AI’s Real Payoff In Telecom

Your carrier has more data than almost any company you interact with, yet most telcos still struggle to turn that advantage into growth. We sit down with Miguel Carames, the Chief Product Officer at Mobileum to sort out what’s real, what’s next, and what’s pure hype when it comes to AI in telecom, 5G monetization, and the future of operators as intelligence-driven businesses. Along the way, we get honest about why “we invested billions” doesn’t automatically translate to new revenue and why regulation and privacy expectations reshape every AI roadmap.

We also challenge the idea that AI only arrived with generative tools. Telecom has used machine learning for years in automation, anomaly detection, and capacity planning, but the story hasn’t been told well. Miguel shares concrete, production-minded examples: using LLM-style interfaces to make deeply technical testing platforms usable for roaming managers and analysts, moving toward automated root cause analysis, and deploying agent workflows in fraud and revenue assurance so cases arrive pre-analyzed with evidence and a human still making the final call.

From there we go into customer experience, where proactive network intelligence can prevent tickets before customers ever feel the pain, and into churn reduction, where the opportunity is huge but the privacy line is delicate. We wrap with fraud and security, the whack-a-mole reality of bad actors, and what it takes to escape pilot purgatory so telecom can move at AI speed. If you found this useful, subscribe, share it with a telecom leader, and leave a review. What’s the best AI use case you’ve seen a telco actually scale?

Everyday AI: Your daily guide to grown with Generative AI
Can't keep up with AI? We've got you. Everyday AI helps you keep up and get ahead.

Listen on: Apple Podcasts   Spotify

Transcript

SPEAKER_00 0:01

Hey everybody, I'm really excited for this conversation talking about uh AI and telecom, what's real, what's next, what's hype with a true insider and innovator at Mobilium. Miguel, how are you? Good, Evan. Good morning. Good to see you. Good to see you. Uh, for those who may not know your background, tell us about your company, your journey, and uh what led you to getting in the telecom space.

SPEAKER_01 0:24

All right. So uh Mobilium is a uh value-added services and analytics company. Um we're a mid-sized vendor, software vendor. Uh, we are a global company. We are uh deployed pretty much in every MNO around the world. Like nine out of every 10 operators are a customer of one or multiple business units, which for a company of our size is no small fit. And uh so yeah, I've been in the industry overall for 26 years, three and a half odd mobility running products here. Uh before that, 13 advertising in different roles. And before that, uh Moderola for nine. So 26 overall in Telecom.

SPEAKER_00 1:05

Well, quite a journey. Uh so let's dive right in. We all know you uh more than anyone that telecom is shifting from a connectivity-driven business to something else, uh maybe intelligence driven. But uh, what is the shift? What's happening, and what are you seeing on the ground?

SPEAKER_01 1:23

Yeah, so um what we are seeing is um, and it's a common thing we're hearing from others too. So operators have made uh a lot of big investments over the last few years uh on 5G, fiber, and and other assets, and the monetization hasn't quite followed, right? And so now the uh the struggle or and the opportunity for everybody is finding new ways to make money, right? And the operators sit on really a gold mine of data. Uh, you know, do we are all um hyperattached to our devices? I have two kids, two teenagers, and uh it was to ask them if they would rather lose a hand or their cell phone, like they would give me their hand, right? Uh and and that's true for grown-ups too. So I think the the opportunity is there to move from yes, the uh being connectivity providers to something else. And then the what that something else will be, uh it's a little bit TBD. Uh we see things evolving very differently in different parts of the world. Uh, this has to do with, I guess, the level of uh risk tolerance or adversion, and also different regulatory environments that uh different countries have for what can and cannot be done with the data. But we are uh we are very bullish that um with maximum respect for privacy and regulation, there is a lot of value that operators can bring to their customers, um, and there is opportunity there.

SPEAKER_00 2:50

Yeah, it's a fantastic uh shift to watch in action. And speaking of which, every operator is is trialing AI in some fashion, front office, back office, network projects, field operations. Um so there's a lot of uh uh project science projects, I'd say, more than anything being done today. But what are some of the early real-world AI wins that you're seeing within operators today? Or are you? Is it is it too early?

