Tony Poulos, TelecomTV (00:16):
Hi, Tony Poulos here at TelecomTV's DSP Leaders World Forum in Windsor. We've just had a fabulous session talking about data foundations for AI-native networks and cloud native beyond the core. And I've asked a couple of experts, in fact, three experts to come and help me cover the topics. Firstly, I've got Beth Cohen, who's a telco industry analyst at Luth Computer. And then I have Diego Lopez, who is a senior technology expert at Telefónica and at ETSI, he's an ETSI fellow as well. And last, but certainly not least, Patrick Smith, who's the EMEA CTO for Everpure. Ladies and gentlemen, thank you for being here. I'm going to ask the most relevant question for today. What is the problem with telco data? All we've heard about today is problems getting that telco data right. Beth, you want to kick that off?

Beth Cohen, Luth Computer (01:04):
Oh, absolutely. So many years of telco dealing with telco data. First of all, much of the data is 40 and 50 years old and it is, let's say, very dirty and all sorts of crazy stuff. And setting aside customer data, which is pretty dirty, you also have the fact that the data is not in any consistent format. So therefore all the data that's for the wireless network is in one format and the data in the wireline network is in another format and all the data in the MPLS network is in yet a third format and you have a mess.

Tony Poulos, TelecomTV (01:49):
That's the starting point.

Diego R Lopez, Telefónica (01:51):
No, that's it. I mean, this is that the telco networks are highly heterogeneous and that implies that the data is highly heterogeneous. If you think about it, we are dealing with different subjects of communication, voice, data, video, I don't know, machines, whatever. And with many different technologies that from radio, optical, IP, whatever. And that implies in each case, implies that the data that you're generating are naturally disjoint from the other data that you have. But you have to relate them because at the end, what you see is something that is end to end. I want to talk to you. I don't care whether it's MPLS or fibre or is the radio. And when there is a problem, you have to navigate all data and if you want to apply whatever smart processing of the data, you need to have it in a way that is understandable, even for an expert.

(02:48):
I'm not saying for a machine or whatever.

Tony Poulos, TelecomTV (02:51):
I'm going to touch that now. Patrick, do you want to add to that before we get onto the difficult question?

Patrick Smith, Everpure (02:55):
And I think that highlights it really well. We often talk about data silos, but within the world of telco, you have business silos as well who each own their data. The key to improving data and data quality is a cross-business data governance approach. And that's very hard to apply in a telco world because every stovepipe likes to look after their own environment. And so you don't have that commonality in data governance, which means data hygiene, data quality, the structure of the data is different in every world. And therefore bringing it all together to get value and monetisation out of it, it's close to impossible.

Tony Poulos, TelecomTV (03:36):
So is that the fragmentation you heard about today, that data fragmentation?

Beth Cohen, Luth Computer (03:40):
I want to add even one more layer, which is we're all talking about within a single telco. The data fragmentation, if you want to do real AI and if you want to really take it to the next level, you need to have standards. Those don't exist today.

Tony Poulos, TelecomTV (03:57):
But for years we've been building data warehouses back in the day, data lakes. How many times we talked about standardising the data? Why haven't we got to that point yet?

Diego R Lopez, Telefónica (04:07):
Because I guess that's because the focus on that cases was about where to drop the data and not because taking account that right now we are not talking about accessing the data and it's about I put my data there and then I collect my data or well, I tell you how you could take part of my data to look them. Now it's about we want machines that are whatever we call them, that they are models and large or whatever, they're machines and they are quite picky about the things being well structured, understandable, processable, et cetera. And then we're talking about a different layer of understandability. Even for humans, it's complicated, as I said before, but for machines is something that is basically impact right now.

Tony Poulos, TelecomTV (04:56):
But Patrick, is AI going to fix that? Is AI going to sort out the data for us or does the data have to be sorted out before AI will work with it?

Patrick Smith, Everpure (05:02):
Well, and I think that's a really interesting point because data is the enabler for AI. We heard about AI projects this morning. Actually, what we didn't talk about so much is AI projects that fail. We were only talking about those that succeed and most of the AI projects that fail fail either because actually there was no real business value in the first place or secondly, because the data doesn't exist to support that AI initiative and the business outcomes, or it's not there with the required quality. And so to successfully drive AI projects, we have to dramatically improve the data landscape.

Tony Poulos, TelecomTV (05:42):
But the owners of the data, are they guarding it as well within the telco? They're saying, "I don't want the billing information to be let loose on my customer or whatever." Is that a big issue?

