
AI Ops & Autonomous Networks: The Future of Telecom
In this Data Insights episode, Andrew De La Torre discusses how Oracle is leveraging AIOps to enable automation and optimize operations, transforming the future of telecom.
In this Data Insights episode, Andrew De La Torre discusses how Oracle is leveraging AIOps to enable automation and optimize operations, transforming the future of telecom.
Transcript
Allyson Klein: Welcome to Tech Arena Data Insights. I'm Allyson Klein, and that means we are back again with Jeniece Wnorowski from Solidigm, and welcome to the program. Jeniece, how are you doing?
Jeniece Wnorowski: Hi, Allyson. Thank you. It's great to be back as always. And I'm doing wonderful. Just recovering from GTC.
Allyson Klein: Yeah, me too. And I think that that blends in really well with our topic today and a really interesting use case of AI.
Obviously we've been talking about AI in terms of a lot of different applications, but we're going into a really interesting domain today. Tell me about who we're talking to.
Jeniece Wnorowski: Yeah. Today we are gonna have some interesting commentary from Andrew De La Torre, who is actually the vice President of technology for Oracle.
Andrew De La Torre: Hey there. It's great to be with you, Jeniece and Allyson. And great to be here with your listeners too.
Allyson Klein: So Andrew, let's just start with an introduction of your role at Oracle and how your engagement in network fits within Oracle's broader purview in every aspect of the tech's landscape.
Andrew De La Torre: Yeah, absolutely. So I am actually in a part of Oracle that is responsible for creating products and solutions for the communications industry. So we work with large service providers around the world. We help them build their 5G networks. We help them build their front office IT systems, like their billing systems, their care systems that they look after their customers with.
And obviously in the broader context, what we're observing in the market, and particularly in the communications industry, is that there's an absolute convergence around the adoption of technology, which is starting to bridge the gaps of typical areas where Oracle are known, i.e. things like database and the back office applications like our ERP and our Human Capital Management solutions.
But integrating those more closely with the front office systems and even the network themselves in the comms carriers. And so, Oracle is really starting to assume a very strong and unique position around the way that it can bring all of these disparate systems together, and AI is obviously front and center in terms of how you can benefit from doing that.
Jeniece Wnorowski: Amazing. And Andrew, you guys are driving some incredible innovation into operation control of the network. Can you describe for our audience how Oracle engages in this arena?
Andrew De La Torre: Yeah so, we're absolutely seeing a shift in the way that our customers are approaching their digital transformations. And whereas, I would say previously you saw a reasonably siloed approach to this where they might have been dealing with it in the network domain, and then there may be a different program around some of the back office or some of the front office systems. It's a much more holistic approach that they're taking at this moment in time.
A lot of that is driven by an understanding now of the need to converge all of the data sets that exist across the various myriad of technical solutions they have in their architecture. And more importantly to the conversation today, the way that you can then use AI and the range of AI technologies to be able to benefit from that.
So, we're really in a place now where we're bringing together a much broader portfolio of offerings as Oracle for our customers to help them drive those transformations, which are really end to end across their business. Now from our perspective, we also want to therefore focus very much on how we can embed AI capabilities into every layer of the technology stack.
Whether that's the vertical domains of which I just mentioned, you know a lot of them, or rather horizontal domains or even the vertical capabilities. So, how do you get it down into the cloud infrastructure layer? How do you do it at the application level? How do you have other capabilities that are sitting around at the autonomy layer?
And for most of our customers there's really two things that are driving these transformations. One is around continued desire to drive operational improvements in their business. So, bringing more automation, which helps them with troubleshooting, it improves the service uptime in their networks. It's obviously lowering their operational costs because they're moving to more autonomous solutions, away from human manpower.
But the second, and probably equally important, if not more important thing is that it's helping to provide a pillar for them on their business transformation around new revenues. Because, as a lot of the industry that we work with, i.e. the communications industry, is moving beyond consumer and into this kind of industry 4.0 space, the need for them to be able to support new digital connected applications is really critical. And we're seeing them needing to equip themselves with much more autonomy, much more AI powered capabilities to be able to move into that digital space.
Allyson Klein: I was so excited to talk to you about this because one of my hot topics for 2025 is really AI ops and how AI ops is advancing to enable more efficient and scalable control of infrastructure across a number of different spaces. You've been talking about this for a few years, but can you just ground our listeners on what AIOps is and what is the current state of AIOps?
Andrew De La Torre: Yeah, absolutely. I mean I would describe AIOps as largely a framework, and it's a framework to fundamentally unlock the premise of an autonomous network, and network in the broader sense here, where obviously in comms one connotation is the network we all think of i.e. the one we access. But that can just as equally apply to an IT network that is really the fabric for all of their business applications that they're running.
