The global telecommunications industry is undergoing a massive shift as providers work to manage increasingly complex networks. Telecom operators have spent billions of dollars deploying next-generation 5G infrastructure over the past several years, yet they continue to struggle to monetize these heavy capital investments. At the same time, the operational costs of maintaining these networks are skyrocketing. Modern networks generate massive volumes of telemetry data, alerts, and performance metrics that overwhelm traditional human operations. To address these challenges, telecom providers are turning away from manual, error-prone human intervention toward highly automated, machine-speed operations.
In a major step forward for autonomous network operations, Google Cloud and Nokia have announced an expanded partnership to integrate Google’s Gemini artificial intelligence models directly into Nokia’s network software portfolio. Specifically, the collaboration will embed specialized AI agents into the Nokia Assurance Center, a central software suite used by telecommunication companies to monitor, diagnose, and maintain network health. By developing a multi-agent ecosystem built on Google Cloud’s Enterprise Agent Platform, the two companies aim to help telecom operators reduce operational overhead, rapidly resolve network anomalies, and transition toward fully self-driving networks.
The partnership represents a significant evolution in how telecommunication systems are managed. Rather than relying on static scripts or simple, rule-based automation, the new integration introduces a fleet of specialized cognitive agents that can independently analyze raw data, distinguish critical infrastructure failures from background noise, and coordinate complex troubleshooting workflows. Initial estimates from Nokia suggest that this agentic framework can reduce network problem-solving times by 50% to 80%, transforming complex issues that once required hours of manual investigation into automated fixes that take only a few minutes.
The Telecom Bottleneck: Why Manual Troubleshooting Is No Longer Sustainable
To understand the value of this partnership, one must examine the severe operational challenges facing modern communication service providers. Traditional network management models were built for simpler architectures, where human engineers could monitor physical switches, track static performance indicators, and manually intervene when a hardware component failed. However, the transition to software-defined networks, virtualized packet cores, and dense 5G cell deployments has created an environment of unprecedented complexity.
Today, a single regional network can generate millions of alarms, alerts, and log entries every single hour. This massive wave of data has triggered severe “alarm fatigue” among network operations center teams. Engineers are forced to spend a significant portion of their shifts manually triaging alerts, cross-referencing performance charts, and trying to determine which warnings indicate a genuine infrastructure failure and which are simply background noise.
Escalating Alerts and the Threat of Costly Downtime
When a genuine network anomaly occurs, such as a localized fiber cut, a software bug in a core router, or a sudden spike in traffic that degrades service quality, the clock starts ticking. Every minute of service degradation or network downtime can carry massive financial consequences. Telcos face immediate revenue losses, potential penalties for violating strict enterprise service-level agreements, and long-term damage to their brand reputation.
Despite these high stakes, manual troubleshooting is slow. An engineer must typically log into multiple separate systems, analyze historical performance trends, form a hypothesis about the root cause, and manually execute a series of diagnostic tests. By the time a human operator identifies the problem and formulates a fix, hours may have passed. As networks continue to scale to support millions of connected Internet of Things (IoT) devices and industrial automation systems, this slow, manual approach is no longer sustainable.
Transitioning from Traditional Automation to True Autonomy
To speed up operations, telecom companies have long utilized basic automation scripts. For example, if a specific interface reports a high error rate, a script might automatically restart the port. While useful, this traditional automation is highly limited. It relies on rigid “if-this-then-that” rules created by human programmers. If a network problem does not match a pre-defined signature, the automation fails, and the issue must be escalated to a human expert.
The integration of Google’s Gemini models into Nokia’s network software aims to bridge this gap. By utilizing large language models that possess advanced logical reasoning capabilities, the software can analyze unfamiliar scenarios, interpret messy, unstructured log files, and formulate creative troubleshooting strategies. This represents a transition from simple automation—doing a pre-programmed task—to true autonomy, where the software can evaluate a situation, make a logical decision, and execute a multi-step solution.