Diego R Lopez, Telefónica (05:50):
Go to the customer. You probably don't want your billing information to be shared with your controlling departments or with the fiscal ... It will say it is because we are humans. I mean, we're talking about humans. We're talking about my piece of power, my piece of my position that corporate pyramid. And this is something that is complicated to break as well.

Beth Cohen, Luth Computer (06:12):
That is, that is. But we also do have a problem with data sovereignty about the network itself. And again, I go back to telcos have been around for a long time. They've been working on the ... They have that sense of how their network works and they have that set of data that has been built within that silo. And to talk to one telco talks to another one, they're literally talking the different language.

Tony Poulos, TelecomTV (06:44):
Yeah. But building use cases must be difficult if you've got to get data from multiple parts of the operation to get a result.

Diego R Lopez, Telefónica (06:50):
No, no, it is.

Tony Poulos, TelecomTV (06:51):
But are we the only industry that has this problem? I mean, we heard some of the speakers talking about other industries having data issues. Are we unique at having these multiple data problems?

Beth Cohen, Luth Computer (07:00):
No, but we're unique in that we have to share it. We're peering. I mean, telcos have to peer. Financial services don't have to peer. Years ago, I worked on a project for the medical records industry. It turns out you need exactly three sets of data to create a medical record for somebody. You need their test results, their pharmacy records, and their insurance records. That's it. But for telco, you need the billing records, you need the usage records, you need everything far more complicated.

Diego R Lopez, Telefónica (07:40):
There are an industry that probably in the future will suffer the same situation that we are suffering right now that is these hyperscalers, cloud providers, whatever, because they have a common substrate of infrastructure or they used to have because to have different pieces of infrastructure and very different kind of customers and workloads, et cetera. Right now, well, they have the cool guys that are doing the new thing and they manage. Let's see in 20 years.

Tony Poulos, TelecomTV (08:08):
The cool guys. There's a few people who questioned that term as well, Diego. Look, is it realistic for operators to complete the journey and unlock the full promise of cloud-native networks? We talked about our cloud-native operations. A couple of speakers said, "Oh, that's the easy part." And then of course Neil McCrae up there said, "No, it's not. I tried. It's really difficult."

Beth Cohen, Luth Computer (08:27):
Been there, done that.

Tony Poulos, TelecomTV (08:29):
How realistic are we and how far are we from having it?

Beth Cohen, Luth Computer (08:32):
It's very difficult. It has to do with the telco's structure as a very infrastructure heavy industry. I mean, it's almost a utility. I mean, telcos pretend they're not a utility, but they are. And because it's so costly to maintain that infrastructure, you want it to last as long as possible. That's not how some of the cloud providers are not doing that, although as they mature, that is actually going to become more of a problem for the cloud providers. So it's really hard. I mean, there are telcos that are still using mainframes to maintain their customer databases because it's so darn expensive to bring those records over.

Tony Poulos, TelecomTV (09:22):
Patrick, do you want to add to that?

Patrick Smith, Everpure (09:23):
So I think the expansion of cloud native beyond the core is an interesting one. The core is such a ... It was talked about in the panel, a homogeneous environment and the cloud-native platform that supports it is it's like a Formula One car. Everything is stripped out so it can just support that one workload, business critical workload. As soon as you expand out beyond the core, you have to support a plethora of different workloads all with different requirements. Suddenly you're bringing in stateful requirements, not stateless requirements. You need to manage data, you need to provide infrastructure level disaster recovery rather than application level, disaster recovery and resilience. So a whole load of dimensions come in as you go beyond the core delivering a cloud native platform.

Beth Cohen, Luth Computer (10:10):
And you're also going past DMARC zones, right? So you're going into customer prem locations, you're going out to IoT devices, you're going across industries, you're going into places you don't have sovereignty over.

Tony Poulos, TelecomTV (10:28):
So we're going to be talking about this for a couple of years to go, I suspect that this-

Diego R Lopez, Telefónica (10:32):
It's an aspiration like many others. For example, I remember sometime ago, people started to talk about friction-free economy and the friction-free economy was supposed to get us free of these cycles in which you were entering, growing and recession and things like that. Well, that has not been the case, but probably you could think that the subprime crisis was way shorter than it could have been.

Tony Poulos, TelecomTV (11:00):
Maybe you'll have to come back next year to tell me if we're achieving it in any way. Beth, thank you very much for being with me. Diego and Patrick, all of you. So good to have you here. And I think we've solved a few problems and created a few more.

Diego R Lopez, Telefónica (11:11):
Well, that's it. That's it.

Tony Poulos, TelecomTV (11:12):
Thank you. Thank you.

Beth Cohen, Luth Computer (11:15):
Thank you.