Ground Truth Survey: Why Network Data Limits Autonomous Networks

Brilliant. Let's get started and welcome again. Thanks for joining us. We are here to talk about our IQGeo Ground Truth Survey that we've done telecoms industry and all that means and of course getting your questions about the results as well. So by way of introduction, my name is David Cottingham. I'm the Chief Technology Officer at IQGeo. But with me, I have two people who know far more than I do about fiber and all of the things that we're going to cover in the survey. So Tom, could you take us through where you're coming from and what you do? Good morning. Good afternoon. I'm Tom Maguire. I was most recently the CEO of a company called Brightspeed, which was a carve out in the States of a incumbent local exchange carrier, primarily a copper footprint. And over the course of the last few years, we've built out close to 50% of the territory with fiber. And so I think that's prior to that I worked for Verizon for a long time. I think it's my experiences are, you know, kind of appropriate to discuss here because a lot of what we're focused on is really, you know, my life story growing up in a copper environment and eventually moving over to a fiber environment, which is a tale of two cities. So I very much look forward to having the discussion today. Brilliant. Thanks Tom and Amaury from the other side of the world. Amaury from Switzerland, So I'm working from Switzerland. So the place to be for snow and to enjoy the snow. Next to the snow, what is important in my life is also Swisscom. We are also here part of the telco companies, quite innovative. Our strategy is to offer the best experience to our customers, the best network. So and the fiber strategy is part of it. Despite of that, I'm leading the Department of Network Strategy by the company Swisscom. Brilliant. Thank you. I said two people who know infinitely more than I do. So I'm very glad to have them with me today. We wanted to start off very briefly with why on earth did we call this thing the ground truth survey and it's really important. There's a lot of discussion about autonomous networks and how we can use AI to automate things and that's great. But to automate your network, you do need to have, as we say garbage in garbage out right you do need to have a good overview of like what's your network, where is it how is it connected together and you need to make sure it's accurate. So no surprise. If you think about that sounding basic. Well, it is basic, but if you think about how many miles of infrastructure people have it's like, there's a lot you could kind of get a bit wrong, right? How many tens of thousands of engineers work on that infrastructure every day and then you work out it's actually quite difficult to know the latest state, not just you know the state once upon a time, but the latest state. And so, as it says on the side as operators are moving into investing in AI that's great. But actually every decision that AI takes is going to relate and depend on that physical network data that lives underneath and you know a couple of examples right. If you have got space in a duct and great you can stick another cable on it, if you haven't then not good which actual strands are have something running on them versus ones that don't do you have capacity leftover. If you have a maintenance visit which nobody documented what they did are you going to end up finding the next technician comes along and they think this strand is unused and it's actually used right and then bad things happen with support calls so. it's why we call it ground truth, because having that up to date view of what's going on is super important even more so now that we're talking about AI and I think a lot of people top of mind AI thing yes true. What we're saying is AI is wonderful and important and top of mind, but you have got to get the basics right as well, in order to deliver the value from AI. Now, who did we talk to? So we went and we surveyed some really senior people so 500 decision makers in the telecom space, two thirds were with senior managements - that gives you a sense of at the top of those companies, what do people think about the importance of ground truth data, AI and so on, so we'll talk about all those things in a minute. And so there's really big companies so 1 million or more subscribers and bigger than that real spread of different operators here, you can see the bottom we've got. everything from the big kind of fixed broadband all the way down to I'm a municipality and I've got a network that I need to run. So lots of responses, really good cross section of the industry. So that's where the data has come from and, of course, what we like some of these things is get some more data and we're going to ask you so we're going to share a poll that will pop up, that basically says, how accurate do you think your network records are. tell us what you think. Is it under 10%, is it three quarters plus? Love to see what you think so we'll give you 30 seconds to chew on that and vote. There's plenty of votes coming at the moment, which is great. Have a think and see where you think you're at. So, results are still coming in, that's great. I would say there's a lot of people, two thirds of people, under 75%, so that's telling and we've got just under half of people where half the network records they think are accurate. So that ties well actually with the data we'll talk about in a minute. So a lot of people who are definitely convinced that their network data is not as perfect as they would like, I think, is the way of describing that. So, good, thank you. Let's go into what we call the ambition. So what are people trying to get done. There's lots of talk, as I said about autonomous networks. I think a lot of people are saying 2030 that sounds like around number it's a long way away. And that's the end of the decade so I'll say i'm going to get that by the end of the decade, I think there are some people who thought perhaps more than that. Here is where people think they are right now, though, so when you look at operators who say i'm fully autonomous, I think that's pretty ambitious, but there are 10% of people who think that so well done to them. Lots of people, if you look at that conditional autonomous. 