Guy Daniels, TelecomTV (00:23):
Hello, you are watching the AI Native Telco Summit part of our year-Round DSP Leaders Coverage. I'm Guy Daniels, and today's discussion looks at the benefits of becoming AI native. Now this will mean adopting new best practices, fostering collaboration with partners and leveraging the capabilities of gen AI. So how will this impact existing processes and how can telco's use AI to achieve their strategic goals? Well, I'm delighted to say that joining me on the program are Manish Singh, CTO Telecom Systems business at Dell Technologies, Danielle Rios, acting CEO of to Togi, Rahul Atri, president of the OSS Business Unit, Rakuten Symphony, and Patrick Kelly, founder and principal analyst of Appledore Research. Hello everyone. Good to see you all. Lots to talk about today because obviously there is a huge interest in telco AI. So let's dive straight in and I'd like to ask, what are the best practices for telcos who are looking to build out their AI capabilities to become truly AI native and how can they manage this build out and subsequent operation? Now, Danielle, could I come to you first and get your views and how telcos can best achieve this?

Danielle Rios, Totogi (01:53):
Absolutely, guy, so glad to be here, but I'm here with maybe the bad news, which is there is no best practice in AI yet. It really is a jagged edge of technology. Every day we wake up and there's new advancements, there's a vendor on the scene that's gotten big funding, that has a big idea and is unveiling it to the market. And so I think telcos can't really rely on their current vendors. The vendors are still figuring it out too. And my message is to telcos is sadly you're going to have to navigate AI yourself. Everyone's learning how to surf this wave. We're all novices and it's still very early days. There's still a lot of improvement, banked and coming out, and we got to be ready to adopt it as soon as it comes out. The way I think about AI is there's really three parts to it.

(02:51):
There's the hardware and the chips, there's obviously the LLMs and the models or the software. And then the last piece is your data. That's a big, big piece to making AI very valuable and useful. The chips in the research, there's probably three x improvements coming every year over the next three years. That's nine x and just the hardware models are improving 10 x per year, and so there's a lot of still value. So right there between the models and the chips, you're looking at probably a thousand x improvements coming in the next five years. So we're really, really early days. And so my question to telcos is where's your data? Data is still really tracked on premise. It's dark data, not very accessible, not very clean. And so what are telcos doing to put themselves into position? And so AI is going to require massive change across the telco.

(03:49):
Literally every job will be impacted the skills you need, the culture you have. And so I probably start by looking at my own culture and trying to really foster a culture of experimentation so that we can tolerate more experimentation within the telco. But I wouldn't wait five years for all the dust to settle and figure out who are the winners and losers on AI. That's kind of like learning how to surf at the wall of skulls in Tahiti like we saw at the Olympics this year. And so that's a bad place to learn how to surf. So start with the easy stuff now, start to surf the baby waves. That's what we're doing over at Totogi. We're trying to build a team of surfers so we can surf the jagged edge of AI.

Guy Daniels, TelecomTV (04:38):
Thanks very much. Danielle and AI is one of these, so-called exponential businesses. There's so much happening, isn't there? And it's early days, so Manish, let me come across to you. Is it too early to be talking about best practices at this stage? Are there any lessons or any guidance that telcos can have when looking to implement and build out AI?

Manish Singh, Dell Technologies (05:00):
Yeah, thank you guy. Great to be here first and foremost. No, the answer is no, it's not too early. Rather it's urgent to get started to really build the muscle around AI, around generative AI in particular. And the reason is very simple.

(05:20):
Telcos cannot, must not, should not be OTTed on AI, on generative AI. And the good news is there are already a number of telcos across different markets who are actually already working on AI, generative AI use cases. They are experimenting, they are actually trialing and in certain areas they've also started to run scale deployments.

(05:51):
Lemme start with the basics. First and foremost, you have the infrastructure, you need the right infrastructure. And the question when we talk about AI, everything has to start with data. Where is the data? Most data is sitting on Preem data has gravity, data has security data, has privacy issues, data has sovereignty issues.

(06:17):
All of these things have obviously given rise to the sovereign AI and we've seen in number of markets, number of telcos who are already building sovereign AI capabilities, move up the infrastructure into the models. The good news again is there are this choice, there are number of models they'll continue to evolve and grow. We will see the parameter sizes of the models continue to evolve and grow, but we are also seeing models on the other side become better and more efficient on the smaller language model sites for specific tasks. And we're going to see this continuum play out between the very large language models to small models. And as part of that, the third thing that comes in on top of it is the use cases and what are the business outcomes. And we can probably come back to that later in the discussion, but if I just summarize it, there are use cases across customer care, billing care and more to drive a whole lot better customer engagement. There are a number of use cases already that are getting worked and into trials on network side of it for network automation, network lifecycle management that start from planning, forecasting all the way into closed loop automation and more. And I think the important thing here really is that generative piece. I mean with the generative AI now with the right infrastructure, the right models, it's definitely unlocking these use cases is now within the reach and that's the work that's underway.

