Guy Daniels, TelecomTV (00:11):
Hello, you are watching TelecomTV and our extended coverage of the DSP Leaders Report series. I'm Guy Daniels. And today's discussion looks at some of the points raised in our most recent report, ai, native telco strategies. Now, given that AI is one of the most talked about topics in the telecom sector, we decided to ask the network operator community about their own AI strategies and how AI is impacting their companies, their market and jobs. If you are a registered view of telecom tv, you can download the report now. There's a link in the text below this video, and if you're not yet registered, then go ahead and do so. After you've watched this, there's no cost. You'll have access to all of our reports and videos. Now for this report, we received responses from many individuals representing more than 45 telcos from around the world, including multinationals and single market operators in both fixed and mobile domains.

(01:16):
And now that we have published the results, it's the perfect time to discuss our findings with our special guests. So joining me on the program are Michael Clegg, who is vice President and general manager for 5G and Edge at Supermicro, Aaron Partouche, who is innovation director for Colt Technology Services, and Andrew Burrell, who is head of portfolio marketing at Nokia Cloud and Network Services. Hello. Good to see you all. Thanks so much for coming on this program to help us make sense of our report findings. Now we're going to focus on some specific points that came out of our research and look at these in more detail. Let's start by looking at organizational structure. We first asked respondents if their own company has a management team executive with specific responsibility for AI strategy, and 77% answered. Yes, they do encouragingly high. And we also asked if their company has a dedicated AI team, slightly less here. 72% said yes, they do. Yet 25% said no, they don't. So let me ask our guests, how important is it to have this organizational responsibility and focus? And Michael, could I start by getting your views first?

Michael Clegg, Supermicro (02:46):
Sure. I think first of all, telcos themselves are very large enterprises, both in running their own network and in customer service. As a company, they would definitely benefit from ai. It wouldn't surprise me that you have an executive focused on AI already. I expect boards are already asking the executive team, what are we doing about ai? One of the nice things about generative AI is we can all experience it personally. Each of us can use it. So that makes it more approachable. I mean, other people are saying, what are we doing with this? I do think going forward maybe a dedicated AI team is certainly necessary to kickstart some of the work, but longer term, because AI can be used throughout the company, we talk about AI in the telco AI for the telco and AI by the telco, I think it will be pervasive across the whole organization. So to me, AI is something that you do as a matter of course, but like most change programs, when the new technology arrives, you really need to sort of kickstart it. And yeah, we're talking mostly about generative AI because we do know that telcos have been doing machine language, which is a form of AI for a long time, but what's capturing everybody's interest is this generative ai.

Guy Daniels, TelecomTV (03:54):
Oh yeah, absolutely. Michael, thanks for that. And Aaron, I'll come across to you because as Michael said there, we're all using ai, we're all experimenting with consumer grade geni especially, so we're all aware of it. So how important is it that telcos have some kind of structure, some kind of oversight, some kind of central team to coordinate and lead here?

Aaron Partouche, Colt Technology Services (04:16):
I think someone or a dedicated team in charge of AI is critically important. It helps to coordinate between the different part of the organization towards the same goal, which have IPN different type of missions AI bring huge opportunities, but also significant challenges. We enc called are supporting our customers toward this data-driven organization but rely on an extreme automation and it means that we will have to implement ourself to lead by example the AI mechanisms or the AI tools in our process. So just to share one kind of example we will have to do, but at the end, having a single organization today, this is what we have today called looking at ai, looking at the different application of AI is for me, essential. There are too many different use cases, too many different organization with different goals to not have this type of organization, it would never work.

Guy Daniels, TelecomTV (05:25):
Great, thanks very much for those comments. Aaron and Andrew, let me pick up on this because there are a lot of use cases, perhaps too many. There's a lot of people experimenting and using it. So what's your views on Telco having this responsibility and focus?

Andrew Burrell, Nokia (05:42):
That's a great point guy. So Michael and Aaron have already covered. I think the fact that it's useful to have a central team taking care of this and providing a center of expertise and some guidance on this. But the issue that you are talking about, and I think Michael sort of hinted at it as well, is potentially shadow ai just the same as we have shadow it. You need to have some kind of governance mechanisms and you need to also I suppose have a culture where people are able to share what they're doing. Otherwise the risk is that even though you've got this central AI team that's supposed to be driving the strategy, the reality is AI is everywhere and you want to make sure that you can connect the different bottom up initiatives that different people are going to be doing. It's not such a straightforward topic anymore.