50-66% those the three bars above manual, you're heading into two thirds territory where we're doing some partially autonomous/conditional autonomous, but highly autonomous means you're actually trusting the machine to make the decisions for you and only raise exceptions. That's really, really ambitious stuff and, as it says in that box 24% of operators serving quite a number of subscribers report that they're highly autonomous - not really surprised, they've got a lot more riding on this and presumably also they're investing more in it. I'll give you one other slide here, this is the 2030 day that I talked about. That's the question of this slide. I think over half of people say we're going to be ready for before 2030 and I said, I think that's just because 2030 feels like a long way away. And then, there are some people who are perhaps more pragmatic or more scared and not sure which. About a third of people say post 2030, some of this is also because they recognize that full autonomy is a big deal so there you go, those are the stats. The key point, though, is that basically, everybody is going "yes, autonomous networks are the decision that we're going to take, the destination we're going for", the never necessary crowd are are 2%. So I wanted to get the views from people as I said, who know far more than I do, so Amaury why don't you start us off with - what do you think of the data that you're seeing here? Does it surprise you that the majority of respondents think autonomous networks are actually realistic before 2030 and if you do, then why? So I think it's quite realistic before 2030 I think that would be quite ambitious. I think the change that we have for the example of Swisscom at the moment we are in the level of partial autonomous network. So yes, we are using a lot of it and but the way going forward will take a lot of time, energy costs and you pointed out for me the key question is quality of data. So we can have the best plans that we have, the money could be even limited if the quality is not there, then we are not going to do anything with it. And our target for Swisscom is by 2030 we want to be highly autonomous networks level. So it's just at the middle between you know before 2030 or after with the level of Swisscom predictability we like to be right on time. We will really achieve this by the end of 2030 we want to reach this level four. So level four is the highly autonomous one, isn't it? Exactly. And for those of you who are familiar with the tele management forums at TM forum that's why that level number comes from, so they have a nice map. The level five is the bottom one here and obviously you can count down from there. Tom any thoughts on this one in particular? You know what does it have to be the case, but what has to be true in order for people to actually get to this autonomous networks thing? And we talked about the data, as well, but anything else you think has to be true? The thing that kind of strikes me if I go back to the universe of individuals that were surveyed, I think we have a tendency, at least in my experience, to be very positive about things. And the higher you go in a company, the more positive people have the tendency to be. So, for example, I think two thirds of the people that you interviewed were senior managers and I'm curious what some of the lower level folks may think, the people that live with the data every single day. At the recent American meetup I think I mentioned when I started in the business as a technician, I remember having an interesting discussion with somebody in our records department and this isn't I work for an I like. And I was telling them I was standing on a pole looking at a terminal that they told me didn't exist. The pole wasn't there and the terminal wasn't there, so apparently I was floating in the air. The other thing that kind of struck me was that about 50% of the survey respondents were either an MNO or a fixed broadband like a fiber provider and a relatively small number 13% were an ILEC. And so, I think that not only where you happen to stand in the company, but what part of the industry you happen to be in will impact your results. One of the things that kind of struck me about the 2030 date, which I think is very ambitious, is we've been trying to migrate customers off of copper now for the better part of 15 years and we're nowhere near where we need to be. And I think those in the traditional telecom space that are saddled by copper that's where the majority of some of the issues lie with respect to data accuracy. The newer technology you have, just from our experience in Brightspeed, our copper data was just about as good as anybody else's which isn't all that great. But our fiber data is much better because we were able to embrace something early that puts us on a good path of full autonomy. And so it depends, do I think that people in newer technologies that focus specifically on newer technologies will have a somewhat easier road to get to the level of full autonomy? Yes. Do I think that people who are still burdened by older technologies will continue to have issues? Yes. I think that also raises a really interesting point that there's been even with Greenfield, so new people on the block they've rushed to roll out these things. And in some cases they've been very careful about it and they tried to gather data and other cases has been like just get it done as fast as possible. And I know from conversations that I've had with people that they're saying, well, now we're not paying the price, but we know that data needs improving again in the greenfield in the last few years. So I don't think it's not just new technology. It's also how fast. One of the things that I think was kind of interesting, I used to kid around and tell people that I was drawn to the industry because everything