Guy Daniels, TelecomTV (08:01):
Great, thanks very much. Mani. There is a lot for telcos to consider here. We'll come to Rahul in a minute, but Patrick, let's get your thoughts first.

Patrick Kelly, Appledore (08:10):
Yeah, guy number one thing is you have to have a clear business case for any AI project and any project at all. We're talking about AI, but in the context of AI, if you don't have a clear business case, there really isn't anything to follow. I think a lot of the excitement right now is around the technology, but you have to align your business problem with the solution that you're going to develop. Second thing that I would say is you have to embrace not only internal projects that are specific to AI or gen AI, but you also have to look at the supplier community because we actually see a lot of solutions making their way into the commercial marketplace. This is around customer care billing, even in the network operations for things like field force dispatch. The third thing I would say in terms of best practice, which has been mentioned from some of the other panelists, is you have to prioritize data governance and quality. So you have to break data silos. If the data is not accessible, the models just aren't going to work when you train 'em. So that's where I would point a lot of the operators in terms of best practices at this stage.

Guy Daniels, TelecomTV (09:33):
Great, thanks very much Patrick. And we've been hearing so much about data recently on telecom TV and association with the deployment of AI. I do wish at times telcos had taken more interesting care and made more attempt at sorting out their data issues earlier. Rahul, I did promise to come across to you. So what do you think? Is it too early for best practices? Is there any guidance? What can telcos do? How can they build this out?

Rahul Atri, Rakuten Symphony (10:01):
I feel this is the moment for telecom to become great again or maybe the tech leaders. Again, we really need to look at it as multiple phases. People talked about infrastructure data, we've been talking about that for long, not only for AI but automation as well.

(10:18):
I personally feel that every telco should have decided their big picture by now the moonshot. How do they want to do the business, how do they want to set up the culture internally? And the technology will evolve. I mean every time open AI releases a new model, a lot of startup die, a lot of things evolve as well and I'm very sure it'll continue going forward. But we look at the brighter side as well. New devices are coming, new way of interacting with the systems are coming. I see this as an opportunity rather than something which needs to be thought through in the future. We really need to look at the moonshot, what next big thing telcos can do. This is again, a moment where the telcos can become more than utility and the center platform for everything.

(11:01):
We can become agile again, we can also become the only technology provider, whether it is gaming fest, whether providing other opportunities because it's not going to be the same challenge of orchestration automation, complex systems, multiple BSS/OSS system collaborating, multiple domain orchestrator collaborating. It'll become a lot more easy for a customer to interact with telcos and I'm very sure on these things. Although on the other side we need to simplify a lot of initiatives which are happening. We were talking about AI alliances, people talking about kamara. I think it all have to come down to few pieces, call it intent, call it the data strategy, call it the context. All those things have to come together. The last piece is because we cannot be the common platforms, models have been designed, there are different strategy of open closed models. There've been a strategy on the infrastructure which somebody has already taken the lead. The best thing for telcos would be define their big strategy and also find a partner. The partner could be anyone providing model or infrastructure together via SaaS play or they can build their own pieces and become the top integrator in the ecosystem.

Guy Daniels, TelecomTV (12:13):
Great, thanks very much Rahul. And I want to pick up on your last point there about partnerships and move across to Manish. Partnerships are so important in telco as we know this. We've seen the evidence of this down the years. How can collaboration between telcos and vendors help to accelerate AI adoption in the telecoms industry?

Manish Singh, Dell Technologies (12:32):
Yeah, guy, I think just as I answer your question, I just want to make sure I at least share my thoughts on the big picture. I do completely agree that telcos have a great opportunity in front of them with the growing adoption and rise of AI. And I want to put this in three buckets. Bucket number one, revenue generation. We haven't talked about that. And I think with the rise of sovereign AI in many markets, it's presenting an opportunity that the telcos have never ever had before. And this is why I say telcos cannot, must not, should not be OTTed in AI, in generative AI. And we are seeing with the rise of sovereign AI, especially in Asia Pacific, Europe and other parts of the world with telcos are already now getting into set up the right infrastructure in place models. When it comes to models.

