The narrative around telco operations is this: as more technology layers got added, things got very complicated, with lots of vertically integrated, vendor-specific, domain-specific systems. Telcos could have dozens or hundreds of them managing different parts of their systems. Complicated, duplicative, over-reliant on human expertise and institutional memory. This was not good.
Then the industry felt that a better solution would be to engage in next generation transformations that integrated all these systems via open data interfaces spewing data into Big Data engines that served functions such as assurance and service orchestration. Lengthy, expensive, reliant on consultants and architects, and in the end not very close to real time. Also not good. Unless you were the one billing for the change management and transformation project.
As the software end of the industry gathers in Copenhagen for TM Forum’s DTW Ignite event, it does so with a new(ish) ambition – unlocking the power of AI to take care of this operational environment, with performance and fault management, assurance, orchestration all aligned as intent-driven processes which are understood, implemented and tracked by AI and AI agents. Where humans are involved, they will be interacting with their networks and systems using natural language interactions and queries.
This is all to be enabled by a new approach to network and system data. Instead of pulling everything into warehouses and lakes, and churning squillions of bytes to find nuggets of truth, AI will exploit a unified data platform that has been pre-contextualised, correlated and formatted, also by using AI. Data can be actioned where and when appropriate by the relevant application.
As with all stories, telling the tale is easier than making the reality come true, but there have been steps towards this AI-enabled future for many years. And this year the challenge has very much been around controlling and shaping the network data story for AI, as this article from MWC 2025 reported.
Ericsson’s Telco DataOps
This morning Ericsson has announced an “evolved” OSS-BSS portfolio that puts some practical steps onto its own AI vision, and puts it squarely in the “fight for your data” narrative.
Amongst a series of measures to evolve its OSS-BSS capabilities, it has invested in its Mediation platform, adding real time data streaming, aggregation and correlation to rechristen it as a Telco DataOps Platform. It has also upgraded its Service Orchestration and Assurance with AI capabilities to enable network operators to deliver an intent-based capability from the network.
Rather than position this as a brave new world, Ericsson is keen that this evolution is seen as a practical approach, that is realistic about progress and speed of change in AI introduction.
Ericsson’s Jason Keane
Jason Keane, Head of Business and Operations Support Systems Portfolio, Ericsson, said that the changes to the Mediation product are exactly about moving towards a more intelligent use of data.
“We learnt that figuring out your problems from all the data doesn’t work at all. We don’t lack data, but we need a knowledge plane concept (which Telstra will talk about at DTW). What’s important is knowing where to get the answer you are seeking. That means that data can be correlated locally and only give you the data you need, summarising events in terms of anomalies and frequency of events, rather than saying ‘Here’s one billion events’.”
“If we have that sort of data you can have a better input into the AI function. Ericsson is not building LLMS, we can’t compete with the investment in that sector. But if we have telco knowledge and good data then we can better construct that into an LLM to get good output.”
We wanted to get away from the hype of what you can do, to providing proof points. It’s not about the agentic future, but about creating value now
Keane describes that data management as a key “secret sauce”, but one that undperins a pragmatic approach.
“It’s an evolution to a platform that sees 40 billion CDRs a day – it scales at that level. So we wanted to get away from the hype of what you can do, to providing proof points. It’s not about the agentic future, but about creating value now.”
But if you do look at that future, Keane acknowledges the need for industry transformation.
“If you look at all the trends, workloads themselves – CNFs, VNFs – are becoming generic workloads running on compute as standardised Kubernetes. There’s a better way to manage than with bespoke systems. Ultimately we are managing software – it will be a long trip to get there but if we do not take it on in the OSS then we are going to struggle.”
Being realistic
Keane is aware that introducing new operational models is not easy. For example he describes 5G slice assurance as a major achievement, and one that took way more work than he thought it would.
Referencing a large US Tier One that Ericsson works with, Keane said Ericsson has been able to combine existing SON and Service Orchestration into one SMO product. This has enabled much more flexible cross-domain management of a network that has tens of thousands of gNodeBs, each generating millions of events every five seconds.
