Guy Daniels, TelecomTV (00:09):
To give this forum some logical structure, we're following the four assessment dimensions that we adopted for the ANTA Index. And so we're starting our sessions proper now with foundation readiness. That's quite appropriate. We've got a panel discussion on this topic after the break, but first of all, we've got two presentations for you. So first up is Guy Davies. So I'm going to ask Guy if you'd like to, which is very rare to find another Guy on stage. I've got to say, it's probably the first time ever. I'm not sure how I feel about this, but if you'd like to take the lectern and start your presentation, Guy.

Guy Davies, HPE (00:44):
Thank you. Good morning, everyone. First of all, I'd like to start with a little level set, if you like, as to what I think of when I talk about AI native cloud, AI native telco, sorry. Basically, we're talking about constructing networks that help you deliver AI, the high capacity platforms, the fabrics and so on. Also, using AI to assist you to operate those networks in the most efficient way. But ultimately the goal, as I think was expressed earlier, was your goal is to deliver a set of services that delight your customers. Ultimately, there's building networks and consuming AI has no value unless you can deliver a high value set of services.

(01:52):
So who are the consumers of the AI? So I see two main classes of consumers. The first are the machines, the networks, the compute, the WAN, the data consumers in order to consume the AI for networking. But primarily we want to be able to deliver services for people, for enterprises, particularly who are willing to pay for those services. And those fall across multiple domains, and many of you are familiar with all of those. So when you deliver services, what are the customers looking for? Well, particularly the machines, they want deterministic. I'd rather say deterministic behavior, not just latency, but there are many other characteristics that the machines want to be particularly deterministic. The next is high availability. And we saw last week what a almost catastrophic behavior when three of the foundation AI platforms all failed almost simultaneously. You had Anthropic, OpenAI, and Groq all went down same morning and the US is, "Ah, the sky's falling."

(03:36):
So high availability of the service needs to be taken into account. You've got to be able to scale, you've got to be able to seek the best cost for each of your tokens, and you've got to be able to deliver services across multiple modes. You've got to be able to develop different vision. We've got vision AI there. I think that's a kind of specific.

(04:17):
When it comes to your end consumers, I think service personalization is a big deal. And both from the perspective of the telco, because that enables you as a telco to enhance your ARPU that was mentioned earlier, the ability for you to sell the ability for other service providers to add value to your customers, but also it enables your customers to consume AI in a way that delivers value most effectively to them. So when I talk to telcos, which I do on a relatively frequent basis, what comes to us from a requirements perspective from the telcos is they want to be able to deliver these kind of AI services with a public cloud-like experience, a self-service, minimal intervention from their own staff. They need to be able to consume or deliver these AI capabilities both in public and private cloud environments. And I'll come to why you might want to do that in a minute.

(05:50):
When we talk to telcos, they also see themselves as a consumer of these services, as an enterprise. Many of the telcos we talk to are tens of thousands of individuals as an enterprise. They're substantial enterprises in addition to delivering services to other enterprises. So all of this has to be end-to-end automated in order to be able in effect to deliver the public cloud-like experience.

(06:28):
So what are the risks to telcos looking to deliver AI services? Well, the first is an obvious one. The capital investments involved in certainly building private cloud AI are enormous. They're an order of magnitude higher than your typical telco cloud compute. In addition, the energy consumption of these platforms and the availability of adequate energy in individual locations is a big challenge. When we talk to telcos, they're also concerned about, "Okay, I can order this stuff. How quickly can I get it? How quickly, once I've got it, can I start developing services? Will my staff be ready? Will they be capable of delivering a set of services that this equipment enables?" So the time to market is a big, big deal. You're making this investment, you want your money back, certainly your CFO does, as quickly as possible. And so if we're looking at delivering this, we've got problems with energy in single locations, maybe we need to deploy this in a distributed fashion.

(07:59):
And we certainly, as Ahmed pointed out, this is a multi-domain problem. This is not a single domain orchestration platform. So how do we deliver or manage this as a coherent environment when we're trying to manage across multiple domains and multiple environments?