was very, structured, very regimented, very documented. And so this way, if something happened, you go to the practice and you pull it out and it tells you what to do. Unfortunately, one of the things that I think a weakness of the industry is that everything is structured, everything is regimented, everything is well documented. And when stateside providers started rolling out fiber networks, they just mirrored whatever they did on the copper side because they had done that for 100 years. And so I think to your point, a lot of people that went out there and built things quickly, they built things quickly relying on some of the older systems and older approaches or processes that they had in place. And so, they didn't really set themselves up for success with respect to moving to an autonomous universe. Again, it all depends. I think you hit the nail on the head. It depends on where you start that will help determine how difficult your journey will be. It's Switzerland, Amaury, it's all perfect, right? No, not at all but we like to be on time. I think quality goes first and that you can really measure it. And in this old transformation, when you think about quality first, it's not about the old style of doing business. Copy paste of the copper, what Tom just said. I think here we started with just process reengineering. And we took really the time to reinvent, to reassess our processes. And that's really the function we need to start from this one. It's not going to an equipment provider or software provider and just listen what he could tell you. But it's a way reverse. Think how you could really review your processes and then derive what type of tool you need. We did one example in to the fiber rollout using the function of AI base for quality control in the field. And this was really doing a really a revolution of our way of controlling quality in our construction processes. And here we did really a good job with the product of extra diplomatic companies where we have been able really to triple the quantity of what we produced. without increasing the quality control and keeping an aim on a target of having more than 99% first time right activations of customer. So that's the way we work, quality first. I told you Switzerland was different. We just think different. But quality has a price. That is extremely important because if we are not reaching the high level of quality at the first time, then you are going to clean all these problems after the rollout by the time of where you are going to activate those customers. You have to go a few times on site. This it's not directly priced into the contractor prices, but you will pay it anyhow on another day without talking about customer satisfaction. And people hate that to be disturbed three times just because the fiber is not fully well built. So that's really reversing our ecosystem, starting with quality, and then think based on that, what do you need to achieve? Absolutely and I think also, by the way, when you add AI to the mix, AI is great at making decisions fast it'll also be really good at making bad decisions fast if we don't feed it the right things, so. more to come on that. Let's skip forward. A bit of piece about the gap basically between where we are, where we want to be. So let me talk a little bit about that. In terms of how many. So the responses, a lot of you i'm sure are part of this survey. How many people actually trust what's in the network records and you've already told us right there's a the the server responses on the poll rather pretty similar I think. About half of the people basically, half of their records they don't trust. That's pretty much identical by the way to what the the poll just showed us. That is kind of scary because at that point you apply AI and the decisions are not going to be where they need to be, they're not going to be accurate. So, do people want to fix it? Funnily enough, yes, they do, and I think fair enough. Everybody thinks this is important to fix that data, but why here are the things where AI would give, if we are having good data, really good value back. So got to fix the data first, and then we can, for example, automate field operations, or we can accelerate planning design. We can reduce this operational cost. All these things hinge on that data quality piece, so there is a lot to do here and a lot of game to be hunt. But as you know both my guests have talked about it's not an easy thing to do. So, I'll flip back actually, on the gap piece. Tom, do you agree that 13% is kind of a true picture that I mean only 13% think that you know three quarters of their data is accurate? Is it that most people don't actually know what the number is at all? Because I've heard some operators that don't know. It seems very reasonable to me and again I think it depends on where you're coming from. It's interesting, it says what they expect to deliver the greatest value and the first thing that they mentioned is improving data network accuracy, That's nice. But to an Amaury's point I would think that if you have accurate data that's going to enable you to provide better customer experience and drive down your operating costs. So the biggest value to me is don't have accurate data just to have accurate data. Have accurate data to do something that's going to benefit the company and that's where customer service and that's where more efficient operations come into play. Not to beat a drum here, but I think that if people got themselves into a certain position with respect to data as opposed to re-engineering the processes as more you're alluded to earlier I think that they're going to be in a position where they back themselves into a corner with respect to accuracy and now they're going to scramble to try to fix it. I've seen that happen before. One of the things that we did was kind of interesting is that we bifurcated the business or split it and we were still operating under a TSA at the time, so we didn't do much to focus on improving that rather we focused on the future of the business which was going to be