(13:29):
I touched on large language models, small models, but what's really essential about models in larger part of the world is language culture context. And what we are talking about is intelligence generation. So setting the AI factory construct with data coming in, tokens getting generated out, intelligence getting generated out with the local language culture and context. That's important, that's required. And this creates the opportunity around sovereign AI now to realize this whole opportunity, it is absolutely essential for the telcos to work with a broad ecosystem and ecosystem that can bring in the right infrastructure, bring that infrastructure closer to where the data is, bring in the right kind of models, not build the foundational models. I mean there are a good set of foundation models in there, but of course need for fine tuning and then adding that context culture and in certain parts of this, the local language pieces that need to be brought, right?

(14:36):
So get those models in and then it breaks in for AI, internal use cases, AI offerings to the markets. And last thing I'll close around, this is the edge opportunity. I think the telcos are sitting on a priceless opportunity with the edge. A lot of the data is going to get created in the physical world, bringing AI closer to the data where the data is, which is where edge is. And again, that's another key opportunity to unlock. But to unlock all of this bringing together a broad ecosystem is required in telcos need to really collaborate on that.

Guy Daniels, TelecomTV (15:21):
Great. Thanks very much. I would be interested to know what our viewers think as well. So viewers please send in your questions. We'll get them into our live show. I know there is an opportunity out there. We'll telcos grasp it. Rahul, lemme come back to you and stay with this theme of partnerships. What are your views? Can you expand on your views here?

Rahul Atri, Rakuten Symphony (15:42):
I think it's a must. We don't have a choice. I think we need to form the alliances and partnership and build ecosystem. Nobody can solve the entire bit of puzzle alone. Telcos have the greatest assets in terms of (A) the data, (B), the industry is so ized that only telco has the data. (C), it's very important that the connection between the telco provider or service provider and the customer is such a unique opportunity that it has been there for decades and it is going to stay. Whether telco takes a opportunity become TechCo or technology savvy, again, provide more use cases or whether they get OTTed as Manish was saying. And we need to believe that what we do best and what we cannot do, some of the infrastructure challenges, some of the edge opportunities, some of the opportunities in terms of models and fine tuning those models into telco specific use cases.

(16:41):
The last bit is it also helps sustainability because every time you query, every time you tune a model, every time you work on something new, it takes a lot of resources out of mother nature. So I think ecosystem helps everyone and the opportunity is so large that there's a win-win for everyone if we partner together. And that's why it's very important to see the moonshot, the big opportunity, the strategy that what would be the future telco or what would they sell or who would be the potential customers and who will be the partner supporting them in the journey.

Guy Daniels, TelecomTV (17:15):
Great, thanks very much Raul and Danielle, let's come across to you for your thoughts on partnerships.

Danielle Rios, Totogi (17:22):
Yeah, I'm an enterprise software girl, right? I'm the vendor trying to sell my software to telcos. And I'm going to say something that is completely contrarian to that, but I think we're entering a time of highly commoditized enterprise software. The days of installing a system and running that same software for five or 10 years and taking a series of upgrades I think is over. And that's not a telco thing, that's a global thing. And that's because I think the big power of generative AI is really being able to generate code. And that was usually previously an elite group of a small group of people could do that, the computer scientist. But it's really lowered that barrier for entry for almost anyone able to create code. And so I think telcos are going to enter a phase where they're going to be serving the market, they're going to be seeing what vendors are doing.

(18:22):
They might adopt a piece from this one, adopt a piece from another one, oh, another vendor is leaping ahead, swap it out. And so we're going to get into this phase of swapping stuff out, building our own stuff, enhancing it with your own code from your own people. And so it's a little bit of a completely different way of buying and consuming software and we almost need to think of it as throw away code, right? Throw away software all the time, which is kind of crazy because that's not how telcos run their business today. They kind like to test it, make sure it's perfect, install it, don't do an upgrade for three years or just do bug fixes only if necessary after massive testing. And we really need to break things down into micro little pieces so that you can throw them away and replace it with a new model, a new capability, a new piece of software, a different vendor. And so that collaboration is going to, I mean it's just going to blow up what we've been doing previously and the telcos that really embrace that rapid experimentation are going to be able to ride that jagged edge of AI the best of the next five years or so as the models are changing rapidly.

Guy Daniels, TelecomTV (19:42):
Oh yeah. Thank you very much Danielle. Well great ideas from everyone there who needs gen AI when we've got the four of you, this is incredible. Well, talking of gen AI, we have a dedicated panel on generative AI tomorrow, but I'd like to ask you now, and Patrick, perhaps I could start with you, how is the increased use of gen AI and chatbots helping telcos not only in customer engagement but also with say maintenance and field operational work

Patrick Kelly, Appledore (20:11):
Guy? What we see in the market as a market research firm is there's accelerated progress in production environments for things like customer care and billing inquiries. That's where we're seeing the early green shoots where we're moving out of PoC into production. I would say in terms of the network gen AI is also showing some promise for things like guided network troubleshooting. So there's, I've talked to a lot of operators where they've archived and documented huge amounts of data they have in manuals and making that available to these models so that you can do assist in and around network troubleshooting and things of that nature. I think also looking out into the future, we should expect to see some products coming into the marketplace that are focused on network optimization and planning. Those are a little bit longer term, but that's where we see the market going.

