Telecom Networks Heal Thyself…with AI!

What happens when you put a bunch of overachieving network engineers in a room and hand them an intractable problem: Can an artificial intelligence (AI)-driven self-healing “network brain” be created to automate and dramatically accelerate fault detection and resolution in live network operations center (NOC) environments?

That is what happened recently at TM Forum’s NeuroNOC Catalyst, an innovation project at DTW Ignite 2025. Many of the leading communications service providers (CSPs) and tech companies gathered to take part in a ground-breaking project exploring how AI agents, closed-loop automation, and high-quality network data can enable a self-healing network brain capable of tackling the costly and time-consuming manual troubleshooting process.

Catalyst Program Aims for Next-Gen Innovation

This year’s event brought together global technology leaders, including Amazon Web Services (AWS), Accenture, Symphonica, Sand Technologies, and NETSCOUT, to collaborate on the development of new operating models for troubleshooting and repairing complex telco networks. The “catalyst” program with NETSCOUT was developed in collaboration with carriers such as BT Group, Telecom Argentina, Omantel, TurkNet, AXIAN Telecom, and Safaricom.

Initiative participants worked together to create an innovative proof of concept demonstrating how AI and automation can dramatically accelerate fault detection and resolution in live network environments. The idea was to harness AI to combine and analyze network operations data from many different sources to enable CSPs to automate the swift and accurate resolution of issues. By bringing AI to bear on this critical area of concern, CSPs would benefit through substantially lower costs, and perhaps most importantly, dramatically improved customer experience.

High-Fidelity Curated Data Is Key to the Proof of Concept

One of the key takeaways from this initiative is the validation that high-quality curated data is needed to drive effective AI solutions. Without high-fidelity packet collection across the network, it is difficult, if not impossible, to identify and correlate issues across multiple data streams, determine root causes, and verify and test automated fixes. AI models require curated data to achieve the full potential of the technology and deliver stellar results.

The AI solution achieved in this case used sophisticated curation of CSP-specific domain data and the fine-tuning of large language models (LLMs) into highly capable small language models (SLMs). This allowed the generative AI to discover network topology, analyze issues encountered, visualize potential customer impacts, identify the root cause quickly, plan resolution steps, and then execute the remediation with remarkable precision and efficiency.

Enabling the collection of the curated data needed to power the solution was the NETSCOUT Omnis AI Insights solution. This entailed the use of Omnis AI Sensors for deep packet inspection (DPI)-based, end-to-end network visibility across the 5G standalone radio access network (SA RAN). Omnis AI Streamer was then used to provide powerful analytics and filtering at the source via an open API-driven dataset.

The solution was put to the test in simulated scenarios where service was adversely impacted. The result was that NOC engineers were able to identify subscriber registration issues, pinpoint the root cause via a curated LLM, and quickly resolve issues with little manual effort. In fact, the solution was able to reduce manual troubleshooting by 80 percent and lower operational costs by as much as 50 percent.

NETSCOUT Senior Director of Product Marketing Richard Fulwiler participated in the Catalyst initiative. “While fully autonomous networks are still in their early stages,” he noted, “this project shows the powerful potential of AI agents—armed with the right data in real time—to support faster and more accurate resolution of network issues.”

Read more about NETSCOUT AIOps solutions for service providers.