BT, Swisscom back AI agents for networks but not closed-loop yet

BT Group, Swisscom, Netcracker and Liberty Global said at FutureNet World in London that AI agents are a long-term part of network operations, but they are not yet ready for closed-loop use. The panel said trust, governance and the risk of network faults must be addressed before operators allow AI agents to act autonomously on mission-critical network functions.

BT Group managing director of mobile networks Reza Rahnama said the operator views its network as a whole and is using that approach to guide AI deployments, rather than relying on multiple tools with incomplete information. Swisscom lead architect for network and infrastructure IT Rudolf Strijkers said Swisscom currently uses AI agents only in very restricted cases. Netcracker AVP of strategy Susan White said consumer large language models, or LLMs, may distort expectations because they are probabilistic systems that simulate reasoning, and she said operators cannot yet allow AI agents onto mission-critical parts of their networks because a fix in one area could create faults elsewhere.

The discussion reflects a broader telecoms industry debate about how quickly operators can move from AI-assisted operations to self-optimising and closed-loop networks. Liberty Global managing director for network and capital allocation Luk Bruynseels said current AI is limited but likely to improve rapidly, and he urged the industry to share how it applies AI to security and data challenges. The panel's comments show operators and vendors see AI agents as useful for network operations, while keeping human oversight in place until governance and reliability improve.