Cisco Extends AgenticOps to Air-Gapped Industrial Networks—AI Agents Now Run Factory Floors and Power Grids

At Cisco Live EMEA 2026, Cisco announced extensions to its AgenticOps framework targeting air-gapped and limited-connectivity industrial environments. The announcement addresses the largest gap in enterprise AI deployment: critical infrastructure networks that cannot connect to the public cloud for security, regulatory, or operational reasons. AI agents can now operate on factory floors, power grids, water treatment facilities, and defense installations using on-premises inference and local decision loops.

Why Air-Gapped Matters

Most enterprise AI assumes cloud connectivity. Models run in AWS, Azure, or Google Cloud. Data flows up, inference flows down. But large segments of industrial infrastructure operate air-gapped by design. Manufacturing facilities handling classified defense contracts cannot send telemetry to public clouds. Power grid control systems are isolated per NERC CIP requirements. Water treatment SCADA networks are segmented from IT networks by federal mandate. These environments have been largely excluded from the AI agent revolution.

What Cisco Is Shipping

AgenticOps for industrial environments runs on Cisco's industrial compute platforms, including the Catalyst IR series and industrial network controllers. The system supports local model inference using compressed versions of large language models and specialized industrial AI models. Agents can monitor OT network health, detect anomalies in SCADA traffic, automate configuration changes on industrial switches and routers, and coordinate with Cisco's broader security fabric through one-way data diodes where connectivity permits.

The OT Security Angle

Cisco is positioning this as both an operational efficiency play and a security play. Industrial networks face increasing threats from nation-state actors targeting critical infrastructure. The 2025 Dragos Year in Review documented a 67 percent increase in OT-targeted malware families. By running AI agents locally on the OT network, Cisco enables continuous monitoring and automated response without requiring the OT network to connect to IT infrastructure or the internet. This addresses the fundamental tension in OT security: you need intelligence to defend the network, but connecting the network to get that intelligence creates the attack surface you are trying to protect.

The Competitive Landscape

Cisco is not alone in targeting industrial AI. Siemens Industrial Copilot, Honeywell Forge, Rockwell Automation FactoryTalk, and ABB Ability all have industrial AI strategies. But Cisco's networking heritage gives it a specific advantage in the infrastructure layer. If you already run Cisco switches, routers, and firewalls on the factory floor, adding AI agent capabilities to that existing infrastructure is an extension, not a new deployment. The bundling playbook applies here as well: platform vendors absorb adjacent capabilities.

What Enterprise Buyers Should Consider

If you operate air-gapped or limited-connectivity industrial environments and have been waiting for AI capabilities that do not require cloud connectivity, this is the first serious enterprise offering in that space. The evaluation should focus on three factors. First, whether your existing OT network infrastructure supports the compute requirements for local inference. Second, whether the agent capabilities map to your specific operational use cases, not just generic monitoring. Third, how the governance model works when AI agents are making autonomous decisions on networks controlling physical processes where errors have safety consequences.