fiber and we completely reinvented the way we do things with that. And so, if you were to ask me you know the question what do I think my data accuracy level is, I would say with respect to fiber it's very high and with respect to copper it's very low. So it depends on where you're coming from and where you started out and it's pretty evident on the prior slide that there's a lot of people that are scrambling to catch up as opposed to doing what Amaury mentioned a little while ago. Come on Amaury just tell us what your thoughts are in terms of whether that 13% is true. And I guess the region it's also very different, right? It really depends, but I like Tom's approach, copper is low fiber is much higher. It's a very good summary, I would definitely sign this statement, because our copper inventory is just hundred years old. So it was paper based inventory and it has been digitalized and we did utilize only what was still used, which means a big portion of our inventory is not digitalized, it just remains on papers that nobody knows where they are. So, when we are doing type of project like pulling out the copper, kind of copper harvest at the end of the migration from copper to fiber, then we are trying to get out also copper. We get a lot of good surprise that we find more copper cables as what we really have. So, which means that we underestimate the value that we have under the ground. It has some positive signal. But on the other hand, I can tell you our CFO doesn't like this type of uncertainties. So, I think the summary of Tom is just great, fiber is much better than copper inventory. The 30% or not 30% it depends so much. I would dream to get more than 90%. But we need it. And then it's real dream for 90 copper, right? 90 to fiber is your dream and then we'll see about the copper piece. Fiber is the dream, copper is the nightmare. Tom, you said it. I'm not going to comment. I've said a lot worse. Let me move us on to why. We know we've got the problem, we're sitting around in a circle, we admit we've got the problem. Why does it keep happening? And you've touched on a lot of this already. But three quarters of the people who respond to survey basically say there is no single source of truth, which obviously makes it really hard, swivel chairing all that kind of stuff. The purple piece, I think, is perhaps the scariest. The orange piece like across the middle of multiple systems, yes, we understand people have different generations of systems or acquisitions and so on, but it's digitized. The purple piece, it's still on paper. We start things in CAD, but some people really do have things on paper. Lots of people still operating with spreadsheets, even at a big big scale. And then you know 25% of people who've arrived, but probably only arrived for fiber. 13 years, how long people have had their their existing network management software package running and that gives you a sense of the multiple systems one as well as these things stick around for a while and there's lots of data in them. And then the next piece is, you don't have a mobility solution, you end up with things being done in the field and they might be scribbled on a piece of paper with a sharpie or something. But then it may never get back to the actual GIS. Some people in the survey saying it's immediate and I think that told you they're not using paper and a sharpie. One to two days, don't know could be between within a week, probably heading to paper certainly within a month and then over a month. There's clearly a backlog. So 48% of people still send their crews out with paper notes printed maps. I think there's this also I can understand that there are sometimes cases where people are wearing gloves and things and you don't want to give them a tablet or whatever it is. We've seen that, but obviously there is technology now that can make that much easier, but if you're sending people out with paper, then I guess the changes may come back on paper as well. So, Tom three quarters of operators have no single source of truth, so, is that a technology problem, a budget problem that you know they can't afford one, or is it culture? Because we all know culture is really hard to fix. Technologies are easy. The answer is yes. It is all of the above. Craig asked a question about our inherited records on the start and that's a copper based question. I'm not sure that everybody remembers TIRKS. Most people don't even know what TIRKS stands for. But you know that system still exists today. Trunks inventory record keeping system, and it's been around longer than I've been around. Not by much, but it's been around longer. That's still in play. I'm not aware of anybody who has really come up with a quick graceful way to get out of TIRKS because things are just so embedded in there. So, we have multiple systems. That's a multiple systems issue. People have developed over the course of time some workarounds for certain systems. We ran into some of that, where people kept records because they happen to serve a particular geography, but they didn't go back into the bigger system. And so that gets into I guess the purple section. And, so you have a combination of these two things. And, people really want to do the right thing. Another I think key characteristic of the industry is that everybody wants to do the right thing. I don't think anybody wakes up and figures out how can I go out and make customers miserable. And so they always want to do the right thing. And so from a cultural perspective, it's okay. But, in today's day and age, especially since the marketplace is a lot more competitive, it's tough to get money allocated to turn around and update a system that's been working for a gazillion years. And so it's really a combination of everything. It's like multiple complicated systems that are not easy to get off of. It's not having enough money because there's never enough money. It's interesting, I think Verizon, AT&T are consistently some of the top two or three companies in the States in terms of investing back into the business. But there's