(08:26):
So from our perspective, the ways we propose to deliver this are primarily to construct modular, scalable platforms, construct platforms that can be delivered initially on public cloud, that can be migrated to private cloud when the scale becomes larger because public cloud, the economics tend to stop working very well when things get big. Use vendor financial services. I work for a vendor, I would say that, but vendor financial services enables you to mitigate the cost as a single capital purchase, make it into an operational expense. With respect to the energy, the primary focus is consume modern, more energy efficient equipment. Each individual item, it will consume more, but it will deliver orders of magnitude, more capability. That's the only way, and that's how you guys have been delivering with 5G compared to 4G. You know the maths.

(09:57):
The time to market is an interesting one. We recommend using pre-validated designs. Talk to the vendors who are delivering the product. They are making designs that enable you, once you've got the product, to build reliable, repeatable, and scalable solutions that enable your teams to start work. They also deliver frameworks within which you can consume the necessary tools to be able to deliver services, which is ultimately your goal. Now for the biggest challenge, which has been mentioned already today, is the multi-domain orchestration. The only way I believe it can be broken down today is in a hierarchical model where each of the domains that you work in retains its own AI models, its own AI capabilities, and you have an overarching orchestration management environment, AI environment that makes calls into each of those areas in order to request capabilities, in order to be able to deliver services and to be able to function.

(11:38):
I see I'm running out of time. So this is just a demonstration of how that is presented end-to-end multi-data center with cloud, with compute storage, all the capabilities are handled. If anyone wants to see this in more detail, you're more than welcome to come and catch me at our desk. And the message really is that the network is fundamental to the ability to be able to deliver services that delight your customer. So then building those capabilities is the first stage, but operationalizing them is really the key to making this work for you and for your customers.

Guy Daniels, TelecomTV (12:38):
Thank you very much indeed. Round of applause. Come and join us for a second. We'll just get just a quick question for you before you go, but I just want to just follow up on a quick one there.

Guy Davies, HPE (12:50):
Sure.

Guy Daniels, TelecomTV (12:52):
The challenges, and the challenges are many.

Guy Davies, HPE (12:55):
Non-trivial.

Guy Daniels, TelecomTV (12:56):
Yeah, non-trivial. You mentioned the orchestration challenge there, as you said, that came up earlier in our session. Can you just explain a little bit more about what you mean by this hierarchical

Guy Davies, HPE (13:04):
Model? Yeah, sure. So if you look today, when I go and talk to service providers, I often talk about. I present the end-to-end image, which I just dashed through, where all of these capabilities are available. You've got a data center and within a data center you have an AI management platform. Within your WAN, you've got AI capable management platforms, you've got AI for your security, you've got AI for your compute and storage, and you can build an end-to-end AI orchestration platform. But when I talk to a service provider, they quite often will say to me, "I love the vision, but we've got that bit and I've got that bit, so I don't want those bits from you. What I want is to be able to take the bits, the white space, if you like, between what we do have from you, but fill the bits or use the bits I've got already and their management platforms to feed the higher level orchestrator." And so that's why I talk about having this kind of hierarchical model, because much as my sales guys would love to be able to come along and me give the story and the telco go, "Right, where's my checkbook?" I'll have all of it, that never is going to happen, never.

(14:41):
So we have to make this modular to be able to be consumed in a modular fashion, and hence the hierarchy is kind of critical.

Guy Daniels, TelecomTV (14:52):
Fantastic. Thanks very much indeed. Now at the end of this session, I promised some audience Q&A and Tony, my colleague Tony, is floating around with microphones, so we'll get to Tony a little bit later, but we're going to move on to our next presentation now. Ohad, please take to the podium and apologies for that extra microphone you got there if you can.