Guy Daniels, TelecomTV (06:32):
No, it isn't. Is it Andrea, thanks very much for those insights. I think the last thing we want is duplication going on again in telcos. Michael, I'm going to come back to you because I think you'd like to add some additional points.

Michael Clegg, Supermicro (06:45):
Yeah, just picking up, I think what Andrew said is key, use the word governance and one of the things we do see with ai, of course it's very dependent on data and telcos have huge data lakes, although I would probably say they're more puns at the moment in the sense they're very distributed and that governance about that data I think is something that's going to be one of the challenges for AI going forward. The privacy and security aspects plus just actually grooming and managing your data to get it usable by all these departments that could actually utilize that data is something that would benefit from certainly from some central oversight.

Guy Daniels, TelecomTV (07:17):
Absolutely. Thanks very much Michael. There's an awful lot to do this coming year, right? Let's move on to how telcos are using ai and we often talk about innovation within telecoms and the role of operators here. So let's look at the results for a question we included in our survey about this and we asked if telcos are well placed to be innovators that can shape the future development and use of AI tools and technologies. And as you can see, the majority said, not surprisingly, perhaps yes, they were at 56% but still 38% said no that there would be mainly would be users. So are telcos well placed to be innovators here? What do you make of the fact that more than half of our respondents think that they are? Aaron, let me come to you to try and make sense of this question.

Aaron Partouche, Colt Technology Services (08:14):
So yes, I was saying before that we need to lead by example, so we have to implement it. So it's also one statement. I mean it's more than we are well placed, we have to do it. I think if we want to help and drive this transformation for our customers, we have to do it now, as Michael said, we are managing a vast amount of data, so this data is so could be so rich and so useful to use AI and test AI tools. But on top of that we have this experience to manage high performing complex infrastructure networks and this experience could be perfect use case for testing ai. And I think the overall of the 40% of the respondent that didn't think that we are telco are well positioned to test ai, they're probably underestimating the impact of predictive maintenance, the impact of customer services we could provide with by using AI for our customers. So clearly I think we are very well positioned to be the early adopters of ai.

Guy Daniels, TelecomTV (09:43):
Thanks very much Aaron. I see what you're saying. There may be a lot of our respondents who are underestimating the potential that the huge potential here, we must develop that with our next report. Andrew, lemme come across to you. What do you make of this split between those telcos who think they will be innovators in developing these AI tools and those who think they'll be mainly just users?

Andrew Burrell, Nokia (10:06):
I think maybe guy is a little bit more nuanced than that in the sense that as Aaron says, the telcos do have a huge amount of data. It's also their responsibility certainly to shape the usage of the II tools. But I think there's a fundamental question to answer which comes back to what's your build versus buy strategy. And I think that's maybe something that we'll touch on later on, but you have to acknowledge that the hyperscale has got very deep pockets. There's a lot of investment already here. It doesn't make sense to try to recreate stuff that's already out there. And in terms of agility and speed of development, AI moves so quickly and there's so many startups active in this space, we should be careful not to try to recreate something which is already out there. And I think that's something that we'll touch on a bit later.

Guy Daniels, TelecomTV (10:56):
Yeah, thanks so much Andrew. There's a dilemma there, isn't there? And it's not just telcos having this dilemma. A lot of enterprises have this dilemma as well. Michael, let's come across to you. What do you make of these particular findings in this question?

Michael Clegg, Supermicro (11:08):
I think if again, we look at telco networks on the AI for the telco, which is really more on the IT side, I think people are most sense that they will be part of the user community more than the creator. And yeah, I'm thinking more chat bots, customer service, customer experience because almost any company that does customers is going to be utilizing that. And what we already see in some cases they probably run in that on a hyper cloud infrastructure anyway. But where telcos can definitely lead, what do they do? They provide connectivity and they run networks. And I think implying AI from an operational side, the AI in the telco there, I think they have an opportunity to really lead some of the industry in terms of security, in terms of sustainability, energy management on their networks in the way they provisioned and determine how to roll out their networks.