Guy Daniels, TelecomTV (21:18):
Great. Thanks very much Patrick. And a few more comments on this question. So Rahul, we'll come across to you next. Is gen AI an essential part of the moonshot you've been talking about today?

Rahul Atri, Rakuten Symphony (21:31):
Yeah, just carrying on what Patrick was saying, we do all these things a lot broader as well. For example, we are divided into three different sectors. A is the gen AI based SDLC program where some of the software teams which we own have already adapted to platforms to do their software development, whether it is writing code, refactoring code testing codes, and a lot more, even the business analysts and the product managers are being deep diving into platforms and understanding where the world's going. We started even designing our platforms the front end based on that, and to Daniel's point earlier, it's probably becoming something which is highly customizable based on the customer we're talking to the use cases we're trying to solve the base platform of the data structure, data governance, everything remains the same. Other parts are, we're doing a lot of change management in our networks.

(22:27):
The CRS or the change requests have been reviewed completely by the AI agents. So if there's something missing like a lab test case report where we really need to make sure that this change would not affect something in the production network or need to do a digital simulation to run, this is all possible. Some of the service requests or the impacts which are there in the network, the bots are able to call us and even before you join the WebEx or a Zoom link, they're able to explain what the situation is, what have been discussed so far, and the guided DevOps platforms where you can actually talk to the platform and say, show me the events which are happening. Show me these events are happening for the first time, or there's a correlation every time you put a code on this branch, this is what is happening.

(23:13):
And then even create a playbook out of that to say, record this next time, create a playbook to replicate the troubleshooting we're doing, we're doing a lot more on the optimization route as well. The tool is able to understand how the consumers are behaving a certain pattern, certain loaded sites drive test logs are being reviewed and the optimization schema, like what changes we need to do the network. So in a way, what we did was take two major steps to say if you were to replace yourself, the AI empowered you, what would your life look like? And then also understand what is the ROI coming through and focus only on those use cases which were giving us banking value for us. This is the definition of the AI native telco look like today.

Guy Daniels, TelecomTV (23:59):
Great. Thanks very much Rahul and Manish. As we've heard from Rahul and Patrick, gen AI does touch a lot of areas of telco's operation. What are your views?

Manish Singh, Dell Technologies (24:09):
Oh, I'll actually give you very real, very practical areas that already are life today. So this is not about talking or future predictions or things that are going to happen. Number one on the customer care billing care side, obviously going towards these chat agents that leverage the value of and the power of the large language models, the conversational AI, but also tie at the backend with your existing OSS/BSS systems to drive automation. We already have been doing a moonshot catalyst project on this with SK Telecom and other partners and we've talked about this, which is all about driving next level of customer engagement, improving that customer experience, put new offers out to the customers device offers, plan offers, roaming offers, understanding customer intent and more, but then automate all of that at the backend with your BSS stacks. So if a customer picks a new device, a new plan, new roaming edition, whatever they do, automate all of that and you can start to see how this now impacts customer care billing care.

(25:26):
You can do churn predictions, sentiment analysis, generating the AI is very, very good in all of this. Then let's talk a bit about the network side of things. We already, as part of our AI for telco programs, we've been working with a curated partner ecosystem on network troubleshooting use case generative AI is very good in doing anomaly detection. So you can take your large large PCAP files, message flows, logs, et cetera, all that unstructured data, put it into the generative AI and identify anomalous flows, what's the outcome? Improving network engineered productivity, reducing the troubleshoot time from hours to minutes. And these are numbers that I'm sharing with you are actually coming from a tier one operator as we speak. So you talk about business impact, business outcomes, your network engineer productivity goes up, network downtime network goes down, network improvement, KPIs improve, et cetera, et cetera.

(26:34):
I want to give you another example of generative AI with time series data. This is around fall prediction. So yes, taking the value of generative AI where you can take all your logs, alarms, et cetera, plus the time series data, and again, we are working here, the unlock we are going for is how do you take the time series and the large language models for the unstructured data piece and can predict faults the outcomes, 90 plus percent accuracy on real data on fault prediction. You can already start to see the implications of this that if you are a chief network officer who's building and operating a network 12 hours before you know a particular card and a particular node is going to go down, this starts to impact everything in terms of how you can plan your truck rolls, your remediation steps, et cetera, et cetera.

