. Manish has mentioned it several times and I think the idea here is that we're going to build data centers or offer up our data centers with GPUs to run AI workloads, workloads for enterprises on their behalf, not necessarily for the telco's own use, probably we'll use it for that, but maybe for enterprises, and I just want us to remember history. History tends to repeat itself and we had the same idea with cloud data centers. We were going to, I think Verizon went and bought Terramark for I don't even know how many billions of dollars and then quietly exited that business. And the key there was that it was not an infrastructure play of cloud, right?

(38:18):
The reason why the hyperscalers won was it started with the infrastructure, but the key was the software that they offered on top of it. And so before we go off and build a bunch of sovereign data centers with GPUs from Nvidia, which are already in short supply, you got Elon Musk coming in and buying, I don't know, a hundred thousand H two, 200 servers and setting up his own shop, right? This is a very expensive game to be playing in. Very expensive where speed and agility matters, where technology capability really matters. What are we adding besides here are our servers come run 'em versus an enterprise going to a hyperscaler like AWS or Google or Microsoft that are offering things like bedrock, multi foundational models, the security piece and all that other stuff. And so I'm not sure it's the no brainer, slam dunk revenue opportunity that everyone thinks it is.

(39:13):
We are seeing some of those telcos starting to sign deals in partnership with Nvidia in different parts of the world, but this is a very expensive game to play. It's more expensive than the cloud game was and our telcos really ready to anti up and do they have the agility and speed and capability to pull this off? I think a lot of people are seeing public cloud as the easiest and cheapest way despite what Mr. CEO of Broadcom is saying. I'm sure he has a vested interest in keeping those on the ground and in private clouds. But what's so great about the public cloud is I can run my workloads and I'm not paying for the server 24/7, I can run my workload and go away. I can very easily switch different foundational models. And so I probably caution the strategy teams before they are recommending this to the board as it's like no, no-brainer thing. When the telcos are pouring already hundreds of millions of dollars building their own chips, offering alternatives to Nvidia and adding the software put on your big boy pants, this is a very expensive game to play.

Guy Daniels, TelecomTV (40:27):
Thanks Danielle. And I have to remind you, if you want to ask another question, you have to use the q and a form on the website please. Like I am pleased to say so many of our viewers are doing and we're getting inundated with questions from our viewers, which is absolutely fantastic. But before I go into the next one, Manish, let me whiz back over to you.

Manish Singh, Dell Technologies (40:47):
Yeah, I'll just say a couple of quick things. I think number one, the world is multi-cloud. It's going to be hybrid cloud, there's going to be public cloud, there's going to be private cloud and that's where we are and where we will continue to be. There are workloads that lend themselves very well in the public cloud. There are workloads that lend themselves very well in the private cloud and we know very well that there is a growing trend just in the cloud, not even AI of repatriation of workload. So I think it's absolutely critical to really bring this up at the telcos to think through very carefully where your data is data, as I said, has gravity and you want to make sure that you are bringing in the workloads and placing the workloads, be it your cloud native workloads are going forward, your AI workloads where your data is.

(41:49):
Second thing there is cost economics, and I'm not going to get into this a lot here, but we are working with a number of telcos really doing very detailed modeling around the cost economic side of it. And again, as I said, depending on what the use case, the workload, where the data is, that economics changes whether you want to do that on the private cloud or public cloud. And then last thing I will say is when it comes to sovereign AI and GPU as a service, there is no reason for any telco to be going and building GPUs or servers or anything like that that was probably getting suggested. As I said, that's a decision of a build by partner and there is a partner ecosystem including an open ecosystem, for example, the Meta Llama models that continue to get open and more, it's not just meta misra, there's a whole body of work going on with hugging face. So there's a very robust ecosystem that's getting built, not for telcos. I mean these are getting built more broadly. As I said, AI is a technology and it's very different verticals that are leveraging this, but the telcos have the opportunity to tap into all of this technology innovation that's happening in and then identify what you want to build, what you want to buy and where you want to partner, most importantly, where you want to place your workloads and where and how the economics work in your favor. It's not just one simple answer. So yeah,

Guy Daniels, TelecomTV (43:18):
Thanks very much. Thank you everyone. I want to bring in another viewer question now because as I said, we have got an awful lot of questions and I'm determined to get a couple of more questions into today's show. So let me move on to our next topic here and next question. And this one concerns the ran, I alluded to that earlier. So here's our RAN question for today. What are the main use cases for AI in the RAN today versus what we might expect to see in the 6G era of networks in five years time or so? So what are we seeing the application of AI in the radio access network today? What's real today as opposed to what is perhaps still on paper and theoretical and going through r and d and may come into networks in the next investment timescale? Scott, can we come across to you for some responses and then Manisha know you want to come in on this one as well? So Scott,

