NGMN says Agentic AI needs alignment for mobile networks

NGMN has published a report on the use of Agentic AI in autonomous mobile networks, saying wider industry alignment is needed for large-scale deployment.

The publication, titled Network Automation and Autonomy Phase III: Agentic AI for Autonomous Mobile Networks, examines the motivations, opportunities, requirements and ecosystem developments linked to applying Agentic AI to network automation. It sets out a reference framework for AI-driven automation, reviews the maturity of standards and open-source work, and outlines the conditions for safe, interoperable and scalable deployment.

Agentic AI is likely to play an important role in creating a path towards trusted Level 4 autonomous mobile networks, the alliance said. However, it argued that commercial adoption will depend less on individual AI models and more on support systems, governance, operational controls and coordination across the telecoms ecosystem.

Among the areas identified as necessary for operators are governance, trust, policy adherence, observability, explainability, operational safety, determinism, assurance, security and cost control. The report also warned of fragmentation if operators, vendors, standards bodies, hyperscalers and open-source groups do not align on architectures, interfaces and responsibilities.

NGMN is a mobile industry alliance whose members include operators, vendors and academic participants. Its work is intended to guide the development of mobile network infrastructure and services in areas including 5G, 6G and network disaggregation.

The report comes as telecoms groups explore more advanced forms of automation beyond analytics and closed-loop optimisation. In this model, software agents would be expected to reason, plan and carry out actions across several operational domains, rather than support only narrow tasks.

That shift raises questions about oversight and operational risk in networks that underpin consumer and business communications. Operators will need controls that allow human intervention where necessary, along with mechanisms to validate the behaviour of software agents in live environments, the publication said.

Laurent Leboucher, Chairman of the NGMN Alliance Board and Orange Group CTO and EVP Networks, said: "Mobile network operators (MNOs) find themselves at a decisive inflection point as the telecoms industry enters a new phase of network automation. AI is no longer restricted to analytics, prediction, and closed-loop optimisation. Indeed, the rapid arrival of Agentic AI has introduced the prospect of autonomous systems able to reason, plan, collaborate and execute actions within and across multiple operational domains.

"However, the complexity of networks and high expectations surrounding customer experience means general-purpose Agentic AI technologies must be used carefully, with the right harness and guardrails to bring humans in the loop when needed. This is a learning process in itself where operators will have to take controlled risks."

Ecosystem gaps

NGMN's review of standards development organisations, industry initiatives and open-source efforts found momentum in Agentic AI work, but described the landscape as uneven. In practice, that means parts of the ecosystem are moving ahead while others remain at an earlier stage of development or lack common approaches.

Guangyi Liu, Chief Expert of China Mobile and NGMN Board Director, said: "Review of standards development organisations (SDOs), industry initiatives and open-source activities demonstrates encouraging momentum across the ecosystem in Agentic AI deployments, but despite this the landscape remains largely complex and unevenly mature.

"Increased collaboration among MNOs, SDOs, vendors, hyperscalers, and open-source communities is therefore important to achieve confidence in the opportunity of large-scale deployments and provide a more unified end-to-end perspective. That means cross-industry coordination on areas such as telecom-grade interfaces and trust mechanisms, as well as shared testing environments, and evidence-based validation of agent behaviour to ensure solutions remain interoperable, secure and fully aligned with operator requirements."

The report also says the market is likely to move from isolated proofs of concept towards operational deployments centred on tightly controlled, high-value use cases. Those use cases are expected to involve agents managing more complex workflows across multiple network domains, increasing the need for common standards and reliable governance.

Operational focus

For operators, the issue is not only whether AI agents can perform tasks, but whether they can do so consistently and safely inside critical telecoms infrastructure. The publication therefore emphasises explainability, observability, interoperability and human oversight as practical requirements for wider use.

Luke Ibbetson, Head of R&D at Vodafone and NGMN Board Director, said: "Capabilities for trustworthiness, security, explainability, observability, interoperability, governance and appropriate human oversight must be in place for Agentic AI in order to accelerate safe, interoperable, and scalable deployment across the telco ecosystem.

"If the related requirements are met, Agentic AI can help create a realistic route toward trusted Level 4 autonomy where networks are highly autonomous and enable automated decision making, self-optimisation, self-healing and self-management."