Autonomous networks are often discussed as an AI milestone. At MWC 2026, the more useful framing was operational: autonomy will depend less on the presence of AI than on whether operators can build OSS environments capable of turning intelligence into action.
That distinction shaped Rakuten Symphony’s panel on AI-powered OSS and the path to autonomous networks, which examined what has to change for telecom to move beyond automation as a support function and toward autonomy as an operating model.
Moderated by James Dartnell, Director, Corporate Communications, Rakuten Symphony, the session featured:
The discussion made one point especially clear: the industry’s challenge is no longer understanding what AI can do. It is building the operational, architectural and organizational conditions that allow AI to do it safely, repeatedly and at scale.
For years, AI in telecom has largely been used to detect patterns, generate insights and support decisions. The next phase is different. As Dr. Gerszberg noted, AI is beginning to influence network parameters directly, which raises the bar for the systems around it.
That is why programmability matters. If the network cannot be exposed, orchestrated and adjusted through software, AI remains observational rather than operational. In that context, the service management and orchestration (SMO) layer becomes more than a technical component. It becomes the control point that allows intelligence to translate into execution.
The panel also reinforced a harder truth: autonomous operations will not be built on models alone. They will be built on data environments that are broad enough, accessible enough and structured enough to support continuous learning and intervention.
Dr. Gerszberg pointed to a familiar telecom constraint: many OSS environments still reflect architectures in which applications own their own data. That limits visibility and weakens the foundation AI depends on. Bringing network and IT data together across domains is not a side project. It is the prerequisite for meaningful autonomy.
Murthy described Rakuten Symphony’s OSS approach in similar terms, with the data layer as the foundation for everything above it. “Once data from legacy and newer network environments can be ingested in a consistent way, orchestration, assurance and remediation become easier to operationalize and AI becomes embedded in the system rather than added onto it.”
One of the more important ideas in the session was that OSS is moving beyond dashboards and copilots toward agents.
Murthy described conversational AI as the entry point. A useful interface for querying systems, surfacing insights and guiding human decisions. But the more consequential shift is toward agentic systems that can execute tasks, manage subprocesses and coordinate actions with human oversight.
That is the logic behind Agent Studio, Rakuten Symphony’s framework for building agents on top of OSS products and data environments. The significance is not just technical. It signals a different operating model, one in which autonomy is introduced in layers, with modular functions, security guardrails and lifecycle management built in from the outset.
The panel made clear that the path forward is not frictionless.
Murthy pointed to the cost and complexity of large-scale AI deployments, particularly in environments where control and repeatability matter. More focused models, machine learning and modular architectures might prove more practical than applying large language models indiscriminately.
“Network operations data is highly time-sensitive and often ephemeral, making it poorly suited to many general-purpose AI environments. And as autonomous systems multiply, operators will also face a new coordination problem, how agents resolve conflicts, negotiate priorities and act without creating instability.” - Dr. Gerszberg, Head of Group Future Networks, Axiata Group.
The decisive issue may be organizational as outlined by Dr. Wu. “AI-powered OSS will only scale if operations teams trust it. That means change management, explainability and a clear understanding of what the system is doing and why. In network operations, trust cannot be assumed. It has to be engineered as carefully as the platform itself.”
What emerged from the discussion was a more grounded view of autonomy. The breakthrough will not come from AI alone, nor from a single product layer. It will come from aligning data, programmability, agent frameworks and human confidence into one operational system.
That is what makes AI-powered OSS strategically important. It is not simply another software category. It is the layer through which autonomous networks become governable, usable and commercially real.