SPEAKER_01 3:19

Uh so I I would say um it is quite interesting. You can probably still go and find some recordings of my previous life when I was televised and talking about how telecom did a really poor job at talking about how much AI has been in use for such a long time, right? So there's been a lot of cool stuff that operators have done in uh terms of automation, anomaly detection. I mean, these networks are super complex now, and uh, and yet you have less people running them, and you have really cool tools and systems out there that operators are using. There's been AI in use for capacity planning, you know, overlined, ROI, consumer and enterprise use cases, and and uh you know, maximizing the use of different assets in different locations and so on. So that's been going on for a very long time, and and we just haven't been particularly good at telling our story, I would say. Um, and then it feels like almost like none of this counts because the world of AI started after October 2023, right? Chat GPT and onwards is the before and after kind of thing. Uh, I think in the after, we do see. So we have a few products that have gone to market um where uh with we do believe adds a lot of value. So if you think about the areas where mobilium operates, so we have our roaming business, so value other services for roaming, we have uh security mostly on the signaling firewalls, voice security and SMS security, we have fraud and business assurance, and then we have testing and uh DPI and analytics, right? So these are our business units, and pretty much in all of them we see very practical use cases where AI adds value, right? And I'll I'll talk about a couple of them that we've brought to market. So on the testing side, it's been all about making a platform that was built by engineers for engineers. It's a very technical platform, uh, which I loved as an engineer, right? Like more options is better, uh, hyper flexible. But now you fast forward 10, 15 years, and the community of users is very different. You still have technical engineers, but you have a lot of uh business users, right? So uh roaming managers, uh data analysts that use this platform on a daily basis to assess uh the roaming agreements performing as they expect, is the quality of experience in all their partners equivalent and so on. And the platform is a little bit overwhelming, and it takes a while to learn and master. So we've overlaid AI to make it simple, right? So uh customers can discover the functionality easier, they can find the reports just by interfacing with the bot, like you would do with uh ChatGPT or Gemini. Uh so that's gone to market, and then this also helped us discover what else do customers need, right? What else would they want this agent to do for them? And so the next phase for us in that uh particular technology is automating root cause analysis. Because if you go along this same um kind of set of uh users that the platform has, many of the users don't want to spend hours downloading pickup files and going through them, right, all the body details to find out what went wrong. They want to be told, right? Like this test case failed, this is what it means, this is what you should be doing next, which suggests that you run these test cases or hey, with high level of confidence, this is a network problem at your side or at your partner's side, open a ticket, right, get it resolved. So that's one example. On the revenue assurance or fraud side of the house, fraud prevention side of the house, similar situation where you have massive amounts of data to identify these um, you know, uh cases that are suspicious of fraud. And so before it would take a data analyst a lot of time to go through all this data and then establish, yeah, with confidence this is uh likely to be fraud and of that type. So now we have the agent do this for them. So the cases come pre-analyzed with the all the evidence. The uh the do we still have human in the loop? And and I, you know, I know this is a big uh topic of conversation in the industry. Uh, how much, how far do you want to take the automation? So at this time we see mostly still this operating and human in the loop. Because there are serious implications uh depending on the action that you take after the case is fully analyzed. So the human can decide that, yeah, they they do agree with the analysis the agent has performed. This is conclusively a case of fraud, and an action should be taken. Um or not. This is a false positive. The logic that the agent is using should be refined because it's some sort of corner case. And so we're seeing that closed loop actually improving the efficiency and the accuracy of the agents uh at play, basically, in production.

SPEAKER_00 8:18

Fantastic. And telecom has sat on a mountain of data traditionally, historically. I mean, I remember analyzing called detailed records for different purposes in the 90s and 80s. So it, you know, this data has been around, and yet monetizing it has historically been very difficult, at least compared to how you know big tech companies monetize data like Google and Facebook and beyond. Why is that? Why has it been such a challenge?

