A few years ago, one of Canada's largest telecom operators experienced a massive outage that left millions of people without mobile and internet access for nearly 15 hours. The disruption affected banking, transportation, government services, and more. ATMs and cashless payment systems ground to a halt, and even some emergency services were impacted.

What was the cause?

A simple routing or configuration error in the operator's network triggered a cascade of failures. Specifically, a configuration change in the core network sparked widespread issues, but the network didn't collapse immediately. It wasn't until the problem propagated through the data plane that traffic essentially stopped flowing. The outage created hidden routing loops and packet-forwarding failures that were not apparent in the control-plane signaling.

Telecom engineers had to piece together scattered logs and partial traces, slowing down root cause analysis. While overall traffic plummeted, the exact choke points remained unclear. The operator lacked granular, end-to-end visibility into user plane flows, so they couldn't quickly pinpoint where the failure originated or how it was spreading across routing and forwarding paths. Control plane metrics alone were insufficient to reveal where user data was breaking.

Control Plane Alone Isn't Enough

Had the operator possessed complete, correlated visibility across the user plane, they could have isolated and resolved the issue far faster, minimizing disruption for millions of people. When investigating a network anomaly, the control plane can indicate that the network is "up and running," but it does not show whether customers are actually receiving service as expected.

A GSMA survey, conducted in partnership with RADCOM, found that only 41% of operators have an end-to-end architecture that integrates data across all departments. That means nearly 59% of operators lack the complete, correlated visibility required to quickly isolate and resolve critical issues, leaving both networks and customers vulnerable.

Agentic Al Missing the Mark

This gap becomes even more critical as operators accelerate the adoption of agentic Al across their networks. Unlike traditional Al or ML, agentic Al and Al agents can detect issues, prioritize what matters, decide on actions, and even trigger remediation. However, agentic Al needs to be grounded in truth, not assumptions or approximations. When deciding what problems to fix first, for example, without user-plane data, the Al might only see generic signals like "packet loss detected" or "cell utilization at 65%". It will not understand the real impact, such as "Netflix streaming degraded at 21:04."

Real-time, accurate, and granular user-plane data, correlated with the control plane, enables Al agents to understand that a generic "packet loss" is actually affecting something specific, such as "video traffic on a particular UPF path." With only control plane data, the Al can see that sessions exist, but it cannot determine whether those sessions delivered a satisfactory customer experience.

Agentic Al promises to reduce churn and enhance the customer journey. However, without connecting the (essential) dots, it risks becoming just another overhyped technology, rather than driving a true shift in user experience. Without accurate, reliable, and granular user-plane data, agentic Al is left to optimize the network based on KPls that overlook real customer pain points, potentially automating the wrong fixes. By leveraging user-plane data, Al agents can measure app-level quality of experience, identify SLA violations, and track session drops on a per-call basis, turning raw data into actionable insights that directly improve customer satisfaction.

Intelligence Assurance: From Blind Spots to Full Visibility

RADCOM's Al-native assurance provides end-to-end network visibility by correlating deep user plane insights, behavior, usage patterns, and service interactions with control-plane data. The solution, which provides data capture at bandwidths reaching up to 800 Gbps on a single server, captures 100% of network traffic on the user plane in real time. This offers a unified, actionable view and lays the data foundation for Al-driven and agentic networks.

With deep insights and predictive reasoning, these Al-driven capabilities can optimize network operations, anticipate anomalies, and proactively enhance customer experience.

In today's Al-driven era, with operators able to leverage vast amounts of customer data, expectations for exceptional service have never been higher. Delivering a truly optimal customer experience requires complete visibility across the RAN and core, spanning both the control and user planes. Without this granular insight, deploying agentic Al is like jumping into the deep end before learning to swim. Ignoring the impact of anomalies on the user plane leaves Al "flying blind," costing operators customer loyalty, retention, and valuable time.