(11:57):
So I think that's an area that telcos are naturally leaders in their space and can definitely contribute. And some of that AI will be broader then than just the telco network itself. And I think we'll discuss a little bit as well. I think AI by the telco is some opportunities over there. I think they've got some competitive challenges. Are they going to see that market do some of their hyperscalers or are they going to actually consider that as a revenue generating service themselves? So those are the three areas and I think depending on the area, they can be user leader or follower.

Guy Daniels, TelecomTV (12:29):
Yeah, absolutely. Thank you very much Michael. And a call out to all telcos out there. Please share your findings with us. Tell us what you're doing. We really want to hear how you are utilizing AI right here right now. Okay. Going back to our next survey question though, we asked our respondents if they believe the use of AI can help with the development of new revenue generating services. It's all about the services. And as you can see from this chart here, the results were overwhelmingly positive. In fact, only one person categorically said no. There's quite a few undecided in there though, but 86% of our respondents believe it can help with the development of revenue generating services. So let me ask our guests, was this to be expected? And let me start by putting that to Aaron.

Aaron Partouche, Colt Technology Services (13:21):
So yeah, telco like called are using AI to improve customer experience and automate our processes. This is obvious, but AI opens new exciting opportunities and I can just list few of them. I mean we could potentially provide some edge capabilities, computing capabilities to process and help the processing of AI tools. We can potentially provide another way of providing our connectivity AI application will require more, let's say strict integration with telco. So AI application may need short duration, high performing SLAs, global coverage or women provide things a bit differently with a Federation of operators based on API for instance. So to help enable new type of AI applications and another area that we may also be able to provide it some data, interesting data, of course we have to be careful of what we are providing in terms of data. But for instance, vehicles, virtual conversational data which are standardized digital record compatible with the current AI LLM tools could be very interesting, very relevant to develop some application like fraud detection. So we have different type of area where we can provide services to our customers, whatever if it is at the end infrastructure oriented or if it's related to our connectivity and the way we deliver connectivity or based on the data we can create.

Guy Daniels, TelecomTV (15:21):
Great. Thanks Aaron. A great list there. Thanks for going through those examples for us Michael. A lot of potential as we've heard there, but this is a fast moving sector. Development's happening so fast. What's your take on this positive response from our respondents that they think AI will help them with this?

Michael Clegg, Supermicro (15:44):
Well, as a supplier to the industry, I was very glad to see the high percentage of people who think we can generate revenue generating services. I must admit it was maybe a little surprise to me this early in the cycle to see an indication that high. But it's good. I think they definitely can. We are some of the industry organizations we're involved in like the AI Van Alliance, this is definitely one of their legs in their work is to either do AI on ran, which is really using some of the RAN infrastructure people, speakers talking there about edge facilities using the RAN infrastructure to also do some AI compute and offer that as an edge service to the end users. So that's one area. We've also been working closely on the whole AI infrastructure, sovereign ai, and we've seen one or two telcos pick that up.

(16:34):
And this is really where telcos become the a primary provider of AI infrastructure leveraging their connectivity scale and IT expertise to do that. Now this may be in countries where hyperscalers are less predominant and there may be also areas where there's some more privacy and governance concerns, particularly when using AI to offer government services. But we get into GPU as a service as an easy way to start and then migrating that onto AI as a service. So I definitely think telcos have a position here to move beyond connectivity into services. AI is going to open that door, but it'll be a competitive environment and they will need to get going.

Guy Daniels, TelecomTV (17:16):
Oh yeah, absolutely. Thank you very much. Michael and Andrew, were you also slightly surprised that the extent of the positivity from telcos believing that AI can help them with new revenue services?

Andrew Burrell, Nokia (17:30):
I'm going to sound like a cynic now, but yes, I was very surprised. We don't have in our industry a fantastic track record of creating new services and unlocking new revenue streams. However, I suppose when you think about it a bit more, I mean AI really is going to be central to everything that we do. If you think about as an example, the concept of autonomous networks that deliver zero weight, zero touch, zero trouble services, AI is critical to that. I mean it's impossible to deliver those services without ai. And in that sense, I mean a hundred percent of the revenues are going to be dependent on ai, whether AI is actually creating new services or is more responsible for delivering those new services. That's a slight sort of nuance, but for sure it's good to see and I guess it's good to have some optimism.