(Cont.) AI’s Real Payoff In Telecom

SPEAKER_01 8:47

Such a good question. And um, you know, it's uh at times it feels like the um rules don't apply to everybody equally, right? And and part of this I honestly has to do with the uh how we all approach our relationship with our operators because we pay a bill, how we approach our relationship with our hyperscaler of choice, because many times the services they offer are free, and you assume, well, okay, they have to make their money somehow, and so you you do realize that in a way we we are the product, right? Like they say. Uh, so I think that's part of it. Uh uh at the same time, we've seen uh and we've had some successes in markets where we've been able to do uh rep share type agreements. We did a um webinar a couple of months ago with uh our customer telecomsell in Indonesia on how we've been able to help them do data monetization. That's a big deployment of uh our DNA platform. And within um all the regulatory constraints and privacy, uh, with anonymized data and aggregated data, we've been able to serve some use cases on the B2B2C side or enterprise side that provide a lot of value to uh these third-party companies that sit on top of the data and they would like to know more about what their customers are doing and what their competitors are doing. Not maybe individually what Evan and Miguel are doing, but what a community of users are doing in a particular region and what should they do to react to that. And so that type of use case uh that's not uh that's very sensitive to privacy, does not disclose individual identities and so on, uh, we think still applies to many other regions and is more of a you know kind of giving it a try, right? And see what can be done within uh regulation.

SPEAKER_00 10:44

Yes, the regulatory environment is changing more than ever. Uh the other area that AI is definitely impacting, we begin to see the impact, is in customer experience for telecom, you know, the much improved chatbots and in some cases customer service experiences, although there's a way to go. That's traditionally been the Achilles heel of many telcos. What do you see as the role of Jet AI uh LLMs into telecom operations? And can we get more than just a customer service? Although that would be a nice start.

SPEAKER_01 11:18

Yeah, so I think it's a tremendous opportunity. And uh that's an area um, like we were talking before, where it's massive amounts of data, uh real-time data as well. And yet in that data, you can with a very high level of confidence predict that something is about to happen in the network, right? And the about to happen might be minutes, might be hours, might be days. But because you've seen all these signatures before, you can identify that an event that will be customer impacting is about to happen and you can take an action up front. And I think this is um one of the areas where operators are improving, but there is a lot of opportunity still, right? Because the moment the worst case scenario is when you react to a customer ticket, right? You've at that point it's already too late, right? That customer has been impacted and impacted enough that they are gonna spend their time letting you know that your service isn't working, right? And uh so in uh in my previous live, we used to say that for every call that you get, there are 10,000 that don't bother to call you, right? But they are equally impacted. So you you need to assume worst case scenario, that uh there are many paying customers, that their business, their everyday life is being impacted by that, whatever the issue is. So being proactive, uh doing the anomaly detection upfront, taking the action uh when you know something is about to happen to prevent the customer from being impacted, and ideally uh being transparent with the customer as well. And we've said I've seen uh previous life some customers that have decided to be proactive, right? They'll let you know, hey, maintenance is happening in your neighborhood. Uh, if you have uh a particular importance, say like a call like the one we are having, they have a plan, you might want to look for an alternative uh way to connect or another location. I mean, wouldn't it be nice, right, if that was uh done by all our operators? Because they uh they're uh not having the service is very impactful. We depend, I depend, and you depend on uh connectivity vendor to conduct our business. And so I think that's a really good opportunity for um being more proactive and uh you know detecting and resolving issues before they impact customers.

SPEAKER_00 13:39

Yeah, brilliant. Um churn, of course, is the enemy of every telco, uh at least in a competitive market. It's the first number you look at when uh you know financials come out every quarter. Um how are you seeing AI preventing or perhaps even reducing churn moving forward?

