rApps for Mobile Networks Autonomy

Autonomy in telecom sounds like a pure technology race until you look at where operators actually get stuck. It turns out the models aren't the bottleneck. The humans around them are.

We're joined by @Ibrahim Eldeftar, who leads Cognitive Software and Services at Ericsson, to unpack the real path from partial automation to Level 4 autonomous networks, and why the hardest part is often the human system around the tools.

Ibrahim walks us through the two hurdle categories every CSP runs into. The first is the technology foundation: multi-vendor support, scalable AI platforms, data management, deployment at scale. The second is the organizational side: change management, upskilling, new ways of working, and breaking down silos that have been cemented in place for decades. The industry keeps underestimating that second category, even when the AI roadmap looks finished on paper. Ibrahim explains why, and what it actually takes to move an operator forward.

From there we get concrete. rApps and a service management and orchestration platform can replace the fragmented automation stack most operators are living with today, giving teams a common SDK, consistent interfaces, and an ecosystem model where operators build apps themselves or source them from partners. Ibrahim shares real proof points from live networks, including modernizing worst cell hunting with AI anomaly detection and root cause analysis, and taming massive MIMO complexity where the search space is simply too large for humans to tune in any reasonable timeframe.

Then we get into what changes when GenAI and agentic coordination enter the picture on public cloud with AWS. Natural language "talk to the network" interfaces. Orchestrating dozens of RApps at once. A shift toward RApps as a service and SaaS delivery, where operators pay for outcomes rather than software licenses.

Subscribe for more deep dives on telco AI and network automation, share this one with a colleague who's living the automation grind, and leave a review if it landed. And think about this while you listen. What would you automate first if you could truly trust the outcome?

Transcript

Meet Ericsson’s Telco AI Team

SPEAKER_01 0:02

Hey everybody. Fascinating discussion today with Ericsson at the leading edge of AI and automation. Ibrahim, how are you? I'm doing great. How are you doing, Ivan? I'm doing well. Great to see you again on the heels of Mobile World Congress. So much going on. Before that, perhaps introduce yourself and your team and the mission within Ericsson.

SPEAKER_00 0:24

Thank you very much, Ivan, for uh having this opportunity again. So my name is Ibrahim Diphtar. I'm heading the Telco AI product management in Ericsson. So myself and my team, we are responsible for creating very interesting type of applications for mobile network automation that we call them our apps. Basically, they stand for radio apps. We can talk a little bit later why we call them that. So we are the team who is responsible for the strategic product management, uh putting the strategy, go to market, and also working closely with our customers on converting their vision for network automation into software products that they can use and they can deploy at scale.

Why Level 4 Autonomy Stalls

SPEAKER_01 1:08

Super exciting. And uh you're helping so many operators uh around the world make incredible progress. Uh and yet from Ericsson's perspective, what are some of the biggest technical and organizational hurdles, challenges still holding CSPs back from moving beyond where we are today with partial autonomy to a true level four, let's say, autonomous network?

SPEAKER_00 1:32

Well, I mean, that the I would say the challenges uh are uh they belong to two different families. One of them is related to the technology, uh, and the other family is actually related to embracing change and uh the readiness from an organizational perspective to do that transformation, required transformation that technology brings. So the first part when it comes to the technology challenge is that having the right answer from a technological perspective. So, do we have a multi-vendor capability that uh CSPs can use at scale for different vendors? Uh, do we have uh platforms uh that can support AI scaling and can actually provide such kind of capability for data management as well as for uh ease of deployment and development of different applications? So these are the technological questions that we work with our customers to resolve and basically put them on the right track when it comes to the technology. But but you know what, Ivan, the interesting part about it is that based on our own surveys and experience, and the same thing also we heard from third-party uh uh uh like assessors, is uh the technology requirement for transformation accounts for something between 30 to 40 percent of the journey, of the transformation journey itself. The rest is about change management and organizational readiness. And this organization readiness is materialized in basically understanding that introducing AI and introducing automation will call for changing ways of working. Uh, there will be a requirement for competence upskilling and rescilling of the workforce for a new future of network autonomy, as well as uh breaking the silos of fragmented automation systems and fragmented organizational responsibility into coherent ways of working across the organization. So these are like some of the challenges that we work very closely with our customers trying to address that so they can move to network autonomy quite fast.