Guy Daniels, TelecomTV (18:21):
Oh, it certainly is. Thank you very much Andrew. And we do want to see these new AI revenue generating services from telcos, so we expect a lot this year. Right. Thanks very much everybody. One more key finding from the report for us to look at and we asked our survey respondents, would telco's benefit from the development of large language models LLMs tailored to their specific gen AI needs? Now in our previous panel discussions, this topic has possibly equally divided the telecom community, but when surveyed for this report, only 17% of operators said no, and more than 70% of our respondents thought they would be useful. So let me ask what you all think and what conclusions we might be able to draw from this result. And Andrew, lemme come across to you first

Andrew Burrell, Nokia (19:15):
Guys. So I think like all of these, there is an element of interpretation. I would say could telcos benefit from specific LLMs? Yes, of course, but should they really be investing in building those from scratch? And I think the answer probably is no. Right? That's back to that build versus buy paradox. And I think there are existing off the shelf commercial LLMs that are available. You can use prompt engineering, you can use RAG to tune those and to give some telco specific context. So I think those are probably better options to look at or potentially to look at small language models. You've got a large language model, it's got 175 billion parameters and a lot of that is of zero use to a telco. I mean, what use is a cooking recipe to you when you're trying to run the network? So I think small language models with eight, 10, 12 billion parameters could be useful specifically for telco task, classic fault management, troubleshooting, service assurance, et cetera. So I think maybe that's the route to go down. Personally, I wouldn't recommend trying to create your own LLM.

Guy Daniels, TelecomTV (20:27):
Thanks so much, Andrew. I'm really pleased you mentioned s SLMs because I think since we constructed this survey, the whole landscape for around LLMs has just moved on huge steps, massive steps. We've got small language models, we've got vertical models, all sorts going on there. So yeah, it's very dynamic. Michael, given the dynamicism in this market is very difficult. But any conclusions you can draw from this,

Michael Clegg, Supermicro (20:53):
For my sense is I think they would benefit to me. Again, this is a language and I think a language is dialects. And for any of us in the telco community, we know this industry is full of three letter acronyms and customized bespoke words and terms that it uses. And well, when you get into the network itself, think of all the telecommunicate standards and it won't have been swept up by an LLM that's looking at the general internet. You start to get the data behind the firewalls that all really need to be fed into the system. Whether you do that post LLM on a fine tuning model or rag or whether you use that to build your own LLM, I think there's some benefits in building your own that's industry specific and maybe this is something Etsy or one of those organizations think about or take up as well.

(21:38):
I think the key point made about LMS versus S lms, I think when we started this it was more all about LMS and maybe what we really just need is an SLM. But again, if I look at how fast the industry is moving, we talk about LMS being billions and s, SL S being millions. I think we're already at the stage where an LLM is trillions of parameters and s LMS are billions of parameters. And with the compute that we're starting to get available now, running a billion parameter or multi-billion parameter SLM is not going to be that big a deal going forward. So I think the models will get bigger and they will get easier to execute on. And I think there is a benefit just of having an industry specific language associated with the telco business.

Guy Daniels, TelecomTV (22:24):
Great. Thanks very much, Michael. It must be a very difficult decision to make this whether you press the button now or you wait a week or a month to see how it all changes. Aaron, let's come across to you. What do you think about these language models and whether telco specific language models could be of use to telcos?

Aaron Partouche, Colt Technology Services (22:44):
We were talking about this vast amount of data we have today, the telco that could be perfectly fit for testing AI tools. Having a telco specific LLM would be beneficial because we will ensure that data privacy and regulatory requirements are met. Firstly, it would help us to automate our troubleshooting. It could help us to integrate new services to play, also to manage some of our customer potentially data to improve the experience. So definitely I think there is a real huge value for us to have a telco specific LLM, but will be adapted to our use case and also to our data.

Guy Daniels, TelecomTV (23:30):
Thank you very much, Aaron. Great place to leave it because we must leave it there for now. Michael, Aaron and Andrew, thank you all for taking part in our discussion. If you haven't already done so, please download the DSP leaders. It's available to all registered viewers of telecom tv. Registration is free as I mentioned earlier, so sign up now and you can not only get our AI report, but you'll also have access to all of our DSP leaders content. And if you are new to telecom tv, please take a look at our dedicated telcos and AI channel where you'll find the latest news analysis and videos including all of the discussions from our AI Native Telcos Summit series plus the recent in-person telcos and AI event. For now though, thank you so much for watching our DSP Leaders Report panel on AI and goodbye for now.