SPEAKER_01 14:02

So this is um uh this is probably one of the areas that's the most impacted by uh the privacy uh concerns that uh apply in some markets. Uh and at the same time, it's one of the biggest opportunities, right? Because in the the um there are a lot of signals that operators can get about churn propensity, right? Like a customer that's had about experience, a customer that is uh thinking about switching operators, very sensitive again, and and all this has to be within uh what the regulations in each market allows. But um that's an area where AI can help uh tremendously, right? Like put together all these signatures and allow the operator to make it right for the customer when there is a risk that they lose it. Because we all know the uh the cost acquisition is tremendous, right? So whatever you can do to avoid losing the customer in the first in the first place, uh the ROI is much, much higher than if you lose the customer and then you have to fight to acquire it back, right? So uh we see a lot of opportunity there. Uh we've had some success in some markets also, but this is truly one area where it's very delicate, right? Like what amount of information can be used so that also you don't spook the customer into uh, you know, the big brother is watching type situation, right? Uh it has to be very carefully done.

SPEAKER_00 15:28

Interesting. You touched on fraud a moment ago. Let's talk about fraud. It's uh still a huge problem in in our industry. Uh fraud security threats, of course, uh outright theft, cybercrime, whatever you want to call it. A problem in every country, some more than many much more than others. Uh you know, even in the U.S. here, I I barely answer my phone given all the fraud calls. I I let Apple screen it. Not a not a really elegant solution. So, how are you seeing carriers that you talk to using AI to fight fraud and security challenges?

SPEAKER_01 16:04

So, and you're right. The reality is the bad actors are really good at what they do, right? And uh it ends up being a game of whack-a-mole. So you know, you identify the the flavor of the week and you address that, and then they find you know ways around it. So AI is particularly important here because any static rule, they are phenomenal at finding the threshold and working within that threshold, right? And adjusting. Uh, we are we we are excited about some of the things that we're doing on the voice side with agents that can get uh instead of the like in the example that you were uh you were using, where I'm myself, I don't answer any call from any number that I don't know anymore. Uh but uh we can have agents that do answer in engage and derive a lot of intelligence as to what's the intent of that call, and and then that information can be applied in the fraud management system, but also can be used for to package it and extend that information to other industries that really need this, like banking and fintech. So that's where we see a lot of opportunity. The places where um we see things progressing faster and better are markets where the regulator is very pro-customer, very pro-protection. Uh and so we we see uh we partner with the regulator, with the operators to build new technology, and uh the regulator brings other actors in, like the um uh police force and so on, cybercrime groups that they have. So there is more of uh sharing of information and then sharing of potential methods for mitigation, detection, and uh you know, prevention. And so there are parts of the world where uh the regulators are very, very proactive, very hands-on uh at testing new technology, and that's where we see kind of a lot of the innovation happening that then ends up applying to other markets, right? And I'm thinking Asia Pack in this case in particular, we see some in the Middle East as well. Um, and then I would say in some markets the regulation just makes it a lot harder, right? To uh to you you can innovate, you can present ideas, but it takes longer for the operators to be able to use these tools and deploy these tools in their network.

SPEAKER_00 18:25

Got it. So, as you know better than anyone, telecom is usually traditionally moved at pretty slow pace, very cautious, and for for good reason, but um, you know, some carriers are are getting better than others, but there still seems to be this uh, you know, pilot purgatory, you know, getting stuck in pilots in the lab for months, sometimes a year or two before services get rolled out. How can we escape that trap and move faster as an industry?