never enough cash out there to do everything that needs to be done. And so, there's a bunch of different reasons. And what we did, we didn't really kind of invest in too much in terms of like trying to figure out what's happening with the older network because, we eventually want to get off of that as quickly as possible. So I guess we kind of simplified things a little bit and we started from scratch to a single integrated platform. And so, half of our business, the small portion of our business is where we need it to be. And that puts us in a good position with respect to moving into, higher level autonomy, but the other stuff there I think we're going to have to suck it up and feel the pain for a bit until we can get everybody move migrated over to the other stuff. Because, we looked a lot at trying to figure out how we were going to get off of some of the legacy systems that are out there, and it's just a really heavy lift. A lot of the people that are very familiar with those systems are no longer with us. So, it's an interesting game. But it's a combination of everything I think that's contributing to where we sit today. I think the summary of that one last piece is it's complicated. It turns out that making changes to systems that have been around for a long time  it's really, really hard. Amaury, only 7% of people see the changes. If you go out, make a change, those changes reflected immediately in your system record, but only 7% see that immediate change. So, why? Where does the time go for everybody else? We talked about paper, but in your experience what's going on? On our side, it's really nearly live. Maybe not on the second level, but it's nearly live when we activate a customer, so between the time where the field technician is able to say quality is okay with all these systems, the light is going through then the service can be activated. And so we have even pushed a situation where we have instant migration systems where once the job is done, once inventory is also done, once the picture of installation is really done, we have green light. So whole thing goes back to the inventory and you are able to activate systems. So, if this is even nearly close to the okay within the next 7-15 seconds. Your world sounds great. I mean you've been around the industry a while. What's your feeling for why you're taking less than a minute and some people are taking a month. It's huge work. So, we started to roll out fibers since 2008 and there we did a decision which was we are starting with a new world with new IT system and we use a present stock and the copper stays with old one. And then we directly invested in a new world for the new system, but I like the question just before with okay, the number of system is huge, how do you do a single source of truth. And this one we solved it or we try to solve it with a data lake. So we are putting out all these systems and integrate that into a lake and don't ask me what is the quality of the data. It's really also stays a challenge between copper and fiber and what is interesting into the view of Swisscom is that so we say that we invested a lot in fiber since about yet now more than 18 years. We are not yet done to finish our roll out. We invested on alternative technology to prolong the life of copper, which means we pronged also the IT side of this, using data and using system and not being able to take out all systems and now we see a whole strategy about autonomous network, we really take it seriously to take out most of the old system and that's why it takes a lot of time to reach 2030. So and all what Tom you said, I fully agree with you. We have also the life cycle of the people, even the name of the system you said I don't know it. I'm a bit too young to know it. I feel harassed. But, there's systems, that like almost is another one out there that we use for the longest period of time. It really is a tale of two worlds because, If you look up technical debt in the dictionary, it'll probably say, something related to copper. And because to Amaury's point, even if we tried to get a little bit more life out of the copper networks and did some other things in order to leverage it to serve broadband, we just took whatever stuff we had and just made it that much more complicated as we rolled forward. And, what we did with fiber is and this wasn't an easy lift because I had a lot of interesting conversations and we stood at Brightspeed. We had a lot of people that were industry veterans. And, I remember having interesting discussions about when we wanted to engineer our fiber network that I wanted to use a GIS based system and people are like, oh, no, you need to get an engineer out there. You need that person in a car with a clipboard walking around, staring up. And I'm like, no, we can't do that anymore. You can't because of all the issues you have in terms of the record lag. I remember having field organizations where I go into a construction office and I would see a bunch of plats sitting on a desk and I would ask people what's going on with this and they're like, oh, those are red lines that we put together a while ago and we have to send them up to engineering. When did you do them? They've been sitting there for a couple of weeks now. We haven't gotten around to it. That doesn't exist in our fiber world because we're able to use either the web based browser or an app and the construction people could walk the job, do a red line, and that becomes the as-built and that becomes the inventory. So now, it's a lot simpler in the fiber world because we just started from a different place. And to Amaury's point, I think, having a data lake is good. It was no such thing as a data lake back when we were doing copper type stuff because it's kind of tough to take a bunch of vellums or paper plats and stick them in the back. But, it's like trying to catch up with literally 100 years of records, keeping and trying to figure out how to move them into the future as opposed to. OK, we're not doing any of that. There is no paper. Everything's electronic. Everything's