Scott Cadzow, ETSI (44:20):
I think in the short term like today we're using AI not very much at the run because it's built where we're seeing AI in use I think is largely in the development and design phase. What we'll see in the future is that we're going to use AI hugely in the design phase, hugely in the planning phase, hugely in developing our software and our sighting of 6G cells essentially. But once we get there operationally you see more and more application of AI for dynamic load balancing, dynamic assignment of bandwidth, dynamic assignment of traffic that will become important. So we're going to change where we use AI from being before deployment into both before and after deployment and into real times operations. And that's going to be significant because it means that the role of AI is going to change, it's going to be less well, it's going to change to be upfront in your business offering from whereas now it's proprietary in your business offering. So the value of that AI becomes different because it's now going to change how you deliver services, whereas as of today it's really not used in Iran to deliver services used to predict service and plan service. And I think that's where we're going to see a change.

Guy Daniels, TelecomTV (45:49):
Okay, thanks very much Scott. So there is a clear change coming there. I'll come to Rahul in a minute, but I promise to come across to you. So what are your thoughts about what we're seeing with AI in the RAN today versus what we may see in the next generation of networks?

Manish Singh, Dell Technologies (46:05):
I think there's AI in the radio access network and there is AI for the radio access network. And I'll start with the good news. The good news is with open ran, the industry now has an architecture in place which really paves the way for AI to be applied. And in this realm I would say ai, machine learning and generative ai. And then there are different use cases where different of those technologies get applied. But where I'm getting to is with the R both near real time and the non realtime rig, the A provides now opportunity for different AX apps or R apps to come in, whether that's to improve network performance, whether that's through improve network energy optimization, whether that's to improve network operations on and on, the architecture enables it. The not so good news is that the market on open rent still has to take off and yes, there are a key set of tier one operators who are leading that, but we still need to see that grow.

(47:15):
And I think in that context I would say you touched on six G, I don't want to talk too much about six G, but in general there is a role for standards to really look at the architectural principles to open up the radio access network in a way that you can apply the AI machine learning generative AI onto the radio access networks and derive the benefits out of it. I'll give you some other examples where I see roadblocks right now, for example, we've seen tremendous improvements on channel estimation. The implication of that you can have massive network performance gains to be had. However go to market for any of those innovators with their channel estimation models is limited because most of the radio access network today is closed. And that's why I'm calling this out, that there is need to open this up if the operators and the networks are to benefit from all the innovation that's coming around ai, machine learning and mode.

(48:19):
Let me just briefly touch on the generative AI piece as well. This is where a lot of the network planning actually starts with forecasting, planning, install, commissioning, deployment, troubleshooting optimization, the whole network lifecycle and automation of that which generative AI is here and now. And again we are seeing and actually we are working with a number of partners to enable these use cases to summarize this now opens up the opportunity for the telcos to talk to your radio access network data and do all sort of counterfactual, reasonings, all sort of different planning. The digital twin now giving you ability to do city scale rate tracing and network planning and more. So it's a rich ground to really transform how radio access networks get built and operated with ai, with generative ai but are equally, and I'll close at that, it's important to get these radio access networks to be open to really pave the way for all this innovation to come into those radio access networks.

Guy Daniels, TelecomTV (49:30):
Great, thanks very much Manish. I'll come to Patrick in a moment, but Rahul, let's get your thoughts on what we're seeing in the RAN now as opposed to what we may say in the near future.

Rahul Atri, Rakuten Symphony (49:42):
I think network obviously need to be more programmable going with the use cases which are coming up. They're much more dynamic, much more vast, much more horizontal in the sense. But the question is about radio, so I'll stick to that obviously the first thing is how do you deploy the network? Where do you deploy, let's say next generation sites and whatnot. But we are using our multiple networks where we're working on our open ran based. Our network in Japan is open ran, so we have Rick already live, which is helping us to convert the intent into action. So whatever your in task cases are, let's say it's a dynamic slice or even an application which requires defined quality of service, you can implement that via Rick and also policies to make the network or radio respond to that. I think where we'll be heading towards is the first phase would be the simulation part, call it digital twin, where you'll be able to talk or define the scheduler logics to solve a use case and also define scenarios.