(27:30):
So fault prediction, fault remediation is another use case. And last one, there are many more, but in the interest of time, let me just give you one another partner we are working with on the radio access network site, again with an operator here with the unlock we are going for is what I would call is talk to your radio access network data. And once you start to do that in natural language, there are so many things where you are looking at it from an energy efficiency perspective to QOE analysis and more things that would take weeks for telcos to get to. You are getting those answers in minutes and hours and that's the unlock with generative AI. So it's really improving your network operations, your network planning, your network life cycle more broadly I would say. And then on the other hand, from a customer care billing care, customer engagement and more, these are all areas that generative AI is going to really is already starting to unlock unprecedented potential for the als.

Guy Daniels, TelecomTV (28:32):
Manish, thank you very much for sharing those examples with us. Danielle, I'll come to you. It's very exciting what gen AI in particular can touch on and the transformation it can enable for telcos.

Danielle Rios, Totogi (28:47):
Yeah, no, absolutely. Just Manish gave some really great examples there, but on my telco in 20 podcast I had Jaime Tati, the chief insight officer at TELUS up in Canada and he had some really great examples of what they're doing with generative AI out in the field and I just wanted to share it really quickly, which is they have technicians that can access AI on their devices and so they're using multimodal capability of an LLM vision, obviously the analysis piece and it's helping that field technician, they're out, they don't have a lot of resources at their fingertips, but they have their device and it's helping them to solve problems more quickly. First time, right when they're out there and the field technicians love it, they're not rejecting this technology. Jaime mentioned that there was 75% adoption by those field technicians, so they welcome it, right? So you're changing their job and you're changing how they work and they are embracing it and they're using it and it's making that business better and it's dropping the bottom line for Telus. So that podcast, he talks for about 15 minutes about what they're doing. It's a really great example of how people are applying generative AI in the telco.

Guy Daniels, TelecomTV (30:10):
Great, thanks Danielle. This leads very nicely to our next question and this is something we've seen with cloud native, but if AI is truly to be regarded as a core competency of telcos, if we are going to build the AI native telco, then how is this going to impact existing culture within a telco? And by that we mean skills, we mean organizational structure, market strategies, et cetera. There's a lot of change there. Rahul, let me come across to you first for your thoughts on how this impacts telco culture.

Rahul Atri, Rakuten Symphony (30:47):
I think again, what we start with is dreaming about the future. How does the organization of future look like? And also like Danielle was saying, embracing the change as a product manager growing up, I think the best KPI we used to calculate and probably measure the success was adaption. And same is for AI. I mean we can base the best of the technologies, but if you give the field technician device with generative AI, if they're not using it, it's of no use. You spend a lot of money. So on that front, they need to know what's out there, the basic education, how they can use the fabric. What we have done in our organization is all the 70 business ecosystem has a common fabric. I call it Rakuten AI. And then you start building use cases, fine tuning and also shared knowledge and learning. For example, one of the use cases we did for our vision model, which was used by the other e-commerce business, we just ported it into telecom.

(31:49):
And within days we were able to test out the site optimization reports and even site inspections earlier, whomever goes onto the field collects a lot of photographs, whether drone or human. And then somebody in the back office have to visualize and create use cases if there needs any field maintenances and not right now that's happening runtime without anyone reviewing those reports. Same happens on the drive test. So for us it's like people should look what their future look like, feels like that they have a partner in AI and what use cases they can evolve, make them feel that this is another evolution toward their job role and what their job role would be in the future. For example, one of the team we were working with very closely, which is the RF optimization team, how do they do work today and how the AI can help them enhance them and how do they record that progress?

(32:39):
Because what we have seen initially is when you give them the enterprise licenses or the models to talk to, they will talk all sort of things to them and some of them would not be productive. We're very cautious of how we solve these problems, the multi-model problems and what kind of queries goes into our internal, local fine tuned model and what queries are broader, which needs to go to the internet or probably a partner like OpenAI. And also that helps us in the efficiency part as well because it's not easy to spend a lot of dollars into every query which we're doing. Also looking at sharing the learning across. But in the culture aspect of it, people need to adapt and understand that it's happening, it'll change, it is changing everything what we do and touching every aspect rather than doing such. I think people are more comfortable talking to the gen AI now and also some of the capabilities which we can do every day.

(33:35):
I mean how we are doing the meeting notes, are we doing meetings or you have all the updates, data uploaded, model trained where you can all basic things. How is my cluster performing? How is 4G versus 5G performing? And also understanding the intent with the context. So how is 4G versus 5G performing and business exec, when they're asking, they probably are asking about arpu. When a customer care agent is asking, they're probably asking about how the customers are feeling or their experience. And when an RF engineer is asking, they probably are asking about the coverage capacity and how the customers are feeling on the radio signals. So that's why we are focusing a lot of in our energy intent context data and the ROI to measure and show people that working together with the technology will only help them groom and also take us to the future organization.