Inside the Multi-Agent Ecosystem Powered by Google Gemini
The centerpiece of the Nokia and Google Cloud partnership is the development of six specialized AI agents. Built using Google Cloud’s Agent Developer Kit on the Gemini Enterprise Agent Platform, these agents are designed to handle specific operational tasks within the Nokia Assurance Center. Rather than relying on a single, monolithic model to manage the entire network, Nokia has adopted a modular, multi-agent architecture where individual agents act as specialized domain experts.
These agents can work independently to resolve simple issues or communicate and collaborate with each other to solve highly complex, multi-layered network failures. This cooperative approach allows the system to scale its intelligence dynamically based on the complexity of the incoming network alerts.
The Six Specialized AI Agents and Their Roles
To deliver comprehensive network oversight, Nokia is introducing six distinct digital workers, each engineered with specialized skills:
- Router Agent: This agent serves as the central orchestration and communication layer of the entire system. It interprets natural language commands and user intent from human operators, translating those requests into specific tasks. The Router Agent then coordinates communication among the other specialized agents, delegating tasks and ensuring that all actions comply with established network policies and security protocols.
- Event Triage Agent: Operating at the front lines of the network operations center, this agent continuously monitors the incoming stream of alarms. It uses pattern recognition to group related alerts, filters out redundant notifications, and distinguishes critical hardware and software failures from routine background noise, ensuring that engineers are only alerted to genuine problems.
- KPI Selector Agent: Key Performance Indicators (KPIs) are the vital signs of a telecom network. The KPI Selector Agent acts as a domain performance expert, identifying and selecting the precise metrics—such as latency, packet loss, or signal-to-noise ratios—that need to be analyzed to diagnose a specific issue.
- Anomaly Reasoner Agent: Once the relevant metrics are selected, the Anomaly Reasoner Agent evaluates the data to identify deviations from normal baseline operations. It uses advanced statistical reasoning to determine if a performance dip is a temporary, normal fluctuation or a symptom of an underlying network failure.
- Action Reasoner Agent: After an anomaly is confirmed and diagnosed, the Action Reasoner Agent takes over to formulate a solution. It analyzes historical repair logs, product documentation, and network topologies to recommend a step-by-step remediation plan, such as rerouting traffic, adjusting antenna tilts, or restarting a software service.
- Dashboard Agent: To ensure full transparency, the Dashboard Agent instantly generates intuitive visual reports, summaries, and interactive dashboards. This allows human engineers to quickly review the actions taken by the AI agents and understand the underlying logic behind their diagnoses.
Glass Box Autonomy: Keeping Humans in the Loop
One of the most significant concerns surrounding the deployment of artificial intelligence in critical infrastructure is the risk of “black box” decision-making, where an AI system makes an opaque decision that human engineers cannot verify or stop. To address this risk, Nokia and Google Cloud have designed the system around a philosophy they call glass box autonomy.
Under this framework, the AI agents operate with complete transparency. Every step of their reasoning process—from the alarms they analyzed to the specific KPIs they selected—is logged and displayed in natural language for human review. Furthermore, the system is configured so that human engineers retain final approval over high-risk or critical actions, such as changing routing tables or deploying software patches to core infrastructure.
For low-risk, highly repetitive scenarios, operators can choose to enable fully automated, closed-loop responses, allowing the system to resolve routine issues instantly without human intervention. This balanced approach allows telecom providers to maximize operational speed while maintaining strict safety and regulatory control over their infrastructure.
Standard Infrastructure and the SaaS Rollout Roadmap
A common barrier to adopting advanced artificial intelligence in the enterprise sector is the high cost and complexity of deploying specialized computing infrastructure. Many legacy AI solutions require companies to build custom on-premise hardware clusters or integrate highly proprietary cloud services that lock them into a single vendor ecosystem.