Now, what is an autonomous network? It's a network that can fundamentally self-monitor, self-optimize, self-heal with minimal, if any, manual intervention. That's really the nirvana here, which is how do you effectively have a closed loop capability around these networks to really let them manage and run themselves.
Now, when we talk in the communication space, one of the things we typically do is we go to the ETSI standards, this is the European Telecommunication Standards Institute, and they've actually been doing work for a long time now around network automation and they have effectively four levels of automation that are defined within the standards that they create, moving your way from basic automation at level one up to really full automation at level four, where there's basically zero touch from a human intervention perspective.
This is really the journey of the industry and where it's trying to go. Now, AI ops, as I said, is a kind of framework to get you to that place. And if I double click, just as one example of the application of this, into a 5G core network, for us as Oracle, that framework materializes I would say in sort of four different pillars.
The first one is that you have to make sure that the actual applications that you are running here, and in this instance, it's the cloud native functions that form the 5G core itself, they have to be cloud native and I mean truly cloud native. And there are many different versions that we see in the industry of what claims to be cloud native. But if you haven't designed your application from the ground up as a full microservices architecture it's really difficult to access the autonomous capabilities of cloud, which is a really, important starting point for this.
The second one is you need to get your hands on data. Data is key for any kind of autonomous system. And so, in Oracle we have a product called Data Director. It's specifically designed to be able to tap into all of the data that resides around the network, whether that's from an Oracle product or not. And it's bringing that together, it allows us to do aggregation, filtering, pre-processing. There's a number of different things it can do, getting that data together is really key.
Then we do have analytics capabilities. So you know, one example is our network data analytics function, and that's one of the standardized components of a 5G core.
This is the place where the magic happens. This is where we have our AI models and they're taking the data that they're receiving from Data Director, they're doing the analysis on it and they're creating the action level insights that come from it.
But then I think I'd finalize this by saying, this all gets wrapped in an automation strategy, which means you need tools and practices around that. So we're applying things like the GitOps practices and the sort of microservices-based approach to things to really be able to unlock the ability to start to manage this whole environment in a closed loop way.
Jeniece Wnorowski: Amazing. Thank you Andrew. That's just awesome information. I do wanna back up for a second though, cause everyone's talking about AI and we wanna understand what kind of AI are we talking about in this application and is there a difference between traditional machine learning and generative AI, and can you provide a perspective?
Andrew De La Torre: Oh, that's a great question because I think right now, obviously there's a lot of industry focus around generative AI as a particular flavor, not without good reason. I mean, it's proving to be such a capable and competent technology. It's becoming so pervasive. But in this particular application, we have to really, I think, realize that what we're trying to tap into is a whole toolkit of different automation capabilities, of which generative AI is just one.
What's really important in the telecom space, first and foremost, is that you focus on creating very specialized trained models because telecoms is not like most other applications and you have to make sure that you've taken, from a modeling perspective, the right approach to get all the data sets, all the algorithms, all the considerations that are unique to a telecoms network into account.
And to that end, what we find is as you move across the various use cases that you might come to want to implement within a telecom environment, there's actually a place for many different types of automation. So at the most basic level, using robotic process automation, the simplest form of algorithmic based automation, still has a role. There are a large number of very routine and well-defined tasks where what we really need is a completely predictable outcome as we move them into an automation domain. And so, we see still the use of those kind of technology solutions to be able to fulfill that kind of use case.
But then of course, you do start to spill into AI and whether that is, at the most fundamental level, sort of machine learning level, either LSTMs or some other kind of recurrent neural network, we are seeing that the benefits of being able to use the historical data to be able to drive better insights and better outcomes from the analytics that have performed are very valuable in some cases.
And of course you can extend that further into natural language processing, which of course is, a large component of the way generative AI is built up, where that benefit of actually also now getting creative outcomes and suggestive outcomes is super important. It's really about recognizing that there isn't a one size fits all in this particular environment. And there's a place for pretty much all of the types of technology we've been talking about for decades.
Allyson Klein: Now, when you look at the network, you're talking about just a vast array of infrastructure from the core of the network to the RAN and everywhere in between. What area of the network is the initial target for this technology and how does data collection work across this broad domain?
Andrew De La Torre: Yeah. I don't think that I would say there's any one area of the network that is a target for this. And I say this because as I mentioned, increasingly our customers are undertaking transformation at a business level. So they're actually looking at everything that exists within their operation. And so, whether they're looking at something like our set of Fusion back-office applications where we have already infused those, even with generative AI capabilities, there's definitely traction being taken on there.
