Autonomous networks driven by AI are exposing the limits of legacy OSS built for static infrastructure. Operators now need an evolved, ontology-driven control layer that can unify fragmented data, govern automation, and orchestrate intent-based, real-time network operations at scale.
The telco industry has arrived at a fundamental crossroads: advanced automation and agentic AI require a new operational model to govern them. Operating Support Systems (OSS), the suite of software used to coordinate, orchestrate and deliver network services, were designed to accommodate a world of static hardware, but modern networks are highly dynamic. What is missing isn’t more automation, but an evolved OSS; a system that can govern autonomous operations at scale.
The continuous pressure AI imposes on networks is at odds with legacy OSS, which was built for predictability. These systems were designed during an era of human-led workflows to manage, stabilize and organize inventory. At that point, service providers had control over the entire planning stage and could feed into the system retrospectively, but ultimately, this was limited to managing the infrastructure, rather than the outcomes. The dynamism of the modern network cannot be provisioned in such a rigid, manual way.
The problem with spaghetti integration
5G, edge, slicing, and multi-domain architectures introduce networks that are no longer static but continuously changing and designed to be self-driven. These features break down legacy OSS assumptions, rendering many traditional systems obsolete. Functions such as bandwidth-on-demand mean networks are in constant flux. Legacy systems struggle because they were not designed to maintain a real-time view across this level of dynamic change.
At the same time, service providers have been layering on more modern systems on top of these legacy foundations. This creates spaghetti infrastructure – a mess of old and new systems integrated without a clear architectural strategy.
This is an incredibly bespoke way of designing infrastructure, which means when it comes to upgrading or sunsetting a single part of the system, it is nearly impossible to know the impact across the entire network. Service providers have found themselves in a situation where old and new systems are integrated without a clear map of the entire network and IT estate, leading to operational risk.
The question of ontology
AI is becoming the catalyst to change, prompting service providers to clean up decades of operational sprawl and decommissioning systems. The data foundations of legacy systems are often insufficient for AI, because you can’t apply AI against bad data. The real-time behaviors of newer systems mean the OSS cannot maintain a consistent record across data and these fragmented systems. In legacy systems, data is spread out across the entire architecture. There are no centralized data stores to house a system of record, because traditionally, they didn’t need them.
Today, this is exacerbated by explosive volumes of data and the east-west traffic patterns (traffic moving laterally between servers or services), which drive more lateral system interactions, requiring real-time, intelligent coordination. To achieve intent-based networking — where you tell the network what you want (the outcome) rather than how to do it — the system requires a unified, context-rich data set: an ontology.
The good news is that evolved systems are more dynamic. They link data from the planning stage through to orchestration and assurance, stitching it together to achieve declarative operations.
An ontology acts as a common language, stitching data together from the planning stage through to orchestration and assurance. When one part of the system changes, that change propagates through the entire network. Without this unified foundation, operators may be data-rich but remain “insight-poor,” unable to effectively operationalize AI.
From automation to governance
If a network lacks an intelligent governing system, it faces severe risks such as autonomous silos, conflicting AI decisions, and fragmented intelligence. Automation might optimize a single task locally, but it can fail the network globally.
Next generation OSS can act as the control system that coordinates and optimizes autonomy at scale. In this model, AI agents execute, humans define intent, and the system orchestrates outcomes across domains. The ultimate industry benchmark is to provide a cloud-like experience for connectivity customers. This means that services are consumed on-demand and shaped by real-time network conditions. If achieved, service providers can monetize their networks as effectively as consumer cloud tech providers.
The industry now has a clear opportunity to evolve. By establishing the right OSS foundation, service providers can move from managing complexity to orchestrating intelligence at scale. The future will not be defined by how much automation a provider deploys, but by how effectively automation can be governed.
Kailem Anderson is vice president of global products and delivery at Blue Planet, a division of Ciena. His responsibilities include global ownership of product management, engineering, delivery, support, and partner ecosystems for Blue Planet. He leads a team of more than 700 employees, to drive the vision, direction, execution, delivery and profit and loss of the Blue Planet suite of products into the telecommunications market segment.
