TM Forum has called for telecom operators to set limits on AI agents’ authority and verify their actions as they take AI into production.

In a summary of discussions at the AI-Native Telco Forum in Düsseldorf, the organisation argued that operators need to judge AI investments by business outcomes and establish controls for agents working across customer accounts, networks, and operational systems.

The report presents governance, data quality, and organisational readiness as requirements for scaling deployment. It does not provide measured deployment results or quantify the financial returns discussed.

According to TM Forum, board-level scrutiny is centred on what AI spending delivers, including revenue, costs, customer experience, and workforce impact. Its proposed approach starts with the intended outcome, then considers the information, response speed, and compute needed. The organisation argues that model selection and the choice between edge and central processing should follow those requirements.

Governing AI agents in telecoms

TM Forum distinguishes between AI agents that recommend actions and those authorised to execute changes. Suggesting a next action carries different operational responsibilities from altering a network configuration, approving a transaction, or acting on a customer account.

Its recommendation is to define which decisions an agent may make, what information it may access, and which actions require human approval. The report cautions against making human review a requirement for every AI decision, which would create additional manual queues.

Verification must keep pace with execution. TM Forum describes a control problem where an agent acts within milliseconds but the operator confirms the result much later. Operators need visibility into the action, its rationale, and compliance with policy, alongside a way to intervene.

These controls depend on the systems around the agent. The organisation identifies inconsistent processes and data as constraints that an agentic layer cannot conceal. It outlines several approaches to legacy estates, including modernisation around existing systems, adding agentic layers, and replacing fragmented systems over time.

Data architecture and agent standards

The report links the scope of AI decision-making to the quality and timeliness of the information available. Delayed or incomplete data limits real-time decisions, regardless of the agent’s capabilities.

TM Forum identifies APIs, events, modular components, and data products as architectural elements that affect how quickly an operator can respond. It calls for AI standards covering agent identity, permitted behaviour, shared semantics, observability, governance, and interactions between agents across organisational and technology boundaries.

The organisation’s concern is that separately developed agent systems could reproduce existing interoperability problems. Its proposed alternative is to organise deployment around an end-to-end business outcome, with the required processes and controls defined across IT, networks, and operations.

Connecting AI workflows across operations

TM Forum uses customer care, fault diagnosis, and engineering assistance to illustrate the limits of isolated deployments. Improvements within one function may leave overall performance constrained by hand-offs elsewhere in the process.

It proposes starting with outcomes such as preventing churn or creating and assuring a network service. Operators would then map the value streams, identify the necessary capabilities, and establish the data, processes, and KPIs required to assess execution.

The proposed sequence includes applying reusable ODA assets, governing execution, and measuring the result.

According to TM Forum: “The AI-native telco will not be defined by how many models or agents it has deployed, but by whether it can sense, decide, act across business, IT, and network domains, verify the result, and learn from it. 

“That is the shift from AI adoption to an AI-based operating model built around value.”

See also: Why network data matters for autonomous telecom networks

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events including the IoT Tech Expo and Cyber Security & Cloud Expo. Click here for more information.

Telecoms is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.