Nokia has announced an enhancement to its comprehensive management and automation platform, Network Services Platform (NSP), with the introduction of an agentic AI framework for IP network operations.

The Finnish telecom has designed the new framework to help network operators deploy AI agents that can reason over real network context and take guided actions according to defined policy and security boundaries.

While AI efforts advance to support network operations, companies have remained cautious amid explainability, trust, and risk concerns in production environments.

Nokia has created its approach with NSP, targeting such concerns by embedding agentic AI capabilities into the platform that already serves as the authoritative controller for IP networks.

“Appledore has been advocating for operators to focus on the primary importance of quality data and ontological relationships – which are proving far more important than specific AI models for efficient and accurate AI reasoning,” Grant Lenahan, partner and principal analyst at telecom industry researcher Appledore Research, said.

“Nokia’s NSP embraces this approach with extensive AI-native infrastructure built on trusted data and operating norms, providing a solid and secure foundation for a myriad of AI use cases. Domain expertise is likely the most critical quality in designing effective automation for complex networks.

Enhancing NSP with AI agents

The telecom explained that NSP grounds AI agents in an accurate and continuously updated view of the network, including topology, protocol behavior, configuration state, service relationships, and recent network changes.

This then drives AI agents to reason based on network truth, rather than inferred or fragmented data, and to operate in operator-defined intent, policies, and access controls. The NSP agent framework also permits communication with external agents via AI-based protocols, such as the model context protocol (MCP), across operators’ multi-vendor, multi-domain networks.

An AI-driven troubleshooting agent is the first use case built on this new framework, designed to help operators identify root causes faster, reduce operational noise, and resolve complex IP network issues. According to Nokia, this is a significant step in its strategy to help operators adopt AI safely, incrementally, and at scale in live networks.

As trust remains the deciding factor when considering AI-native operations, Sasa Nijemcevic, vice president and general manager of IP network automation software at Nokia, said: “We are enhancing NSP with AI agents built on an agent framework in a way that respects how networks are actually operated.

“This will have a major impact on the way operators manage their networks and will enable them to enhance their operations significantly and accelerate their journey toward autonomous networks with a focus on solving real operational problems, starting with high-impact use cases like troubleshooting. This is an incremental, pragmatic step toward AI‑native networks.”

Nokia said that the new agentic framework can assist network operators with a flexible foundation to introduce multiple AI use cases over time, without creating siloed solutions. Operators can start with focused, high-confidence scenarios and gradually expand the role of AI as trust builds.

This NSP enhancement is expected to be commercially available by the end of 2026.

Turning to other AI developments, Nokia’s recent showcases at Mobile World Congress 2026 saw the giant double down on AI-augmented radio access network (AI-RAN) efforts, revealing new additions to its AirScale portfolio and results of successful field tests with Nvidia.

In March, Nokia launched its Doksui line of remote radio heads (RRHs), with new radios “engineered to meet the performance, sustainability, and automation demands of the AI era,” the operator said.

Nokia claimed the new foundational component of its AI-RAN efforts offers up to 30% improvement in power efficiency and reduces site footprint with Doksuri radios being 25% lighter than prior systems.