(50:49):
We're not talking about the embedded coding, we're not talking about people who build L1, L2 layer of the radios, but we're talking about the radio optimization engineers or the customer care executive at certain time. So the intent converted into similar simulations and twin, even if you have to let's say sell a slice, you need to know the network would be able to sustain or serve and then only the slice should be activated. So that's the second phase. The third phase is I think and more towards whether we call it a 6G or AI-Native telcos or radios is where the radio networks or the networks will start sensing and understanding the customer behavior patterns, requirements, use cases and maybe turn off, turn on by themselves and also understand how and what need to be reconfigured, whether it is resources, whether it is bandwidth, whether it is the network end-to-end or even the radio coverage footprint areas or the head net network maybe tomorrow having even the satellite network to configure to give the user best experience or the application use case to be solved rather than somebody configuring it manually. So intent going towards the sensing of the use cases and radio network becoming sense of it.

Guy Daniels, TelecomTV (52:01):
Great, thanks very much for that step program there. Rahul and Patrick, let's hear from you on this one.

Patrick Kelly, Appledore Research (52:08):
Yeah, guy biggest thing is coming back to where is the opportunity in the ran, it's around energy management. So most mobile operators, they spend between 30 and 45% of their total cost on power consumption. So that is going to be a huge area for operators to take advantage of the technology and get better energy management and that includes things like putting cells to sleep, network optimization, things of that nature. Just coming back on open ran, I'll just make a quick comment. I mean I think the whole notion around open RAN is disaggregation of the RAND and I agree with a lot of the panelists. We're not there yet. A lot of the RAN is still closed, it's proprietary. I think the opportunity in front of us in the context of AI is once you move to open RAN and you're able to disaggregate a lot of those components, you're then able to leverage the data to do things that have been mentioned, digital twinning, optimizing for your high value subscribers, better energy management. And so I think those are areas in the marketplace that we expect to see some adoption, but clearly open RAN has been slow on the adoption curve.

Guy Daniels, TelecomTV (53:42):
Yeah, thank you very much Patrick. And what has been mentioned today is energy efficiency. We've got the Green Network summit early next year. The agenda has not yet been published, but I can confirm that we will have a day where we're going to look at AI in the RAN because that's so important. And let me also plug the interview that's up today with Alex Choi who is the chair of the AI Run Alliance that's well worth watching. Oh, look at this. We're about four minutes away from the end of the program. Can I squeeze in a lightning fast question? Yes. Let's try and do it. One more question just come in. How much oversight and governance is needed to ensure secure responsible use of AI and data and could this be a hindrance to innovation and development? Scott, this sounds right up your street. Can I come to you? And I'm sorry for brief requesting brief answers here, but we are running out of time. So Scott, over to you.

Scott Cadzow, ETSI (54:41):
Regulation never impedes innovation. That's the short answer. Regulation governance gives you guardrails and we need those because we are not just technologists, we're also members of the society we live in and regulation helps us make that society work. Governance helps make that society work. Technology shouldn't be bound by that, but guided by it. And I think if we accept that regulation gives us that guidance, we can innovate within those limits. It is not a bad thing. Overdoing regulation over doing any kind of governance is a bad thing, but overdoing anything is a bad thing. Let's take it in moderation exist in the framework we have. Yeah, it opens up innovation because it gives us that lead to where we have to go.

Guy Daniels, TelecomTV (55:30):
Nicely said. Thank you very much Scott. And Danielle, I think we've just got time for you to have a quick comment to yourself.

Danielle Rios, Totogi (55:37):
Yeah, I'll go really quick. So yeah, I wouldn't let it slow you down, so be sure to really go fast where you can. I would look into things like Philanthropics Claude, where they have a constitutional ai. There's a podcast for me today dropping with philanthropic where they talk about their constitutional AI and how they keep it safe, how they never use your data for training. And then use things like AWS Bedrock, which adds an additional layer that allows you to control how your data is being used. And so those would be the things that I would tell Telco.

Guy Daniels, TelecomTV (56:14):
Great. Thank you very much, Danielle. Well, we are out of time. Thank you to all of our guests who joined us for this live program. Do remember to send in your questions for tomorrow's live show as soon as you can. Don't leave it too late. I've just seen two questions pop up next to me and it's way too late, unfortunately. But we'll try and look at them in tomorrow's show. And please do take part in the poll. There is still lots of time for you to have your say and you can find the full agenda for day two of the summit on the telecom TV website. We have two panels for you tomorrow. The first discussion investigates where Telco should focus their gen AI activities and the second panel looks at the large language model, the LLM opportunity for telcos and whether or not we need a domain specific telco, LLM. And remember, you can see both of these panels on demand from tomorrow morning and for our viewers watching live. In case you missed today's earlier panel discussion, we are going to broadcast it in just a few moments, so don't go away. Meanwhile, we'll be back tomorrow with our final live q and A show, same time, same place. Until then, thanks for watching and goodbye.