(34:23):
We're talking about some of the targets which we're setting for all of us as. For example, in the operations there will not be L1 and L2 layer will directly talk to L3 and L4. We are also working on taking RAN traces, ran KPIs, all the logs sys logs into the system. So an L3 kind of person can actually go and talk to and understand what's happening on the radio layer or probably transport layer and also going forward build playbooks where the AI bot can actually talk and do analysis and raise an alert or a change request going forward. So some of these things are where we are a mix of where you are today, where we want to be in the future, and how we take the leap of faith and journey together to reach to the future state, but with the same people, they should not feel scared, they should feel that they're a part of the journey and this is about evolving them and organization and capabilities efficiencies together.

Guy Daniels, TelecomTV (35:15):
Yeah, absolutely. Thank you very much, Rahul. Well, I know you all want to come in on this question. So Patrick, I'm going to come across to you next and how the AI native telco, how it impacts telco culture. Are we fortunate that AI in general has more acceptance within society and so it's seen as something more helpful now?

Patrick Kelly, Appledore (35:36):
Yeah, guy, I think you hit on something important right at the start, and that is it is just another discipline, right? So if you look at core competency that's going to be required for AI, it's accepting some of the same principles we used in cloud DevOps. And Rahul is like, I mean he's a practitioner, he rolls up his sleeves, he gets his hands dirty. And so I think some of the examples that he was running through are clear areas where the market's evolving and the market's developing. I think one of the things I would say is automation, reusability, data model governance, model management, all those things are critical for an organization to really be AI native. And so if we look at automating the ML lifecycle, you have to look at various stages of the ML pipeline. The teams can develop and deploy models more quickly.

(36:34):
If this practice is used, you're getting consistency and reliability, things like model monitoring and management. So once a model's in production, you need to monitor, make sure the predictions remain accurate and relevant to whatever problem or business case that you're looking to solve. There's been a lot of discussion over the last half hour or so around model governance and compliance and data. This is clearly something that organizations need to think about. And then the whole, again, another cloud DevOps is CI/CD. So we know to Danielle's point, some of the codes going to get thrown away, some of these models get thrown away. So applying that whole CI/CD principle is something that organizations are going to need to accept.

Guy Daniels, TelecomTV (37:32):
Yeah, great points. Thanks very much Patrick and Danielle, lemme come to you next because in your previous answer you were talking about employees accepting and getting buy-in to using these tools. What are your thoughts? Can you expand more on how AI will impact telco culture?

Danielle Rios, Totogi (37:51):
Yeah, I know I said before, I'm the enterprise software girl, but I actually spent 10 years in hr and if there was ever a moment for the HR department to shine, it's now, right? This is a pervasive culture change where every single job is going to be impacted. And so it starts with strategic HR in that leadership boardroom level, aligning the 10 or 20 top people of a telco and aligning 10 or 20 people. It's not that difficult. You can go have 20 conversations and see where everyone's at. But in a telco with a hundred thousand people in it, how are you aligning the other 99,980 people that you need to get around what we're changing, why we're doing it? And not only why is it good for the company, why is it good for the telco? Why is it good for you as an employee and how are we addressing the elephant in the room of aren't we going to need less people, fewer people?

(38:53):
As we roll this out, I think the HR department needs to work with the senior leadership team around how we're incentivizing people to help us on this journey. Because when you get those other 99,980 people aligned, your telco flywheel, the flywheel of change is just going to really, really accelerate. But just throwing AI into the culture of the organization and expecting productivity improvements to happen. I mean, we've had AI now for, what has it been like two years since the chatbots have really come out from open AI and philanthropic? How come we're not seeing natural improvements happening in enterprises around the world? And that's because people are either scared to share their successes, they don't want their job to go away, and so they secretly use it and don't really share what the team, how it's improving their work. They might be hiding it so they can get ahead and get the promotion over their peers.

(39:55):
But I think organizations really need to put, it's not always money. It could be promotions, it could be visibility, kind of highlighting the team within the organization, but this is hrs moment to really be a partner to that leadership team, cascade the message through the organization and answer the elephant in the room of what this means for my job and will I have a job when the AI transformation is over? And I think hiding from that is a really, really bad habit that leaders have. They don't want to address the tough questions. They're like, yeah, yeah, yeah, just go work on the AI stuff. But there's such a huge opportunity because humanity will rise again. We've had technology advancements from horse and buggy to now electric cars, we will rise, we will figure it out, but how are we helping the 99,000 people in our telcos to get new skills and maybe they don't stay with us forever, maybe they move on to other organizations and other industries. Such a huge opportunity to lift everyone up and really go fast inside that telco with hr.