Ohad Barnoy, Amdocs (15:10):
Morning, everybody. Agentic-Ready Data Foundation. I'm going to try to talk a little bit today about how we see data as the basis for success in transforming into an Agentic operator. And I'm going to start just with a very quick number. So Ahmed mentioned this morning that don't offer me POCs. If you're a vendor, don't offer me POCs. And I couldn't agree more. And my first takeaway is don't ask for POCs. And if somebody's asking you as a vendor for POCs, chances are they're not the decision maker. So you need to find the decision maker. Why are POCs failing and not transforming actually to production? So I'll tell this through an anecdotal example that I recently had. I met with an amazing startup that offered me an AI solution to manage my pipeline better, accelerate deals, manage my pipeline better, and they wanted to do a POC.

(16:13):
They wanted to show me how I can use their AI in order to close deals faster, bigger deals for our sales team. I told them, "I don't want to do a POC. I want to do a pilot to production." And they said, "Okay, what does that mean?" I said, "Well, I use Teams for all my communication with my customers. We also use their facilitator, Copilot. I'm not advertising Microsoft, but this is the tool we use. Obviously, we use Outlook. We use a CRM to manage all the deals. And my salespeople where they're communicating with the customers, if they're in Europe, they're using WhatsApp. If they're in the US, they're using the Apple, don't remember the name. And if they're in Eastern Europe or APAC, they're using Rakuten Viber.

(16:58):
Now all I need is for you guys to integrate with all of these to see that you can actually have your AI use case work. And they basically said, "Oh, we've never done this before. We don't know how to do everything you're asking for." And that is when I told them, "Guys, just go sit down, try to do integration with Microsoft, try to do integration with one of them, and come back to me when you can actually do this." So in my view, this is why POCs are failing. POCs are failing not because they're not proving a value of a technology, it's because they can't work in our complex ecosystem. And that's why you need to demand and suggest, demand as a telco, suggest as a vendor to do pilots to production, and you'll see a significant increase both in the success and the adoption rates.

(17:42):
Next slide might get me kicked out of ANTA, so I apologize in advance.

(17:47):
What are we all doing today? Today, and I didn't think about Germany when I had the BMW there, but regardless, today what we're doing is we're taking all our systems, we're taking all our processes, and we're infusing AI into them. We're adding AI capabilities, sometimes strategically, sometimes tactically. Sometimes we're looking use case by use case, by use case, and sometimes we're looking across our entire gamut of operations. But what we're doing is we're doing AI add-ons. That's good. We don't have any choice. We have to do that. Where we should aspire to is to become a truly AI native telco. I know that it was supposed to be a Tesla, but I thought it might be too controversial because today some people love to hate Elon Musk and so on. The bottom line is when Elon Musk was interviewed about what's the difference between his EV and a German EV, which are amazing, I had a BMW i4 when I lived in Seattle and I loved it.

(18:47):
He said they took a BMW 4 and they changed the engine. I build it from the ground up. Now, we all understand we can't build our systems, all our systems from the ground up. But to Danielle's point from the previous session, if you're looking at 6G, for example, this is where you should be thinking, I don't want to take the processes that I manage my RAN in 5G and infuse AI into them. I actually want to build an AI native radio access network.

(19:25):
Let's just move quickly into what I mentioned before, and again, Ahmed mentioned it in the morning. 2025 and the beginning of 2026, we're all hysterical about using AI, infusing AI, cost reduction, revenue generation, better customer experience. And most of us looked into very tactical, focused pilots, projects that would drive business outcome. There's nothing wrong with that. It drives business outcome. You can absolutely use AI tools. I'll give the classic example for your conversational IVR for your customer engagement in order to improve customer experience and reduce calls that are coming into the contact center.

(20:17):
Tactically, that's the right approach that we need to take. But the more we look into where we need to think, we need to look at the process of why did the customer even call the 1-800 number? What caused them to call the 1-800 number? And can I predict that ahead of time? My entire customer journey, the entire process that I did before I even acquired this customer, when I acquired this customer, when I activated their SIM, when they connected to the first tower, when they had the experience, when they got their first bill, maybe they even walked into the store or did my digital, that entire experience for some reason caused them to call the 1-800 number. How do I take that entire process, look at it end-to-end, and don't automate it, but actually AI native it, if you want to think about that. So when we're looking at what we need to provide the telcos moving forward from an Amdocs standpoint, we're looking at how do you take these complex business and network workflows and actually look at them end-to-end and turn them into an AI native process.

