From Data Quality To Autonomous Networks In Telecom

AI is finally forcing a scoreboard moment in telecom: some organizations are seeing massive productivity gains, while others are stuck between fear of missing out and fear of getting it wrong. From the floor at IBM Think in Boston, we sit down with IBM’s global CTO for telecom, media, and entertainment Eoin Coughlan to get practical about what separates “AI pilots” from AI that actually lands in production and earns trust.

We start with the hard truth that hasn’t changed for decades: data is the bottleneck. Clean, timely, governed data determines whether AI helps you run a network or quietly amplifies bad decisions. From there we move into telecom operations where fragmented observability makes it hard to see what’s really happening. We talk about pulling signals into a unified view, using AI to correlate root causes, and keeping control as you introduce agentic AI. Autonomous networks come up as a real journey, not a magic switch: time series models for network telemetry, multiple agents that can read tickets and vendor manuals, and then automation that begins with humans in the loop and expands as trust grows.

Then we zoom out to the ecosystem: hyperscaler dependence, rising sovereignty requirements, and what it means to run compliant, air-gapped platforms that enterprises can rely on. One of the biggest opportunities may be hiding in plain sight: SMEs often trust their telecom provider more than software vendors or hyperscalers, opening the door for CSPs to deliver packaged AI assistants and managed platforms. We also hit legacy modernization and 5G monetization realities, and finish with what might surprise us next, including early quantum computing use cases. If you found this valuable, subscribe, share it with a telecom leader on your team, and leave a review with the AI or automation challenge you’re tackling right now.

Transcript

Welcome From IBM Think Boston

SPEAKER_00 0:00

And it's Evan here in IBM Think in Boston, my neck of the woods for a change. And we're talking with the global CTO for telecom, media and entertainment IBM, and lots of interesting discussions on the table on how operators are building, running, and monetizing their networks and where the industry is headed beyond the hype. Owen, how are you?

SPEAKER_01 0:22

Very good. Thanks for having me here.

SPEAKER_00 0:24

Well, thanks for being here. You're based out of Ireland, so Boston must feel like a second home. Before that, um uh meeting you for the first time, uh, who are you and what do you do exactly within the global innovation powerhouse that is IBM?

SPEAKER_01 0:40

Sure, yeah, as a double CTO in the telecom uh area uh for technology, I get to look at our portfolio and match that against the needs of telecommunications all over the world. So I have the privilege of meeting quite a lot of our customers all of the time, and understanding their business, which is a key element of how you can look towards um you know adding AI into your business, uh managing that hybrid cloud infrastructure, uh, how are you going to monetize uh assets that you want to bring to market? So that that whole kind of process works right across operations to enterprise sales, everything. So we're we're helping all different departments right across their businesses.

Why This AI Wave Is Different

SPEAKER_00 1:24

Oh, it's an exciting time uh at IBM and for customers. And you've been through many waves of uh technology innovation, uh hype cycles, if you will, in some cases. How does this AI wave feel different from past industry, let's say, hype cycles uh in telecom?