Nokia and Google Cloud have addressed this barrier by ensuring that the Gemini-powered agents run on standard, open cloud infrastructure. The agents are designed to run seamlessly on Google Cloud’s standard managed Kubernetes and Google Cloud Storage services. By avoiding the need for custom, proprietary managed services, telecom operators can deploy the new agents quickly within their existing cloud environments, lowering the total cost of ownership and accelerating their time to value.
The Phased Deployment and Marketplace Launch
The rollout of the new agentic platform will follow a structured, phased roadmap to ensure technical stability and allow operators to gain confidence in the system’s capabilities.
- Initial Testing: The router and event triage agents are already operational and undergoing active testing with select telecom partners.
- Live Demonstrations: Nokia and Google Cloud are showcasing live, interactive demonstrations of the initial agents at the DTW Ignite event in Copenhagen, allowing industry representatives to observe the system’s troubleshooting capabilities firsthand.
- SaaS Launch: Nokia plans to officially launch the full agentic platform as a software-as-a-service (SaaS) product on the Google Cloud Marketplace in September 2026. At that point, telecom operators worldwide will be able to easily purchase and deploy the router and event triage agents directly through their Google Cloud accounts.
- Rolling Updates: The remaining four specialized agents—including the anomaly and action reasoners—will be delivered to customers through a series of rolling software updates starting in late 2026 and continuing throughout 2027.
The Network as Code Platform and Broad Market Implications
The integration of Gemini AI into the Nokia Assurance Center is closely linked to Nokia’s broader strategy to monetize modern 5G networks through software development. In late 2023, Nokia launched its Network as Code (NaC) platform, a developer portal designed to expose complex mobile network capabilities through standardized, easy-to-use application programming interfaces (APIs).
Traditionally, software developers who wanted to build applications that utilized advanced network features—such as high-precision location tracking, on-demand network slicing, or dynamic quality of service adjustments—had to possess specialized telecom-specific engineering knowledge. The NaC platform removes this barrier, allowing mainstream software developers to interact with the mobile network using the same API-based workflows they use for standard cloud applications. Today, the NaC platform connects over 70 ecosystem partners and utilizes more than 20 standardized APIs, including collaborations with major global operators like Deutsche Telekom and Vodafone.
Simplifying the Developer Experience with Intent-Based Workflows
By combining the Network as Code platform with Google’s Gemini models and the Model Context Protocol (MCP), Nokia and Google Cloud are taking this developer simplification a step further. Instead of requiring developers to manually write complex code to call different telecom APIs, the AI agents can interpret a developer’s broad intent and automatically generate the necessary on-demand workflows.
For example, a logistics company building an autonomous drone delivery application could simply tell the system: “Ensure high-priority, low-latency connectivity for our delivery drone fleet as they fly through this specific geographic corridor.” The AI agents would process this natural language request, communicate with the underlying network APIs, allocate a dedicated network slice with the required quality of service, and monitor the drone’s connectivity in real-time. This seamless integration transforms the mobile network from a passive transmission pipe into an active, programmable, and AI-native platform.
Lowering Total Cost of Ownership and Unlocking Revenue
Ultimately, the partnership between Nokia and Google Cloud addresses the core financial challenge facing modern telecommunication companies: how to lower the cost of network operations while unlocking new, high-margin revenue streams. By automating the vast majority of routine troubleshooting and performance-monitoring tasks, the Gemini-powered agents can help operators drastically lower their day-to-day OpEx.
Relieved of the burden of manual maintenance and alarm fatigue, telecom engineering teams can redirect their valuable time and resources toward developing innovative new services. Operators can focus on creating monetizable, on-demand network APIs for enterprise clients—such as high-security network slices for healthcare providers, location verification systems for financial institutions, or dynamic priority routing for autonomous vehicle fleets. As the telecommunications industry enters this new, agentic era of connectivity, the integration of advanced artificial intelligence into core network software will serve as the foundation for the next generation of digital services.