going to be done immediately. It's just a different world. So we're going to be saddled with that, at least from a wireline perspective until we get away from that copper and that copper network. I think the best thing to do is just get rid of copper, get rid of the costs of copper. I'm talking to some some other folks about that. Luckily, we see a little bit of regulatory relief in the states coming from the FCC, but in the United States, if you get down to state specifics, a lot of them are still like you need to have copper. And as long as we're saddled with old regulation and things like carrier blast resort obligations that's going to continue to force us to be in the orange piece of this continuum. And I think what's really interesting it's on one hand, yes, we can put a new system in for the new fiber network. But actually also getting rid of all the old systems and just simplifying down, not having the orange bar effectively is really, really important as well. So let me shift us on. Obviously, if you don't have good data it's not just that the people who worry about data get upset, there's also other things that it impacts, so what is it that impacts? Slide basically says it all really. It's not just in one area. It's lots of them. So project delays, yes surprisingly, if you've got to go and roll out a new piece of network and your data about the existing network isn't right, that's not good. We've touched on service activation, no surprise, I think if a truck roll is $200 at a time, I need two or three of those to get some activated. That is not a good statistic for your margins. Getting close to customers, no surprise outages are resolving them or causing or resolving them, I guess, and of course regulatory issues. I thin, if nearly 70% of people said if we don't have good data we're going to have issues with a regulator, I think we often overlook that one, but it can end up being quite expensive to Tom's point just now. And then, another one, as everyone knows there's lots of smaller alternative network providers around. We're seeing consolidation in the industry. The way those alt nets are valued in a big chunk of that is around basically can you prove that your network is what you say it is, where it is, what's capacity what its usage is and so on. And so you see the respondents here basically nearly everybody says that the value of my company will depend on how good my data is. 70% say it's critical or major, no surprise. Lots of people who want to invest want to go to you and say really have you got it? By the way, there was a an alternate I think in the UK where they didn't have good quality data and it turned out that they had rats eating their fiber cables and so tell up that their network wasn't worth very much, but they didn't know it at the time. And it was only when they went to audit it they discovered these things. So data really matters. Amaury, let's start with you on this one. Obviously we know bad data causes lots of issues. That's a given. But how much does that occupy you, how much do you end up talking about that in your business? Because, if all these things are true, presumably there's lots of conversations. Or is it that actually people don't think about making the connection really bad data and these things? For me, I talk about it each day. Because decision is based on data, and my main concern is what I am going to see correct and easy to type matching with kind of reality. And so here you didn't mention it, but for me, we use a future goals decision based on data so we know all when bad quality in bad quality out. And then for me it's if we go for the decision if you go for the most network, it has to be a topic. And even in strategy environments, we are starting really to use decision based based on data and here it's fully clear. And the first question when I get the report is what did we should get. So it's switch on our brain, be critical and ask all as a test questions before starting really to decide and I think that's the way we proceed, or at least I try to proceed. Tom, I wanted to get your view on this next slide which you're familiar with. Acquisitions and buying other people's data kind of buying other people's network and the data that comes with it. What's your experience of how important the data quality is for the acquisition price, but then you're acquiring these things, how many of them say they've got amazing data and it turns out that they haven't and you find out much later? It's kind of interesting. Not telling tales out of school, but if you go back, say five, six, seven years, it was kind of interesting because companies that offer fiber services were really valued based on how many households they have and at one point in time a fiber household was worth like thirty five hundred dollars. And I think there's a couple of things that have happened in the industry over the course of the last five six years. Back when we started looking at things and this is why I think this chart is very important as we look forward because in my opinion there will continue to be consolidation in the marketplace. I think it has to happen. And so but back then, things were a lot simpler. You didn't have any competition from LEOs, you didn't have a lot of competition from fixed wireless. And so customer decisions were pretty simple. Either I'm going to stick with the local cable provider or I'm going to go to what effectively is the incumbent local exchange provider, the local telecom. And like was simple, and the faster you got fiber out there, the greater the likelihood of winning. And so, that it was all about fiber, less about some of the existing records, because fiber is like if we could build it fast. if we could build it economically, that's a great thing. Since then, I think people have started to recognize that the business is a little bit more complicated, and especially if you have a lot of old stuff sitting around in that orange segment of that line that previous slide. And so I think nowadays it's not so much about how many fiber households