In the front office solutions like their CRM systems that they service their customers with, we have product there as well as Oracle and we have AI integrations, which are really helping them to assist the customer care agents. They're providing sentiment analysis on the customer interactions, they're providing the ability for you to recommend next best products and next best steps for the customers.
And then in the network space, we're seeing definitely a focus around areas like the planning of the network around service optimization and more and more around troubleshooting. So, not one area I pick out, but certainly use cases in a multitude of areas are starting to stand out as the lead use cases.
Jeniece Wnorowski: And then, what challenges are you trying to solve from an operational control standpoint, and then how much of that data is a challenge?
Andrew De La Torre: Yeah. I mean, look, let's maybe talk to this by sharing one sort of statistic around the state of the industry environment at the moment. There was a recent report that was published by Omdia, and they'd done an analysis across the service providers around the world. And their report basically stated that half of those service providers were basically at around ETSI level two, which is fundamentally partial automation, which requires human oversight. Only 29% of the people they spoke to claim to be at true automation levels with minimal human intervention.
Now I will actually also caveat this by saying that the way this report was done was it was like a self-exam. So I would probably hypothesize that these are wildly optimistic and that the levels of automation we see in the network at the moment are actually even much lower than these across the industry. And so, we come from a starting point where the opportunity to drive automation, the opportunity to use AI to be able to control that automation is still very significant because the progress of the industry is quite slow.
Now, when we think about some of the barriers to adoption that we see when we speak to our customers, I think some of them are around their business and around their ability to be able to transform both their business processes and their cultures to be able to adopt some of these different ways of doing things. That is definitely one that we observe. Another one is the continued struggle, particularly in this industry, around on-premise versus cloud-based solutions where it is still quite heavily oriented, particularly on the network domain to on-premise deployment. And, that gives you limited access to the true power of public cloud and the AI large language models that exist in that space.
But there are also, issues around legacy infrastructure. There's a lot of old equipment in these networks which actually are not designed for AI integration. There are genuine concerns around security and compliance. This is a heavily regulated industry and license conditions make it very difficult for them to understand how to correctly and safely adopt AI decision making.
And then there's data sources and the fragmentation of their data sources, I would say is probably one of the biggest ones because it really holds them back from getting a kind of really unified operational visibility on which they can then of course make actionable insights.
Allyson Klein: Now, Andrew, you brought up a lot of points there that took me into a slightly different space, which is the human element. You've got teams that are managing generations of infrastructure that both the infrastructure and the people are not necessarily akin to a DevOps cloud native type of mindset for automation. How do you work with the operators that are your clients to address that human element and take advantage of these capabilities where they can?
Andrew De La Torre: Yeah, it's certainly a journey. And you know, the one thing I would say is it isn't without some good reason as to why perhaps the telecoms industry moves a little bit cautiously. I mean, we are fundamentally talking about an industry that is now responsible for delivering critical national infrastructure. I mean, every country in the world now regards it in that way. And of course, they're understandably very cautious and protective of the way that they build and run those networks because they're so fundamental to the economies of the countries in which they reside.
But we absolutely see the difficulties that they go through because their business processes and their cultures inside of their organizations, they've been designed for decades on a waterfall approach and they're almost the antithesis of what Cloud native and DevOps really requires from organizations.
Now, you know, when we work with our customers on things like 5G where you know, everything is cloud native, one of the things that we do try to do as Oracle is we try to help them not just with delivering products, but actually delivering tools and capabilities that can help them move into the DevOps world. And things like our software repositories, we have automated test tools. All of these things we have developed for ourselves as Oracle so that we can deliver in a truly cloud native way, but we actually make those available to our customers if they choose to take them. And that's an important starting point for them in terms of being able to have the capabilities in their business to be able to change the way they do things and what they do.
That's just a little bit of an example of where we try to extend our responsibility as Oracle a little bit in the way that we partner with customers and, not just make it, simply a product transaction, but more a what else can we give you to help you with your business transformation.
Jeniece Wnorowski: That is amazing. I feel like that is unfortunately a wrap. You just gave us so much good stuff, Andrew. Where can folks go to learn more about all we've discussed today?
Andrew De La Torre: Easiest places? Go to oracle.com, click on industries, and click on communications, and you'll find hopefully everything you need right there.
Allyson Klein: Andrew, thank you so much for being on the program. I love this topic. I love what you guys are doing in this space. It's such a pleasure to have you on tech arena. And thank you so much for your time and Jeniece, this was a fantastic podcast. I can't wait till you're back for more.
Jeniece Wnorowski: Thank you. This was a lot of fun and again, well done Andrew.
Andrew De La Torre: Thank you. Great to be with you both.