“The problem was connecting data models, so merging into one product mean we had one data model and one product.”
That means he is realistic about the hype about AI and Agentic AI.
“If you break it down, Agentic AI describes programmes that all add together to give an output. When we started with agents there were two problems. One was that not every workflow needed an agent, where the process is to describe an outcome and keep looking for it. The second was that what you asked for was not always what you got.
“So we found there’s a lot more to it. But if you have good data you can tune the model, and make a start with making sure agents can control products such as service orchestration and monetisation. And we have published a number of agents that can control products.
“An important step was to have somewhere where you could design and test AI agents before you push the button. For that we picked AWS. Assuring agentic output is super-hard, so we have used the fundamentals of AWS, Bedrock, governance and observability,” Keane said.
This focus on adding AI capabilities to OSS is the subject of a separate press release from Ericsson. This includes the launch of the Gen-AI Lab with AWS, the Ericsson Telco Agentic AI Studio and Ericsson Telco IT AI Apps.
Ericsson also has two operator references to back up its new Agentic AI and – with Grameenphone and Orange.
Grameenphone will be the first CSP to use the new Ericsson and AWS Gen-AI Lab to create an Agentic-AI powered OSS/BSS solution. Grameenphone is using the Lab to create and develop an Agentic-AI based solution to automate the process of migrating its legacy product catalog to Ericsson Catalog Manager.
The solution uses a series of Gen-AI-powered agents to understand Grameenphone’s existing product catalog, translate those products into a set of business requirements and high-level designs, configure the products in the Ericsson Catalog Manager platform and test each product to ensure accurate migration and creation.
Orange has selected Ericsson to automate the creation of 5G services for enterprises and subscribers in Europe. A blended Ericsson and Orange team will develop a range of market offerings, underpinned by Ericsson Service Orchestration and Assurance and other Ericsson portfolio elements.
Laurent Leboucher, Group Chief Technology Officer at Orange, said, “This partnership with Ericsson is an important step in our journey towards cloud native and demand-centric, autonomous networks. Ericsson’s service orchestration solution will run above our horizontal Orange telco cloud infrastructure and will enable our 5G networks to deliver differentiated experience to our business and retail customers.”
Winning the AI interface
Ericsson’s AI for OSS update follows a release a few days ago from Rakuten Symphony, which detailed how AT&T has expanded use of its AI-enhanced Site Management software to handle deployment and management of its cell site estate. Along with its Intelligent Network Operations suite, Site Management is one of the two foundational products via which Rakuten Symphony markets and bases its OSS capabilities. These wrap up capabilities such as assurance, fault and performance management into a structure which sees automated applications fed by a common data and observability layer.
The structure leverages data extraction and observability to enable use cases such as anomaly detection, RIC/rApp operation, automated configuration, zero touch provisioning and so on.
Rakuten Symphony was keen to expand on the meaning of that announcement, telling us that it is driving towards a future network operations paradigm, where the interface with and between AI agents and operations will be the key aspect that determines success.
Its head of OSS Vivek Murthy said that Rakuten Symphony is building its own LLMs and SLMs, using technologies such as Retrieval and Cache-Augmented Generation to build an AI layer on top of its data. The aim is to be able to create a Natural Language Programming interface that can intelligently understand the intention of the operative, as well as co-ordinate between AI capabilities such as AI agents, making the necessary links and connections to fulfil an operation.
There’s no doubt that the network operations part of the DTW Ignite programme is going to be dominated by discussions about AI, in all its forms. Companies, even the biggest, are having to decide where and how they play in the space, and make sense of that in the context of their current presence. Ericsson is clearly positioning itself as a pragmatic hand, instrumenting its existing capabilities with AI use cases of proven value, leveraging its managed services capability. But it is also willing to look outside of its competencies – indeed acknowledges that it has to do so – with partnerships with AWS and Google Cloud a key part of its make-up.
The story of network operations is at the beginning of this new chapter, and it looks likely to have many authors.