you have, but what else is going on in the business. And so, I think people are starting to consider what are the costs, right? Because the cost associated with a copper network, if they exceed the revenue you're pulling in from that's a problem, and you need to figure out how to rectify that problem. So, I know for a fact that certain people that were looking for fiber, they were looking for fiber builders. They didn't want to be, well, I don't think anybody really wants to inherit a company that has a heavy copper burden hanging over its head. And so now people are starting to look into what's the quality of information regarding that older network, whereas five, six, seven years ago was like, who cares about that? We're going to build fiber and everybody's going to be happy and customers are gonna run to us as quickly as possible. Those days are gone. And so I think people are taking a long hard look now about what is it that you guys have or what information do we need in order to properly assess how the business is going to get to where eventually it's going to become some sort of fiber powerhouse. And so I think you're going to see major influence or critical, now I know what's up around, it's approaching a 78%, it's going to be even more important as you go forward. Makes sense. Got a final bit on what's changing and, by the way, people have questions, please do post them in the questions panels as we go along. We're very happy to try and cover those as we go. In terms of what's changing so. Visual AI is one piece of the AI puzzle. If I look at how many how many companies require those technicians to take photos. Basically everybody does. 9 in 10 operators require field photos. Sounds wonderful. The problem is that about 1 in 10 of those photos actually get looked at and so really what we're doing you know we talked about paper. Now we've got a digital filing cabinet. People are taking photos and they're putting them away very carefully. Some of them less carefully, it's a mess of digital storage, but they take photos. That must be good, right? Most of those don't ever get looked at. So we're ticking a box. And the question really then is what should you be doing with those? You should be using AI on those photos so that you get value out of them there and then. This isn't just about ensuring your technicians have done the right job, although, of course, that's massively important, it's also about getting the data back into your system, so that it's correct. To a data problem again. So, 58% of operators that audit photos now use AI, but if we look at the distribution of that it's very, very stark in terms of who's using it where. The US right now not really using anything like as much as the UK, Canada and I would have seen Europe, not on this slide but as well, so there's a big divide going on here. I want to get a poll started with everyone on the call, though, because maybe this is wrong, maybe actually you know our survey data is not representative of you, so if we can do the poll, do you look at them? Do you take them? Are you one of the 1 in 10 that actually does something useful with them? Do you use AI? Where are you at? Love to get your views. So, pick an option. Let's see how far we get in terms of responses and the balance. Plenty of people require photos. That's great. We saw in the survey results that we take photos, but we don't really look at them. Honestly, I think that's probably the open secret everybody knows. Everybody takes photos and nobody looks them in great part. We have about 20% of people don't require field photos. That's also interesting. And we have we take photos, but we review them manually, that's about a third of people. So, definitely a good spread there. In terms of using AI enabled analysis, the answer is nobody at all on the school. So clearly something to think about there. Let me just shift now to Amaury and let me get your view. Why do you think that so few operators are or to see all those photos given I know you are in this 11% I guess not the 89%. We are using photo like crazy. But we don't look at them manually. Never in live, because to activate the line we are taking about 11 photos on the customer premises so in the apartments, in the basement, in the manholes. And we are doing about I think about 200 checks. We source photo over the lines, so we don't do it manually no way. And so we fully use AI for that and also it's fully embedded in our processes so that's really something you didn't mention is what do we do with it. Because using AI is nice, but why? We are working with total contractors, which means that we are doing internal is a design and then the total contractor is doing all specification detail inventory and building. At the end of this process, we want to have customer activated. And this is the key. He is not going away of the place, he's working without having green line from his device. So devices are simple phone. But he has to take those photos so device tells him did your job well or not, and he has no right to go away without having the job done. And we have an exclusion process where his own boss can on the remote level force an acceptance if the tool itself makes a mistake, but you see at which level, so there are a lot of photos which are done. We are not looking at it manually and the tool himself tells you job done yes, no, you can go for where yes or no, which means that the payment of the job which is done by our total contractor is also driven by this quality control. Which means the whole chain is looking for quality and even the field technician likes it. At the beginning, it was quite tough to make them accept, but at the end, they like it because nobody looks at if they do it really bad or good. They are just learning on the field and the machine tells you yes, do it again, it's not fully good. And then the job is done so he can automatically also learn. So you have not only the pressure to leave a job which is done, you have also the chance to learn on the field without having your boss next to you who tells you