Guy Daniels, TelecomTV (41:09):
Great, thanks Danielle. And Manish, I come across to you because we've been talking about the opportunity during this panel session, but to embrace that opportunity, it does require a lot of change in the organization. It will impact all aspects of the organization. What are your thoughts here?

Manish Singh, Dell Technologies (41:25):
Yeah, great question. Every transformation eventually comes down to technology, people and process, right? And all of these need to evolve. I think if I just take first the technology piece, clearly the technology in this realm is moving at a very, very fast pace. We've talked about it, how the infrastructure and the optionality in there continues to evolve. The models themselves are continuing to get smarter and evolving. And where we are today, if you think about it, we are at a point where a lot of these generative AI is getting used as copilots or assistant, or whether that's for a network engineer to troubleshoot things faster or for a field and technician for install, commissioning, troubleshooting, whatever. We are doing a much better work for the knowledge information being made available through all these rag architectures and whatnot. So that's where we are. But where we look at this technology is going to move to even with the more agent frameworks around the horizon, we're not only working as a copilot, but these agents now being able to take complex tasks, break them down, plan, reason, and execute, and all of that means there's going to be more and more opportunities to drive more automation.

(42:53):
So that's just a bit of a view around the technology piece of width and how things are moving. When I look at it from a people lens, I think it's absolutely critical that we need to have teams build muscle on this new technology and continue to evolve their skills and understanding of the technology, where the technology is going. And I think that's exciting for any engineer who wants to learn and understand where the technology is going, what's new, what can be done, how can it be used to unlock new opportunities and more. I see that as very exciting times to me, this is about eventually coming down to driving productivity gains, driving faster outcomes for the problems that we have had and which would otherwise take longer times. So whether you're a software developer developing software and at times doing got mundane tasks for a network engineer who's troubleshooting network workflows, hey, now you've got new tools that makes it much more faster.

(43:53):
Or if you're a call center agent taking customer calls, trying to understand the customer issues, well again, you've got new tools to understand customer intent context, how you can better serve the customer, et cetera. So I think to me it's about again, the productivity gains, doing your work much better, and of course requiring the need to upskill the teams across the board. And then last pillar is the process. And I think again, this will require process transformation. First of all, anything you want to automate will require clear definition, standardization, paving the way for that process to be automated. So it sounds simple, but it's complex. How do you simplify your processes? How do you standardize and then pave the way for them to be automated? And last thing, just keeping it short in the interest of time is changing processes, especially in the world of telcos who are very used to working with a few vendors.

(44:51):
This is going to be about an open ecosystem. And so transforming processes from procurement to implementation to roll out at scale. This, again will require that process transformation to work with a broad ecosystem of partners, whether that's on the infrastructure side, models, use cases, and more to bring all of that together and be comfortable in working with that ecosystem and defining the processes accordingly. So those are all the things I see across technology, people and process that will require transformation. It's a great opportunity. It's exciting times, but obviously a really good bit of work to be done.

Guy Daniels, TelecomTV (45:37):
Great. Thanks very much, Manish. Well, we've talked about all the new opportunities, but also cost reduction is at the top of every telco agenda as well. I'd like to end by looking at this and asking you, in what practical ways do you think AI can enable a more cost efficient telco? I wonder, Patrick, can we come to you first with some thoughts about the cost efficiency angle of the AI native telco?

Patrick Kelly, Appledore (46:10):
Yeah, guys, so just in the area of operating expense, one area that we see potential opportunities for AI and gen AI would be for field dispatch. The other is network quality. Those remain the highest cost factors for operators. And with the right data and statistical models, I think AI has the potential to drive down 30% of the total cost in a dispatch task. There's other areas again where in the care and billing areas where they're further along and we're seeing dramatic improvements both on the OPEX side and the ability to drive more revenue through customization and market awareness.

Guy Daniels, TelecomTV (46:56):
Great. Thanks very much Patrick. A lot of good real examples we're already seeing and hearing about there. Manishh, let's come across to you and then Danielle, I'll come to you next, but Manishh, what are your thoughts on the cost efficiency gains of AI?

