Nokia and Google Cloud ship six Gemini agents for telecom

Nokia and Google Cloud announced a six-agent Gemini stack for telecom network operations on June 22, embedding Google's multimodal models inside the Nokia Assurance Center. Two of the six agents are already live, and the rest ship as a SaaS package on the Google Cloud Marketplace in September 2026. The deal lands the same week telecom operators are pulling agent tooling into the same procurement cycles as their radio and core network gear.

Nokia Assurance Center is the company's flagship network management platform, sold to tier-one telecom operators as a unified software layer over their radio, transport, and core networks. The new partnership integrates Google's Gemini models into the Assurance Center through a six-agent architecture that Nokia developed on Google Cloud's Agent Development Kit, the same framework Google uses internally to build production agents. The agents are designed to operate independently or in concert, and they cover the full lifecycle of a network incident: from initial signal, to root cause analysis, to action recommendation, to human approval, to automated remediation. The six agents are Router, Event triage, KPI selector, Anomaly reasoner, Action reasoner, and Dashboard. Two of the six, the Router agent and the Event triage agent, are already fully functional in the current Nokia Assurance Center release. The remaining four ship in rolling software updates over the rest of 2026, with the full SaaS package landing on the Google Cloud Marketplace in September 2026.

The Router agent acts as the central orchestration layer. It interprets user intent, manages communication between the other agents, and enforces compliance with the operational guardrails the operator has set. The Event triage agent analyzes ongoing alarms and compares them against historical patterns to identify root causes and assess operational impact. The KPI selector agent provides domain-expert interpretation of complex network performance metrics, definitions, and measurement units to aid reasoning. The Anomaly reasoner agent investigates unusual network behavior to determine whether a deviation is a genuine issue or a false alarm. The Action reasoner agent matches active events against automation catalogs to recommend specific remediation steps. The Dashboard agent enables teams to quickly generate visual analytics and tracking screens using natural language prompts. Each agent runs on standard Google Cloud compute and storage, deployed via Kubernetes, so operators do not need a custom managed-services overlay to adopt the stack.

The glass box autonomy framing and what changes for operators

The most distinctive part of the partnership is the operational philosophy Nokia calls glass box autonomy. Rather than removing operators from the loop, the framework introduces a confidence-based recommendation layer that presents proposed actions to human engineers and retains final approval over critical control points before any fix is executed. The action reasoner agent serves as the advisory layer in that loop. For low-risk, policy-approved scenarios, the same architecture can also support fully closed-loop automation, but the default is human-in-the-middle, with the agent proposing and the engineer approving. The framing is a deliberate counterpoint to fully autonomous agent narratives and reflects what tier-one operators have been asking for since the first wave of agent deployments in 2024 and 2025: agents that do the analytical work and surface the action, with the human engineer retaining accountability for the final call.

Nokia's published performance claims are aggressive. The company says the agent stack can cut network problem-solving times by 50% to 80 percent, with complex issues such as voice degradation or software errors that historically took hours to isolate now flagged and resolved in minutes. The agents can filter out data fluctuations to focus on genuine problems, reducing false alarms and unnecessary escalations. Staff can use simple language to instantly generate dashboards and performance reports, removing a long-standing bottleneck in network operations centers where dashboard construction typically requires specialist tools. The cost story is also explicit: the agents run on top of an operator's existing cloud setups, avoiding the need for expensive custom software overlays. The pitch to a tier-one CFO is straightforward: agent tooling on existing cloud spend, with no additional managed-services contract.

How the partnership compares to the rest of the telecom agent stack

The competitive context is mixed. Ericsson has been shipping AI-based automation inside its network management suite for years through the Ericsson Operations Engine, and Huawei offers its own AUTIN framework, but those are largely closed stacks with proprietary models. The Nokia + Google Cloud deal is the first major network vendor partnership to anchor an agent stack on a third-party frontier model family, and the choice of Gemini as the underlying model gives Nokia a different technical and commercial posture than the Ericsson or Huawei approaches. The two named executive voices in the launch are Vivek Jaiswal, Nokia's senior vice president of Autonomous Networks, and Sridhar Gollapudi, Google Cloud's global telco market lead. The partnership is a bet that telecom operators will adopt agents faster when the underlying model is already familiar to the procurement teams buying it, and the answer in 2026 appears to be yes for several of the largest carriers, who are already running Gemini workloads in adjacent areas such as customer service and field operations.

The wider agent story is well documented in our recent piece on the Nvidia telecom agent stack at DTW Ignite 2026, which covered the parallel move by Nvidia, SoftBank, and others to put agentic tooling into the same telecom operations workflows. The difference is that Nokia's bet is on Gemini as the model and Google Cloud as the runtime, while the Nvidia DTW story was about a multi-vendor RTX PRO 6000 Blackwell hardware stack with multiple model partners. The two approaches are not exclusive, and several tier-one operators are likely to run both in different parts of their network management portfolio: Nvidia for hardware-accelerated RAN planning and digital twin workloads, and Nokia + Google Cloud for Assurance Center operations and incident response.

What the partnership means for agent procurement in 2026

The most useful framing is that the partnership positions Gemini as a real option for telecom network management, not just customer-facing agent flows. The same Gemini Enterprise Agent Platform that powers Google's own customer service agents and a growing list of third-party enterprise agent deployments is now the foundation for Nokia's Assurance Center. The agent development kit gives Nokia engineers the same primitives Google uses internally, and the deployment model runs on standard cloud infrastructure rather than a custom managed-services overlay. The shift in telecom network management is the same shift in the rest of the stack. The system has to be agent-readable, not just human-readable, and the procurement decisions are increasingly about which model family and which cloud the agent stack will run on.

The risks are real. The partnership makes Nokia more dependent on Google Cloud for a critical operational product, and any Gemini model deprecation, pricing change, or availability issue ripples directly into the Assurance Center. The glass box autonomy framing is also a bet that tier-one operators will be comfortable with the action reasoner agent's recommendations, and that bet has not been tested at scale yet. The 50 to 80 percent reduction in problem-solving times is a vendor claim, not a third-party benchmark, and the September 2026 SaaS launch will be the first time operators can independently verify the performance numbers in their own networks. The rollout cadence, with two of the six agents live today and the remaining four coming via rolling software updates, also means the full stack will not be in production for most operators until late 2026 or early 2027. The release lands the same week Anthropic's Claude Code can be tricked into a reverse shell by a clean GitHub repository through a DNS TXT record, and the same week vendor after vendor is shipping best AI coding agents in 2026 and adjacent tooling that needs a runtime like Gemini or a competing frontier model to actually do work. Nokia's bet is that the answer for tier-one telecom is Gemini on Google Cloud, and the September 2026 launch will be the first public test. The full launch post and the primary source are on the Nokia newsroom announcement and the PRNewswire release dated June 22, 2026.