here you did it bad. So the computer tells you rather than a human that you can embarrass in front of. I do think also the value of knowing you've done the job and then not having to come back three hours later driving all the way back to the same house to bother someone who's not gonna be happy with you is super important. Tom, I wanted to get a quick two minute take on why is the US the 45% bar here and obviously Europe and Amaury just mentioned, UK, Canada as well but why the US being down here? I think because there was a euphoria of people taking pictures of everything and it's like once you take the pictures nobody ever looks at it so it's like you know you look at your phone and it's like you know I have 93 pictures of cats and what the hell am I going to do with them? So it's a matter of trying to figure out what do you use these pictures for. And I think that some folks are kind of struggling with that. To an Amaury's point we use them to make sure the technicians were installing things properly. And you know they couldn't leave unless they did certain things by whatever they were doing at the customers prem that they were doing them accurately. But it's only recently that we started. I'm actually going to do a little commercial for an IQGeo product called NetLux and it's only recently that we started looking at this. So now, all of a sudden we have a reason to look at these pictures. Because using NetLux on polls we can anticipate make-ready work costs. I know somebody was asking about building out the network and how much did record accuracy impact. This is a big thing for us because we can take pictures of polls. And we can anticipate before using AI whether or not we're going to have significant make-ready work, and the associated cost with that. The other thing that we're using it for is to identify people that are not a compliance. So you know surprisingly some carriers or some providers out there may not do what they are supposed to do with respect to poll attachments. So we can find out where we're falling short of requirements with respect to polls, but we could also look for people who are attaching to our polls. We have like over a million polls, but we could look to see people who are attaching that we're not charging them for those attachments or if those attachments are improper and they're going to impede us from going out and building. So it's like all of a sudden we have a purpose for those photos and we could run them through NetLux and we could figure out whether or not we have an issue that we need to work on. So I do believe the 45% is going to increase as we start to see more opportunities to improve our efficiency and potentially pull in some additional revenue because people aren't paying us rent for the polls. It's a combination of are you doing the job right, which I think is super critical. I think the training element of that is also very good because, there aren't as many supervisors as you need out there to train people so to utilize AI to help train those folks I think is a benefit. All the way up to t's the stuff in the field and then it even gets to the compliance piece because we've had regulators that question us about the way certain things were run. And now we're able to prove to them that you know those aren't our facilities, are somebody else's and if you don't like them that's who you need to go after. So once people start to see the benefit and how it's going to impact everything that they do, and it's going to improve their ability to kind of build out additional fiber or maintain the infrastructure that's there I think you're going to see more, you'll see that 45% creep up. Tom's called it, says he reckons 45 is going to come 70. We have a couple of minutes for Q&A. If you haven't already please do see your question in the Q&A panel that there are some there. While people get the questions in if you need to get the summary or the full findings. They're available at this link and you will get those emailed to you as well. So weird networks - how do you document ongoing maintenance moves and changes? It depends which changes we mean. When it's network change, we put that directly in the inventory of Swisscom. When it's about cable change, when it's device change and we have all the inventory and then it's also putting in our system. But I also think anytime we do a maintenance change or something else that's out there you got to capture a new set of pictures to reflect what's out there. Every time you touch a network, I think, is the way we've always thought about it. Pictures are cheap and we know that. Everybody takes them, but you need to use them. Let's do one more question, because I know we're coming on time. Tom, how you at Brightspeed working on improving data accuracy and Amaury talked a lot about his side of the house, but i'm just interested to get your take. So copper not so much. I don't think I haven't talked to anybody about this recently, but they're not going to invest a lot of money in trying. They'll have a data lake they work, they're looking at things as Amaury described in terms of like access to data make it a little bit more accessible or easier to attain, but I don't think people are really kind of investing a lot in terms of improving accuracy. I think they're more focused on what they're going to do with the fiber network because that's the future. So at some point in time, copper needs a DNR and then you just walk away and let it do whatever happens to it. Thank you. I know that there have been a couple of other questions in the panel. We will certainly answer those offline. Thank you very much for coming along. Thank you to Tom and Amaury for a very insightful discussion. I said they have proven they know far more than I do, so thank you for being here and thank you for just giving the commentary on the survey and of course reminder you'll get that in your email or you can download it as you wish. With that we're going to stop. Thank you very much, and we'll see you on next one of these. Thank you. Thank you. Thanks. Goodbye.