Towards autonomous networks with agentic AI and real-time simulation

Telecom operators want networks that manage themselves, but today most automation still needs a lot of human oversight. Vendors are now pushing "agentic" AI and real-time digital twins as the next step. In this model, software agents watch live network data, test changes in a simulated copy of the network, and then push approved actions back into production.

Supporters say this could speed fault detection, cut manual engineering work, and let networks adapt faster to changing demand. The approach depends on accurate, continuously updated simulations, clear guardrails on what AI agents can change, and ways for humans to review and override decisions. The industry is still early in deployment, with most projects in trials or limited domains rather than running entire networks end to end.