How Telcos Build Autonomous Networks with Agentic AI

Summary

Telecom operators currently operate mostly within Levels 2 and 3 of the TM Forum autonomous networks taxonomy, relying on predefined scripts. Progressing to Levels 4 and 5 demands autonomous systems capable of intent comprehension, real-time telemetry sensing, and cross-domain action coordination. To address this, an autonomy platform architecture combines domain-specific reasoning models, agent harnesses, and isolated sandboxed runtimes. Workflows separate responsibilities among on-demand agents, long-running execution loops, and deep-research multi-agent systems that evaluate ranked remediation paths or design wireless algorithms. Initial experiments in simulated SR-MPLS routing and wireless physical or MAC layer design show autonomous discovery can deliver measurable gains over standard operational and engineering baselines.

Takeaways

1. 1

   Network operations follow three agentic problem patterns: execute paths for known issues, optimize paths for measurable outcome improvements, and discovery paths for unencountered failures.

2. 2

   Deep research multi-agent systems organize into planner, researcher, and orchestrator agents to simulate and rank alternative operational plans against digital twins and telemetry.

3. 3

   In wireless R&D experiments, an agentic evolutionary search using NVIDIA Sionna delivered over a 3% spectral-efficiency gain against standard link adaptation baselines.