By: Ram Ramanathan

AI is already transforming network operations, but realizing its full value begins with understanding the problem it is meant to solve. AI assistants, copilots, and analytics tools offer meaningful benefits to teams managing increasingly complex networks with limited resources. These technologies can surface relevant information, summarize events, identify anomalies, and recommend next steps. While such improvements represent a practical first step toward AI adoption, they should not be confused with autonomous operations. True autonomy requires a fundamental change in how networks are managed and operated. Faster processes layered on top of fragmented data, manual intervention, and reactive workflows may improve efficiency, but they do not transform operations. 

This distinction is becoming increasingly important for communications service providers. Many organizations are turning to AI extensions as the fastest path toward automation and operational transformation. While these capabilities deliver clear benefits, a more strategic question remains: can the operational architecture beneath them support the next generation of intelligent networks, which rely on real-time data, contextual understanding, and increasingly autonomous decision-making? AI extensions may be a practical starting point, but they are unlikely to deliver the full level of autonomy operators ultimately seek. Achieving that objective requires a foundation built on unified data, contextual awareness and automation.

When AI Runs into Legacy Limitations

Operators have made significant investments in infrastructure, operational systems, processes and skills over many years. Those investments cannot simply be discarded as new technology cycles emerge, making the idea of introducing AI into existing OSS environments understandably appealing. AI extensions offer a relatively efficient and minimally disruptive way to introduce new capabilities into established workflows while reducing manual effort and improving productivity. 

But achieving autonomous operations requires more than incremental efficiency gains. Traditional OSS platforms were designed to support human-led, reactive workflows rather than real-time, autonomous decision-making. AI extensions frequently rely on customized integrations connecting multiple legacy systems together, making it difficult to create the unified operational environment autonomous networks require. As a result, AI often improves existing workflows without fundamentally redesigning them. 


In many environments, data remains incomplete, delayed, or disconnected from related systems. An AI assistant may successfully identify an anomaly, but engineers may still need to consult multiple operational systems to determine business impact, identify root cause, and execute remediation. The workflow becomes faster yet remains fragmented. 

This is the central weakness of treating AI as an add-on rather than as part of a broader operational redesign. Extensions can improve access to information and streamline user interactions, but they do not automatically solve the underlying issues of fragmented data, limited context, and manual execution. In some cases, they may simply make existing processes more efficient rather than helping operators rethink how decisions should be made and acted upon. 

As networks become more dynamic, distributed, and interconnected, these limitations become increasingly difficult to overcome through AI overlays alone. To move beyond incremental gains, operators must evaluate the operational foundation that supports AI, not just the AI capabilities themselves. 

Autonomous Operations Require the Right Foundation

AI systems must understand how events, services, and infrastructure relate to one another across the network in order to support meaningful autonomy. An alarm on a network component may indicate a hardware failure, a configuration issue, a capacity constraint, or a downstream service problem. Without visibility into broader operational context, AI can identify symptoms without understanding their significance. 

This becomes even more important in today’s multi-vendor, multi-domain environments. Modern networks span physical infrastructure, virtualized functions, cloud-native applications, transport networks, routing domains, and customer-facing services. Events occurring in one layer frequently affect another. A domain-specific view may fail to reveal how a performance issue affects service quality, customer experience, or operational risk elsewhere in the network. Autonomous operations require end-to-end visibility and systems capable of understanding these relationships.