Data Trust And Unified Observability

SPEAKER_01 1:43

Yeah, so I think the the interesting thing about this is that we have clear evidence from some of the AI first leaders that they're getting significant benefits from this up to 70% um productivity gains for them, that they're developing projects and delivering them 67% quicker than they used to. So that's causing an acceleration for those companies. Uh, and then we have the other companies who are you know concerned about missing out on AI and they're concerned about messing up AI. So now what we're looking at is we're trying to bring AI to them in a manner that's consumable, uh usable, uh, and that they can trust it. All right, so that's the key. So we've brought quite a lot of capabilities. We've seen challenges down the years, which haven't changed. Data is still a massive challenge, probably even more so now as we look at AI. When we bring solutions to market, you need to have the right data at the right time, the right quality that you can trust. So IBM has been investing heavily in building What'sNex.data and also in acquiring companies like Confluent, which was the$11 billion uh investment that we've made, in order to make sure that we have access to real-time data all of the time with the trust that's needed on that. And that's how you build use cases that will deliver. Um, otherwise, you you fall into certain traps of not having the right data, making decisions based on the wrong data, uh, or just not being up to date. So I think that's uh that's one area. Um also then you know, we're helping with operations. We see uh across the telecom operations and IT operations in our customers that they have disparate systems which are providing observability at a certain point, but what they struggle to do is to bring all of that together uh into one unified dashboard that can use AI to help them figure out and correlate what the real issues are and what the priorities are to fix in their networks and their enterprise and IT. So that's uh that's uh one of the challenges. And we have all of these different areas, so we're helping them build agents, manage agents, agent sprawl is a big thing. Um we saw ourselves when we built our client zero, um, which has now yielded$4.5 billion in productivity for us.

SPEAKER_00 4:04

Wow.

SPEAKER_01 4:04

Um, over the last uh you know two and a half years. I think we've been working on that. That has taught us a lot of lessons. That's led to some of the acquisitions, it's led to some of the product choices that we've made. And that's why we're focused so heavily on hybrid cloud, building the AI capability uh and managing data right across the piece, because that's how you build towards value.

Autonomous Networking With Multi-Agent AI

SPEAKER_00 4:26

Wow, what an amazing portfolio of really exciting stuff. And um, one of the uh you know buzzwords out there is autonomous networking, autonomous networks. Sure. Uh tremendous opportunity for operators for CapEx reduction, OpEx, NetEx, and beyond. But what what um what are you seeing there in terms of the reality of that journey, challenges, and uh how close are we to a true autonomous network?

SPEAKER_01 4:50

Yeah, I mean it's fair to say that it is a journey because there's a lot uh of work to do to get to that point. Uh we talked about data, getting all of those data silos, uh, getting the accuracy of that data, even finding the missing data that's you know built across legacy platforms for the last number of years, all of that needs to be managed. Uh, and then you need to build the uh AI capability to process that. And what we found is that when we looked at uh large language models and how they could handle that, uh they struggle with um time series data. So we worked with uh our research teams and we've built a number of time series models which are available in open source. And then we've also moved to the next step of saying, I think people are gonna want an agentic framework to run this because you need to be able to decide uh which agents are looking at the time series data, which agents are now going to look at ServiceNow tickets or other ticketing platforms to find out what's happened in the past, other agents that are going to read manuals from the likes of Ericsson and Nokia, etc., and starting to figure out a plan of what is the anomaly that the time series is identifying and what is the fix that's required. Has it been done before, or do we need to build one? And of course, then we also link that with Bob, where if we want to build one, we can build a solution. Um, and the important thing I think is once you get to that AI uh insight, you want to then move it on to automation. So uh we acquired a company called Pliant a number of years back, uh, which is now a key part of our concert platform, which provides that automation capability. Um, so that's that's moving us you know from managing that data streaming to processing the time series data, generating the solutions that are required for it, and then activating automations that can fix issues in the network. All of that can be done with human in the loop, or it can be automated further. Um, as you go through trust cycles, you build up enough trust to say, look, if this happens again, you know the fix, let's automate that process. So that keeps control over everything for you. But what's important is that it's you know a multi-agent and orchestrated platform so that you have that capability.

Sovereign Core And The Hyperscaler Balance

SPEAKER_00 7:13

Wow, impressive. It's great to see all of this data turned into real business intelligence, actionable intelligence that's taken us a long time to get here. Yeah, let's talk about the ecosystem, the uh landscape. You know, you've seen this evolving relationship between telecom and MSPs and cloud providers, hyperscalers, and that sort of dance continues. Uh, how do you see that relationship evolving over the next couple of years?

