Cisco builds AI model to automate network operations

Cisco is preparing to release a proprietary AI model trained specifically for networking tasks, claiming it can diagnose and resolve network issues more effectively than general-purpose frontier AI models, as the company expands its push into AI-driven network operations.

The networking giant said the model, which understands routing and switching configurations, will be published on Hugging Face in the coming weeks alongside its previously released Antares security models, according to the Register.

Cisco plans to use the model to power Cloud Control, a new agentic AI operations platform that provides a unified interface for managing networking, security and infrastructure environments.

Cloud Control is designed to reduce the need for administrators to switch between multiple management consoles by allowing them to query infrastructure using natural language. The platform employs AI agents to collect telemetry from different Cisco products, analyse operational data and identify the root cause of network issues.

According to Cisco, the platform can investigate incidents such as a device failing to connect to a Wi-Fi network by correlating information from wireless infrastructure, security policies and network configurations. Where corrective actions such as device reboots are required, the platform can also assess the operational impact before recommending changes.

Cisco expects Cloud Control to become available to customers in the United States by the end of August, with rollouts across Europe and the Asia-Pacific region to follow. The company said deployment timelines outside the U.S. are being influenced by infrastructure readiness as well as regional data governance and regulatory requirements, since the service requires access to customer environments.

The platform will be bundled with Cisco's existing products, although the company has yet to finalise its pricing strategy. Like other generative AI services, Cloud Control will operate on an AI token-based consumption model, with customers receiving a bundled allocation of tokens under existing subscriptions and paying for additional usage as required.

Recognising that some organisations may prefer not to rely on cloud-hosted AI services, Cisco also plans to make its networking AI models available for on-premises deployment, including on Cisco UCS servers. The approach is expected to appeal to enterprises with strict data residency, compliance or security requirements.

The announcement reflects a broader industry trend in which infrastructure vendors are developing domain-specific AI models rather than relying solely on general-purpose large language models. By combining specialised networking models with agentic AI workflows, Cisco aims to automate complex operational tasks while improving the accuracy of network diagnostics and reducing the time required to identify and resolve infrastructure issues.

With enterprises increasingly looking to simplify hybrid IT operations, Cisco is betting that specialised AI models and agentic automation will become a key differentiator in network management, even as traditional product-specific management consoles continue to coexist alongside AI-driven operational platforms.