(21:29):
And I'll give an example of a live production use case that we have, which is actually looking at what I just said. So we have a concept that's called personality engineering. The practical term is customer digital twin, where most of you are from the network, you know the network digital twin now think exactly the same, but about Ohad's digital twin. So imagine right here next to me is standing my digital twin, and he knows all my behaviors. He knows how I think, he knows how I act, he knows how I'm going to react to somebody saying to me the word Apple, yuck. He knows everything about me.

(22:14):
We can use this customer digital twin because you guys have all the data, the telcos. As Sachin mentioned from Rakuten, the data is yours. Our system may be the ones creating it. We may need to have to expose it to you as MCPs in order for your AI to be able to consume it, but the data is yours. So you have all the data about your customer experience. Yes, it's in different fragmented systems. Some of it is in the retail application, some of it is on eDigital, some of it is in the network, some of it is in your BSS. It's all over the place, but you have all that data and you can actually map. And this is what we did for. I'll show in a second, the example from the Canadian Tier 1 operator. We created what we call archetypes. Archetypes are personality aggregators of a person.

(23:08):
So for example, I am a cost saver. When you offer me something or when you talk to me about something, it has to be in the sense of why would I want to spend this money on this? What's the value for me out of this in some use cases? My boss, on the other hand, the one that convinced me to get the BMW i4, he's all about innovative capabilities, or in other words, wants to look good and so on. So when you're thinking about those traits, how does AI know to talk to me differently than to my boss? If I give the AI, the sales agent, the care agent, the virtual agents, that information, they can actually help me make better decisions. So these numbers don't look huge, but the actual outcome here is amazing. 9%, almost 10%, and by the way, I looked at the numbers this morning, this has been running in production for the last three months.

(24:13):
It's actually up to above 15% reduction in human handover in the group of sales agents and care agents, virtual care sales and sales agents that are getting the archetype of the customer. So when they know who they're talking to, not just the name of the person, but the personality, they change the entire conversation. And this goes back to Danielle's point about context, about understanding who the customer is.

(24:42):
The engagement has also increased by almost 10%. Now, what's the engagement? The engagement is this is a sales agent that's reaching out to customers and with a conversational IVR and talking to them. In this case, almost, I think it was about 40% of the customers were hanging up before even the greeting. Before the greeting took place, before, "Hi Ohad, this is the sales agent from Telco X, and I have this amazing promotion." 40% of the customers were hanging up. So we were able to reduce that number by 10%. So again, this is just by knowing and adjusting the greeting to the different personality of the person. These numbers are going to continue to grow. I wanted to use this example to show you how leveraging all your data and building the right data foundation across all your towers, leveraging context and leveraging ontology can actually drive significant business outcome in areas that are impacting both your customers and are not just focused on cost reduction, which we're all implementing.

(25:48):
Thank you very much.

Guy Daniels, TelecomTV (25:49):
Thanks so much, Ohad. Right. Excellent. It's the sort of CX we've been wanting and promising for years and years and years now, isn't it? And it looks like you finally have the ability to get it and make it happen. I do want to open this up to the room though. So if I cover my eyes, I can see Tony at the back. Hi Tony, you've got a question? No, I've got a microphone question. Okay. Right. Please help Tony out. There's a hand in the middle. There's a hand at anyone at the back as well.

Tony Poulos, TelecomTV (26:22):
We've got somebody.

Guy Daniels, TelecomTV (26:23):
Anywhere you can find, Tony.

Tony Poulos, TelecomTV (26:24):
Hang on a sec.

Guy Daniels, TelecomTV (26:26):
It's a big room. You've got little legs.

Tony Poulos, TelecomTV (26:28):
Thank you. I need your name and company. Mic close to your mouth.

Francis Haysom, Appledore Research (26:33):
Francis Haysom, Appledore Research. Great presentations this morning. One thing I haven't heard talking about AI is the concept of conflict. We talk about intents, for example, but there are lots of intents in an organization and they often conflict. I'd be interested in getting some view as to what you think you're doing about how AI can help that process of conflict resolution, finding the best optimum between many intents rather than just seeing this intent as being a single thing.