SPEAKER_01 7:39

Yeah, it's been a it's been uh an interesting relationship because there's a you know, I think at this point there's quite a high dependence as well. We have the challenge that's come into play now as well, then in a lot of areas around the world where sovereignty is a big question mark and whether they can depend on the hyperscalers to provide that sovereignty for them. Um, what we're seeing is that a lot of our customers are very interested in, or we announced a product yesterday called Sovereign Core. And what that provides is a software platform which will sit on whichever vendors' hardware that you want to roll out as your AI factory landscape as such. Uh, we will build a uh an based on open source uh a platform on top of that, which will allow you to build and govern across the infrastructure. But you'll be able to build your own applications there, you'll be able to bring applications in, you'll be able to put IBM software in there, third-party software in there, and that's all air-gapped and protected. And it also has like 162, is the current count of compliances against it. So it's constantly checking for compliance. If you were ever audited by a you know the regulator to say, can you prove that this platform is sovereign because that's the regulation? The platform is built to take those reports all the time. So it's available at the touch and button. It's no problem. So that's a really good kind of starting point to be able to build a sovereign platform. But what's really interesting for telecoms, I think, is that this provides them uh the ability to also add tenancy. So they can bring in their enterprise customers, provide them with access to this infrastructure, but also then they can choose from different offerings of software uh that they might want to bring in. We did an interesting uh study um in March with the GSMX intelligence, um, where we looked at the SME market because that was 90% of the market if you if you count it all up. And what we saw is that they actually trust their CSPs more than they trust software vendors and more than they trust hyperscalers, and that they would be very willing to purchase um more from them if they had available um capability. So when we look at things that we did uh to build agents that are managing HR, managing finance, managing marketing, those kinds of assistants could be made available by the CSPs to all of these SMEs, which would mean that they wouldn't have to go out separately, buy software, you know, integrate it, manage it, etc., um, and wonder which ones to buy. They can trust their CSP to be their CIO office assistant as such and provide some of these agents with their business. And even as far as you know, providing them with the capability to do performance management on their own connectivity or across their cloud, if they have cloud environments. So those kind of software pieces they can manage themselves and and consume. I think that's a quick road to market for the enterprise divisions of CSVs and much needed revenue that people have been looking for as optionality for years. So yeah, we're very excited about that one.

Turning CSP Trust Into SME Revenue

SPEAKER_00 10:50

Brilliant. Can't wait to see that unfold. Um, speaking of the future, we're also looking back at telcos a lot of legacy infrastructure, perhaps more than any other industry, technical debt to say the least. What are some of the biggest mistakes you think operators are making in trying to modernize all of that legacy that's out there?

Modernizing Decades Of Legacy Code

SPEAKER_01 11:12

Well, I think that um sometimes the decisions can be made easier um by different solutions, like we announced uh IBM Bob, the show here. We're using that with many customers, and it's looking back through code for the last 30 and 40 years that they've been building. And in some cases, it's the first time that that code has ever been documented. Wow. So it's creating documentation, it's creating architecture diagrams, it's creating uh reasoning across what those programs are doing, but it's also looking at them from the perspective of governance and security, because everything to do with Bob is always premised upon security and governance. So it looks at all that code and it can suggest modernization. It may suggest keeping some of that code as it is, but it may also say you should actually, if you translate this into a different language, it might help. If you change this to uh parallel programming rather than sequential programming, you're going to get a performance increase. Um so it'll look for all of those gaps and it'll help you with the whole modernization process because it understands the context of it. It's not just an LLM for programming, it's a partner for development. So you can use it to ideate around what you're trying to achieve, and it will ask you questions as well to make sure that it's uh it optimizes its own internal queries and use of LLMs or small language models as well, and that allows you to build your structure, your plan with the partner development kit, and then it'll help you to roll that out and manage it over time. So it's uh it's a real uh leg up. Our own developers are now at 80,000. They're using Bob all the time and wouldn't give it back if we tried to take it off them. Um, and we're seeing an average there of 45% productivity gains in our development, and that's a huge help to us because, like every software company, you always have a long roadmap and you can only create so much software in a year. If you can get productivity gains at 45%, you can start to roll out a lot more features, functionality, and value for your customers.