Manish Singh, Dell Technologies (47:12):
I think there are two parts to it. The way I'd want to answer the cost efficiency gains are to be had, as I touched on through different use cases, whether it's improving productivity, improving customer care, reducing network downtime, improving network KPIs and more. And so the use cases are many, but I always get asked this question, Hey, what's the first use case I work? I say, Hey, here's a simple two by two, sort out your spend by CapEx and opex. Where are you spending the most? Then layer the second layer, what's your data readiness? Two by two, it'll give you the right use cases that you should start working in clearly. Field technician, field dispatch, as was mentioned previously, great area of opportunity, network troubleshooting, network engineer productivity and it's grain area of opportunity software development greater. So the areas are many, but the question comes down to where are you spending most of your money and how ready are you with your data to unlock that, right?

(48:12):
So that helps. That framework helps in identifying the right use cases. The second part also I just quickly want to touch is doing your AI the right way. Bring AI to your data data. Where is the data? 80% of that data is sitting OnPrem. And so bringing AI to your data, doing it more efficiently, unlocking those use cases in a very efficient manner is equally important to get those costs efficient outcomes. And that's one of our big focused areas to bring AI to your data, to bring AI on-prem, bring AI on the edge closer to where the data is and help unlock all of these use cases in a cost efficient manner.

Guy Daniels, TelecomTV (48:59):
Interesting. Thanks Manish and Danielle, what are your thoughts on how AI can help telcos become more cost efficient?

Danielle Rios, Totogi (49:06):
Yeah, I mean pretty much in every single area of the telco, there's opportunities to apply AI. And when you do it in the right way, you save money. I think Patrick said at the top of the panel, focus on the areas that are really bringing business value, where the AI application is really driving a cost reduction. And so in a very practical way, what I tell telcos when I'm talking with them is start with customer support. This is a huge cost center of the telco. It's usually multiple percentage points of revenue that you're spending in this cost center. And it is a sort of universally accepted great place to start with AI. And my question I asked telcos is what percentage of your customer support inquiries are a hundred percent solved by AI and what are you doing to increase that percentage every single day?

(50:02):
And I believe that number should be north of 50%, and it could be even as much as north of 80%. And customer support costs are really linear with headcount costs. So if 80% of your tickets are now being autonomously solved by AI, and now you're just paying for API calls to an LLM on top of your already pretty clean data, right? So great about customer support is the data is already pretty clean because people have been writing the answers to tickets. Tickets have been coming in online or coming in through different channels where it's largely written or you have voice transcripts, it's pretty clean and there's an answer there. And that's why customer support been working so fabulously well and it's been the early winner on the AI side. But the key is the autonomously solving of the ticket, not the copilot with a human is the quality bar and evaluating whether or not I'm going to use the answer or not, and they're copy pasting, that's going to give you maybe 20% productivity improvement.

(51:07):
But if you can actually get the AI to autonomously do the job completely and really drive that number up, you're going to see really, really big cost improvement. And from there, you can start to take your work in support and start to apply it to other areas of your business, like maybe sales or maybe internal finance queries that you're getting from your group. So start with the use case that everyone's really adopted and then roll it out to different adjacencies. But I think is the key is measuring the result and making sure that it really is translating into this cost savings.

Guy Daniels, TelecomTV (51:44):
Thanks very much, Danielle. Well, let's also go across to Rahul and Rahul. Let's get your views on cost efficiency.

Rahul Atri, Rakuten Symphony (51:52):
The best part about the AIr part is you can actually measure the efficiency, unlike other aspects which we've been talking about, whether moving to cloud is efficient or not, whether automation helps you saving money or not, what is the ROI aspects? But with AI, everything becomes so transparent and fluidic, you can actually measure the efficiency. And like everyone is saying, you can push for more. For example, CXO doesn't have to wait for 20 people to be in the room for the presentation asking the question, understanding strategies, how the strategies are playing out. They can just check the statistics to the last level, even the customer experience if they want to from one of the screens. So I think even that aspect of becoming a transparent fluid organization where everyone have access to the data is itself a huge movement forward. Think about a fresh grad who has joined the organization. Think about somebody moving from horizontally, from one department to another. Think about people moving organization. It takes a lot of time for them to understand the culture, understand where the document is lying and understand the processes. I think the huge transparency itself is great. And then everything is measurable. You can actually measure the ROI, which is super cool.

Guy Daniels, TelecomTV (53:14):
Absolutely is. It's an amazing range of possibilities that TURs are facing at the moment. We must leave it there. I'm sure we're going to continue this debate during our live q and A show later, and I'm looking forward to seeing what our audience has to say about this too. But for now, thank you all very much for taking part in our discussion. And if you are watching this on day one of our AI native Telco summit, then as I say, please do send us your questions. We really do want to hear from you and we'll answer as many of them as we can in a live q and a show, which starts at 4:00 PM UK time. And the full schedule of programs and speakers can be found on the telecom TV website, which is where you'll find the q and a form that you're going to need. And also our poll question for now though. Thank you so much for watching and goodbye.