Guy Daniels, TelecomTV (27:08):
Thanks, Francis.

Ohad Barnoy, Amdocs (27:10):
I can try to take that. So I'll actually answer you with the example that I showed, and if that's not enough, we can continue later on. When we define archetype of a customer, obviously there's no one archetype that's a one size fits all. This customer is 80% cost saver and 20% something else and 30% something else. They have different intents. So we're doing two things in order to. First of all, we're intentional about how we use intent. And what we're doing is two things, right? And I think this is what everybody should be looking at is one, put the right weights and balances for each intent to make sure that you're identifying the right intent and the most important intent for that interaction point. So for example, if a customer is defined as a detailed oriented, like my friend Guy here, which is very detailed oriented, when he calls a care call and he calls about his wifi not working or dropped calls or anything like that, I'm not going to just give him an answer that, "Oh, the tower was down in your area." Okay?

(28:23):
I'm probably going to tell him, "Well, it looks like you were driving and going between towers, the tower that..." Or something like that may be a bad example, but you get the idea. I wouldn't just give him a very basic answer. I would give him a more detailed. But if it was a sales call and he's detailed oriented, I might not even look at the detailed oriented part of his intent. I would actually look at the fact that he's cost conscious and that he's looking for saving money here. So I think the idea is to put weights and balances for each one of the intents and making sure that you're understanding the right intent in the context of this specific interaction that you're having, both internally in the organization and with your customers.

Guy Davies, HPE (29:01):
Yeah. I think from my perspective, which is more down in the network, there definitely is a hierarchy of intent value, if you like. And to some extent, that's driven by business policy. Your business may decide that the intent to save money is the absolute prime directive, or it may be that a particular customer service value is the prime value, but that becomes a business policy decision rather necessarily than a technical one. The implementation is technical, the decisions about how those Intents are graded effectively on the metrics applied to them is business policy.

Guy Daniels, TelecomTV (30:06):
Great. Thanks very much, Guy. Tony, do you have a second question or are we all waiting for coffees?

Tony Poulos, TelecomTV (30:14):
Of course it had to be the other side of the room.

Guy Daniels, TelecomTV (30:17):
Don't run, Tony. Health and safety. We've got no insurance.

Sylvia Feng, Belden Inc. (30:19):
Hey Guy, I have a question for you. When you were talking about the determinist - Oh, I'm Sylvia from Belden. You talk about the determinist network and you also have the box with robots and physical AI. So can you talk more about how the physical AI is demanding the network to rise up to meet or how is our network needs to change and adapt to meet the physical AI applications?

Guy Davies, HPE (30:50):
Absolutely. And that's a really particular and key example, if you like, of where determinism becomes really important. If you imagine that you've got physical robots moving around within a space and they are being fed by AI that is not local to, not actually on them, then the latency budgets become a question of safety. For example, if you have to avoid the arm of a robot swinging and hitting somebody or hitting something, then the latency between the detection of the thing, the action choice and the action being implemented could be a matter of sub millisecond in order to avoid something fairly catastrophic. So that's where I see a lot of these physical machine driven imperatives being created with respect to the determinism. With respect to people, people are fuzzy. People don't necessarily have those same demands for. They don't have the demands for low latency. They have a demand for predictable latency.

(32:27):
So for example, with voice, if it goes above 300 or 400 milliseconds, it becomes noticeable and kind of annoying. Whereas 300, 400 milliseconds is kind of fine. Within the same country, you're good. So the scales on which we're talking about determinism become different.

Guy Daniels, TelecomTV (32:59):
Great. Well, thanks so much for the question. Look at that, exactly 11 o'clock to the second, fantastic timing. That is all the time we have for our first session. We have a 30-minute networking break. Coffee and refreshments are available in the lobby and do please visit our exhibitor pods. Let's show our appreciation for our presenters. We'll see you back here in half an hour.