Monetizing 5G Beyond The Obvious

SPEAKER_00 13:17

Amazing. And I see around IBM Think, Bob is everywhere. It's free to try. There's QR codes floating around. That's right. So uh why not jump right in? I'm gonna try it myself about telco. But uh, but it looks fantastic. In addition to modernization, you have the challenge of monetization, lots of pressure on telcos to monetize that huge investment they made in 5G on the consumer side and look for more enterprise use cases. Um, where do you think the biggest use cases and opportunities still exist?

SPEAKER_01 13:47

Well, I think what we're seeing at the moment is in the AI side, if I look at enterprises and SMEs across the board, they are looking for partners to help them to, because like they don't have the CIO office that's maybe the size of a telco department or IBM or others. So what they want is they want a partner who's going to provide them with AI capability, the ability to build their own uh platforms that they offer their services on using an Agentic framework, for instance, and they need that type of uh help from a partner to provide the platform from a hybrid cloud to the tool sets to the management of those assets, and all of that can be provided for them by the outgoers. And I think that's a really good opportunity to use the relationship with those customers to develop new revenue streams. I think there will be other revenue streams that come. We always see it over time as technologies advance, we see more. I know when we we often talk about 5G and we say that there's not enough use cases out there, and I think I'd agree with that. I we would love to see a lot more, but the the fact is that if you turned off your 5G network now, yeah, I think you'll find your 4G network won't be as performant as you think it is, um, because it can't handle the same capacity and the same speeds that the 5G networks can. So there is a little bit of a misunderstanding maybe of what might happen if you were to take that step backwards. Um, so I think 5G is paying for itself in that way. You definitely want to be upgrading your networks on 5G purely for capacity, you know, cost per gigabit perspective. Um, but there will be more and more use cases. I mean, we can see the consumption of social media and other uh platforms is gone up. I do remember the days when the iPhone came out, people said you'll never watch a movie on you'll never watch a football match, the ball will be too small, but that's not what you see on the buses and the trains, right? So Yes, yes, indeed. It's an exciting time from a technology perspective.

SPEAKER_00 15:50

It is. And speaking of exciting, uh, I'm not gonna ask you for predictions for three, four, five, ten years because no one can tell. I can't tell next week what's gonna happen. But what what changes in this industry, telecom in particular, do you think might surprise us uh on the upside?

SPEAKER_01 16:06

Yeah, I think the solutions that are becoming available to telecoms providers now will help them with the productivity that they need and hopefully give them some of the headroom that they need to reinvest in the business side in order to expand their own portfolio of capability. Um, I think the changes in regulations um with regard to sovereignty, I think help their cause. Uh, they're a natural um partner to government and uh they're very familiar with regulations and they have the technical capability and skill set, and they are certainly used to the scale that is required to build complex platforms. So I think they're in a good position to take advantage of a lot of that requirement over the next while. Um, I think when I look at the business in general, AI is going to allow us to create a lot more products for them and for them to create them for their customers. And then when I look a little bit down the road, I see quantum computing is on the horizon. We're doing a lot of work now with uh with our customers, uh including in telco, looking at different use cases which will use the specific types of mathematics and probability that quantum leans in very well on and can answer questions that we simply cannot do with uh standard technologies. So we will see more use cases coming there. Hopefully, we'll find you know better, better uh materials uh using material science and quantum uh in order to find batteries that will last for more than 24 hours on our phones and all of that kind of thing. So there's so much more that will come, I think, in that space um between AI and quantum. I think we're gonna see some exciting times over the next five years.

SPEAKER_00 17:55

Indeed. And certainly the most exciting time in the industry in my 35 years. So I'm always looking forward to what's next. Thanks for the chat at IBM Think. Great. Thanks for being here. It's really good. Thank you. Thanks everyone for listening and watching. Take care.