RApps And SMO Beat Siloed Automation

SPEAKER_01 3:38

Incredible. Um, so given the many R app uh uh applications being deployed together with the service management and orchestration platform, how can they actually improve the speed or practicality of reaching this advanced level of automation more compared to more traditional or siloed automation approaches?

SPEAKER_00 4:01

In the siloed automation approach, you have different challenges. Some of the challenges will come to how you manage such kind of fragmented landscape when it comes to security, when it comes to enhanced ways of deploying the software, doing the lifecycle management of the software, and also the readiness of the people that they can develop applications and scale them on a network-wide uh uh uh you know scenario. Uh so the service management and orchestration answers this need at the very right time for the industry transformation. So, with a service management and orchestration platform, we defragment this landscape of automation tools. We bring a common user interaction to how to develop applications, what we call the software development kit, the SDK. So, in order to develop different use cases in different parts of the network, you just use the same platform, you use the same SDK in order to do that. So this is unprecedented in the in the industry. And not only that, you have the opportunity to build it yourself, you will be able to procure it from very large number of CSVs and uh sorry, sorry, very large number of ISVs as well as CSVs who are creating the R apps themselves as well. So the true openness in the interfaces, the capability to build R apps, and also either by the CSV themselves or by like you know, any other players in the industry, Ericsson, of course, is one of them, uh, is unprecedented at this kind of scale at this time in the in in our industry.

Real RApp Results In Live Networks

SPEAKER_01 5:43

Fantastic. And add some commentary on which R apps-driven use cases are having the biggest impact for CSPs today, aiming to reach this higher level of autonomy. Maybe you could share a few real examples or anecdotes of the value um they're delivering today.

