Ray Le Maistre, TelecomTV (00:05):
We're at Future Net World 2025 in London. I'm here with Thomas van Briel. He is Chief Network Architect at Deutsche Telekom. Thomas, thanks so much for joining us.

Thomas van Briel, Deutsche Telekom (00:15):
Great to be here with you.

Ray Le Maistre, TelecomTV (00:16):
Good to see you. So this event, it's really all about, or mostly about ai, APIs, network automation. How would you describe Deutsche Telekom's AI strategy as it relates particularly to network operations?

Thomas van Briel, Deutsche Telekom (00:34):
We strive for autonomous networks as to other operators as well, and we see that the combination of all the good traits of putting the right foundation and then putting AI on top is really a match made in heaven. So we see a lot of potential to apply it across our operation. Let me just point out two aspects. So we go very much for a value driven approach, so looking very structured across all our value chains and seeing where's the biggest value tip we had in which network domains or which part, and basically it's the operations where we have the biggest leverage. That's one aspect. The other aspect is that we then bring together also the shift from the classical ML ops or also getting better accessibility to our data, whether Jen and I now to agent AI and really build the power to build agents to substitute human processes and we try to do that as quickly as possible, possible board and instil the right skills into the organization to go for that.

Ray Le Maistre, TelecomTV (01:42):
That's actually something that you're not just talking about, but you are actually doing, aren't you in the mobile network with the Google alienated? Exactly. Okay. How is that progressing? What are you seeing as a result of the deployment of that particular tool? Is that living up to expectations?

Thomas van Briel, Deutsche Telekom (02:01):
Yes, of course we are learning. We are seeing that we really have good results. I won't go into KPIs that share right now, but it's something that's really working. The difficult thing is really how to put it into operations so that you can keep it properly in the lifecycle that it's coexisting with all the other processes that are going on. So that's where we are still learning a lot, but we see that if you go for clearly defined business problems that are sufficiently narrow, then it's also manageable to go through the learning curve comparably quickly.

Ray Le Maistre, TelecomTV (02:37):
Are there any other particular telco AI use cases being tried out in particular parts of the Deutsche Telekom Empire for example, in certain European operations where you are trying it in one country first or even part of a country to kind of prove it in and see how it works before rolling it out in other markets?

Thomas van Briel, Deutsche Telekom (02:59):
That is a very usual pattern that we have. We have communicated also at the last Capital Markets day last year that we strive for higher synergies across the footprint and that's one of the basic patterns that we apply. If it's of interest, you ask what is then something else that we look into where we see a big value. It's definitely in fiber rollout where we see that we have made good progress to really bring the data together across all parts of the operations or we have a good foundation and then for example, substitute doing human work to reconcile the actual construction progress taken on with a camera on site with bolts being sent by a partner and then matching. Now is the depth of the digging really to what we see on the bill and is the actual progress in line with what we expected and then bringing them up out of line situations. These kind of things are accessible to applying agents. We are going for these kind of use cases. Just as an example.

Ray Le Maistre, TelecomTV (04:06):
Now you mentioned the data there and obviously the data foundations are absolutely key here. What is Deutsche telekom's strategy around data management and how you are collecting and using this data and what needs to be put in place before AI can have a real impact on operational efficiency?

Thomas van Briel, Deutsche Telekom (04:27):
Okay, two parts of an answer here. We experience in practice the most specific, the use cases that we go for, the easier it is to make sure that you have the prerequisites. And I would not subscribe to a statement that says we only can leverage the power of AI if everything in terms of data is right there. Having said that, going for broader use cases to more powerful use cases requires that you really get your act together on data. And we really try to apply industry practices that are not even limited to TechCo with the right data management, data governance, having discoverability of the data in place with a defined quality so that as a consumer you have an easy task and as a producer of data you take on the accountability to make sure that that is available for any type of use case that needs to be there. So that is in my expectation, anything from special, bringing an organization across the board to that level is a challenge that we are working on.

Ray Le Maistre, TelecomTV (05:45):
And then obviously as this progresses and the use of AI becomes more prevalent in the telecom sector, there's a lot of talk about certain levels of or advances that can be made, but how can you measure that? How can the impact that the use of AI tools is having, is that something that, is there an industry standard for this or does Deutsche Telekom have its own way of measuring the impact of AI?

Thomas van Briel, Deutsche Telekom (06:21):
First statement on a specific use case? You better know what you're striving for and you can, if you, for example, go to the wrong guardian agent, we can very clearly say before with that many people, we were able to specifically cater to that many events afterwards. We can increase that by factor 10 or factor a hundred in terms of taking action for specific events that happen in network. And you can count that and measure that. You can also then reconcile what is the actual impact on customer experience. Do we see that the average experience that the user has in these events has improved? So that is possible, but you do it per use case to really optimize it going forward. And that's also a local responsibility of the DevOps team that take care of this specific, this case. Second direction on looking at the question that you are asking is how do you identify where the value is and what you should go after? So we follow very pragmatic approach where we basically build the metrics across the whole value chain across all the network domains, allocate our spend, our FTEs, and then identify with expert knowledge what are the high value fields that we should go after and describe for those what is the goal that we want to reach. So that's answering the question, how do we identify what we really should do.

Ray Le Maistre, TelecomTV (07:53):
Okay. I was just wondering if this is something at the moment that, for example, the way you measure and the way you model that, is that something that you've developed maybe different from other telcos or is there a kind of...

Thomas van Briel, Deutsche Telekom (08:15):
We have looked very closely at what Team Forum is doing what also key players on there doing China Mova, something like that, and have done some things that are similar and then some things we deviate a little. We see that it's quite challenging to come up with repeatable, accurate industry frameworks to measure progress in terms of autonomy rates or something like that. It is doable, but it's very, very tedious work and as an industry as a whole, we are still very much in the beginning as far as this is concerned, so tend to be very pragmatic and rather solve for how do we make sure that the use cases that they're striving for actually deliver the value and being able to identify, we very closely watch what's going on there and adopt more advanced practices as they become mature. And I think the approaches that we have are very much compatible.

Ray Le Maistre, TelecomTV (09:17):
Yeah, I guess the thing that everybody wants to avoid is just using AI because it looks like you should or because it's exciting, but then you need to have a tangible gain, don't you? At the end of the day. Okay. Definitely a work in progress. I can feel that sense that from the conversations here at the event, a work in progress, but at least things are happening now in the sector and things are changing. So Thomas, great to talk to you. Thanks very much for taking the time and look forward to speaking to you again in the future.

Thomas van Briel, Deutsche Telekom (09:46):
It was a pleasure. Thank you.