SPEAKER_01 19:00

Um so to be honest with you, I think a lot of this is just inertia, right? And uh, you know, having, you know, like we we come from a kind of a different time where the expectations were maybe a little bit different, but now uh you know, I use this every week. We're like everybody's a friend of me in the industry. You partner and compete with hyperscalers, you partner and compete with startups. Often you partner and compete with your own customers, right? In in different parts of the industry. And I think uh we we just have to um kind of let go of some of the risk aversion that we've had, because otherwise we won't be able to remain competitive, right? Because the the reality is some of these, definitely the startups, but uh the hyperscalers as well, they've created their their entire environment based on this. Try a lot of different things, fail fast, and move on, learn and move on. And I think the um we can do the same in telecom. We probably need to uh make sure that we explain to customers what this means to them, uh like when you and just find their the right set of customers that might be, you know, like technology forward that would like to be part of this type of pilot and and such. Uh uh, you know, there is no reason why we cannot do it, right? Uh uh that's my my belief. And uh with the tech that we have now with orchestration, with automation, the move from pilot to network-wide deployment can occur a lot faster. And and operators are doing better at this. Uh, but I I still I agree with you that the um there is a lot of hesitation to move from the pilot to the scale production. And oftentimes the pilot, the goal of the pilot seems to be the pilot itself, right? Rather than deciding that, okay, what are we learning out of this? What are the success measures that will determine whether we move forward with this technology, even if it's for use cases different than when you started the pilot, because it happens all the time, right? You learn something along the way that you didn't envision up front. And I think that as an industry, we need to be more aggressive.

SPEAKER_00 21:15

Yeah, I agree. And uh it used to be oh, you know, telecom needs to move at hyperscaler speed. Now it's telecom needs to move at AI speed. I I'm a Claude uh fanatic. I'm using Claude Code and Claude co-work in the background. I'm actually building apps, and there's an update to Claude almost every other day. So I think they're they're having this notion of okay, we're just gonna move fast and break things from time to time, but it's not gonna stop our innovations. So yeah, definitely a new generation of uh telecom leadership will probably uh speed things up as well. Uh, tell us about your roadmap and plans for the future as far as you can share. What's on your radar horizon? What kind of things are you working on uh over the next weeks, months? Months?

SPEAKER_01 22:01

Yeah, so um I think we touch on some. So definitely um, you know, bringing all these AI capabilities throughout the entire portfolio, all five business units. Uh, that's uh a top of mind for us. Um we are also making strides into the enterprise space. So Mobilium historically did all nearly all the business with uh uh with the uh MOs, but we're seeing there is an opportunity to participate on the enterprise opportunity with the MOs and at times also direct to enterprise. Um we are also doing some interesting partnerships with companies that do things. Um that we've been doing a lot of innovation on and peer activity, uh, one of our good partners called No Holt, which is a company that's been in AI for 20 some years, before AI was a sexy thing. And so we are uh exploring solutions for the MOS to sell to small businesses to enable them to also participate in the AI economy. So we're very excited about that. Um and then uh overall, in uh some of the business units we're building towards the supporting this TM Forum Autonomous Networks kind of aspirations, right? Like this whole uh layer cake from zero to five. Um, five, to be completely honest with you, still feels a little bit aspirational. Um, you know, I I hope I'll get to see it, but it'll it'll take a while to get there. But we see all these uh use cases that already can be in the three to four range, like uh you know, highly autonomous, right? That's how TM Forum puts it. So we are putting some calories behind that. And then last but not least for us at Mobilium is uh, you know, we we are in a way almost a conglomerate, right? Mobilium went and acquired uh over the last five years five or six different companies. Oh, wow. We are adjacencies, so this is what made these five business units that we have today. And uh we are discovering and we're with our customers that a lot of the value is created when we put these discrete products together as a solution, almost like playing with Legos. And so that's where we are spending uh quite a bit of time. Security and risk goes hand in hand in many of the deployments, testing and DPI and analytics goes hand in hand because it can provide a 360-degree visibility. So that's an area of opportunity and uh focus for us, and we're bringing some solutions to market between now and the end of the year.

SPEAKER_00 24:33

Brilliant. Well, I can't wait to talk about those in the near future, and congratulations on all the success. Thank you, Evan, and thanks for uh thanks for having me. It was a lot of fun. Yes, it's been fun and even more fun. I see Grogu hiding in your background there. Uh if you haven't seen The Mandalorian, the new Mandalorian movie, if you're a fan, it's the absolute blast. Uh it's two and a half hours, so it's definitely a journey. But uh good times. Uh, thanks for chatting. All right, thanks, Evan.

SPEAKER_01 25:00

Have a good day.

SPEAKER_00 25:00

Take care, everyone. Bye bye.