SPEAKER_00 6:01

Well, uh the the the R apps, when maybe one thing, Ivan, when when we talk about the design of the R apps, I mean R apps are actually built for a purpose, is to create value for CSPs. And the value could be materialized on different uh uh fronts. Uh could be for optimizing efficiency and optimizing OPEX that the customers are spending for network operation. It could be also for optimizing their assets, what they have invested in Spectrum. Uh, it is also to improve network performance and user experience. So users who are basically today, mostly humans, I would put it, who are enjoying networks, they actually find it uh uh uh fulfilling to their needs. And the need as well, like when it comes to the value for running greener, so having a sustainable operation. So these are different value pillars that we think about, we have it in mind when we design our apps. So when we design, we when we have this kind of design mindset, we aim to introduce the AI technology and the R apps that can fulfill those, maybe at the same time, you can hit it with the same R app or different R apps doing different things. So, give you an example. Uh, some of our R apps address a very uh old problem in network optimization that you know what, Ivan, has not actually changed since the times of 2G introduction is how you address sales that are performing badly in the network. So, and what in the process that we typically refer to in the industry uh as worst cell hunting, we created our apps that revamp this whole process. So rather than looking at limited number of sites, which are basically guided by the human capacity, we can take full network 100%, analyze it, unidentify the issues. Not only that, we will be able to do root cause analysis for the issues that we detect. But not only that, Ivan, we actually have the machine recommending what kind of actions that need to be done in order to improve the network performance. So these apps, uh, for example, the first one that I talked about, we call it a cell anomaly detector, the second is anomaly root cause explainer, the third is anomaly general optimizer, are all AI-powered apps that actually have proven in the field that they are able to improve network uh uh uh efficiency by more than 50%. So basically you spend less time analyzing and less time coming to action in the network to improve network performance. They have reduced the number of sales with issues by more than 60%. Doing that, they have also we have seen that the customers we deployed those were able to get better spectral efficiency in case of congestion. Sometimes it goes to uh like you know between six to nine percent. Uh, we have seen improvement in user throughput in the uplink and downlink. So these are like examples of what you do for a very old traditional uh uh problem that we try to resolve a little bit differently now. But think also about uh one of the new challenges that we have for massive MIMU networks. Uh Ivan, you probably know that in order to manage a 5G advanced network with massive MIMU uh coverage, you will have different options of how you tune the cell shapes, what we call them the cell shapes. In Ericsson, today for each cell, we have actually like we we with the in the most recent that we have, we have 13 cell shapes. So imagine, even that you would like to optimize a big city, and this city has hundreds of thousands of cells, hundreds of thousands of radios. You take this, multiply it by the number of frequency carriers, multiply it by the number of cell shapes, and you end up with an astronomically large number of probabilities that humans can control. This is a new process, this is a new challenge that was not actually before in the older generations. But we were able to use it with our own RAPs, with AI algorithms that can actually pick and choose for each and every cell, tailor the coverage footprint for it. We have put this in live networks, we have seen improvement in strictural efficiency by up to 4%. We have seen improvement in downlink and uplink throughput. We have seen improvement in voice, uh uh network quality as well, all with like with different variants of I would say digits because the networks are different. But the consistent result is there is an improvement in OPEX efficiency. There is improvement, some of the networks actually reached out even 75%. I was personally impressed with that. Yeah, I mean, because that's that's a significant improvement compared to before. And this is a real number that uh it's an interesting thing that I can bring to you in our own estimation. We these we said that you know what is gonna save between maybe like 70%, the maximum. But when we put it with the customer, the customer came back to us and validated, and he said, you know what, guys, it saved actually 75%. So so that's that's something that you know you that that is real, that we have heard from our customers and we listen to them. Uh so again, the the the there are different field proven proof points that we have brought with with this AI technology that we bring for network automation. And either that it is a new process, it's a new challenge that comes with the new G, uh be it 5G today or maybe 6G tomorrow, uh, or a very old process that started with the times of GSM. We started to have a mindset change with CSPs on how to approach that and how we solve it using artificial intelligence in radio applications or RAMs. Incredible.

GenAI On Public Cloud For Autonomy

SPEAKER_01 11:58

Wow, such uh such a huge opportunity. Of course, the whole industry is excited about agentic AI. How does Erickson view the role of RAT combined with agentic and intent-based AI in really reshaping operational decision makers making as CSPs move to this level four autonomy you're describing?

SPEAKER_00 12:19

Well, so what I described right now is not powered by AgenTech AI. So we thought that because AgenTech AI is such a powerful technology. How can we bring the best of what we and Ericsson has perfected over the past decade when it comes to AI, uh domain expertise, and software engineering, we bring this to the public cloud. When we bring this to the public cloud, we unlock new capabilities from a technological perspective. So with public cloud, uh our partners that we, our partner that we have chosen for this journey is Amazon Web Services in AWS. We bring the best of the two worlds together. So we bring Ericsson domain expertise in mobile networks, scalability with a number of CSPs, as well as our own knowledge and capability in and understanding networks, together with elasticity, speed, and AI stack from AWS. Doing that, we do a level shift of network autonomy, if I put it this way. So the value, the the the proof points that I mentioned, they actually become elevated more with capabilities of Genii. So rather than working with these apps and consuming the typical charts and GOIs, you can actually talk to the network. We actually call it talk to the network. Is that in the natural language, you say, like, okay, what are the most uh pressing issues in the network today? Or how could I improve you know this coverage area because there is a stadium, like you know, a stadium event, maybe a concert, maybe even like you know, a match or like you know, uh a sports match in this area. How can you know I improve that? You can communicate, talk to the network in a different uh way that is a different interaction between humans and machines. That's what we bring with public cloud capabilities, but it's not only that, because we leverage the genetic AI stack that AWS makes available as different services for us uh in order to coordinate the different R apps. So now, Ivan, think about the world when we have not just like four or five RAPs that we have designed already to be coordinated, but we are having, I don't know, 40 different RAPs, not only four. And this is the future where we are moving. You know, we sit with we sit down with CSPs and they tell us I'm expecting that I will have 50 R apps within 33 years. So, how would you manage and coordinate that? Autonomy needs to be coordinated, autonomy needs to be harmonized, and this is what a genetic capabilities bring on top. And this is what we have actually brought. You know, when when we met Ivan, you know, uh at MWC, we were just doing the launch of that, and it created tremendous interest from our customers because we showed them a world where you bring coordinated autonomy on top of public cloud. So you leverage the best of the two worlds, the uh uh Ericsson telco expertise together with the public cloud capabilities. So you unlock new possibilities and you have a faster transformation to L4 for autonomous networks.

RApps As A Service And SaaS Shift

SPEAKER_01 15:49

Amazing to see the capabilities from public cloud providers going beyond basic compute and helping you scale globally. What a great opportunity for carriers worldwide. And uh speaking of that, Ericsson recently introduced RApps as a service. Uh, what are some of the key benefits CSPs could be looking at from this offering? How do you describe it?

SPEAKER_00 16:14

The the CSPs will be able to consume uh the network automation in a different way compared to traditional deployments. So this consumption uh will be first made available elastically on demand. So you would like to use an R app, you would like to have a use case that you, you know what, I would like to deploy it maybe for a small part of the network, uh, or I would like to deploy it on at a smaller scale. So can I do that? Yes, you can. So this kind of capability of elasticity where you can scale up and scale down is something that uh CSPs see the benefit, see the benefit of it do being on public cloud. Another thing that CSPs also will be able to leverage is the capability, the ability for them to consume that on the marketplace. So a single price object that includes software, like you know, services on public cloud, uh services, professional services needed to operate that, where Ericsson is actually taking care of the full operation of the R. So for them, it's a single price on the marketplace where you actually pay for a trusted outcome rather than just consuming a software. So these are the capabilities that that sorry, these are the values that they can actually get out of this new offering. A new GTM, new types of technologies that we explained with the Gentec AI and Genai, as well as the elasticity on public cloud.

SPEAKER_01 17:54

Wow, what a blockbuster. And indeed, the latest analysis Mason report points to SaaS as being the dominant delivery model for our apps. Um, do you see that influencing how CSPs adopt new capabilities in general on their journey towards autonomous networking?

SPEAKER_00 18:12

Well, I mean, the the analysis that the the the research that Analysis Mason uh recently published actually shows us very interesting pattern of development of how CSPs consume software. Closer to three-quarters of them say, like, we would like to consume it as software as a service. So that's a very interesting insight because it confirms that the preferred mode of operation for the future for network autonomy will be based on outcome that we promise to our customers. The outcome that they can see in the network is not just the technological capabilities, like you know, the bells and whistles of AI. That's absolutely important to make sure that you are deploying and using the latest and greatest in technology, but they are may very much focused on outcome. Uh, and this is what we see on the movement towards SaaS. Uh, more and more in this happening in the telco. We brought this in Ericsson be like you know a little bit earlier with what we call Ericsson in demand, which is like making core network available as a service. Now we are also making automation available as a service. So more and more towards this movement where you get the best of the two worlds again together, the telco cloud, the telco uh uh world of standardization and scale with a public cloud world of elasticity, speed, and AI capabilities.

Final Thoughts And Farewell

SPEAKER_01 19:38

Incredible. Well, such an exciting time to be leading the industry charge here. Congratulations on all the success, onwards and upwards. Thank you very much, Ivan. Thank you. Thank you. And thanks everyone for listening, watching, and sharing. And until next time, take care. Bye bye. Bye bye.