SAN DIEGO–A wide range of new AI-based software and technology from Cisco is designed to manage, secure and automate network operations as well as to cement Cisco’s place at the center of customers’ AI infrastructure plans.
Cisco rolled out the new wares at its Cisco Live event, where it also unveiled new Nexus data center management software, campus Smart Switches, branch routers and a raft of security updates.
“The overarching message to enterprise customers is that we’re launching the ability to support the use of AI, and then we’re introducing AI infrastructure to support it and products to make it secure, and then you combine that with what we are doing with Nvidia, and that makes Cisco a full-stack AI player and gives customers the security and management tools to support it all,” DJ Sampath, senior vice president of Cisco’s AI software and platform group, told Network World. “We have a rapidly developing, intentional strategy for AI development.”
A key part of Cisco’s strategy is using its own trove of networking telemetry data gleaned from Cisco ThousandEyes, AppDynamics, and NetFlow to build a network-specific large language model (LLM) called the Deep Network Model. Education and learning materials from Cisco U. and Cisco Certified Internetwork Expert (CCIE) are also part of the Deep Network Model.
The Deep Network Model will be used by a variety of Cisco systems to help customers identify network issues and tackle vulnerability remediation more quickly. It will be a central piece of another AI-based product called AI Canvas, which was demonstrated at the event and will be available in October. AI Canvas will tie together networking, security, and observability data and help customers spot and fix all manner of trouble spots throughout the enterprise estate, Sampath said.
AI Canvas can also gather information from non-Cisco security systems, such as firewalls, according to Anurag Dhingra, senior vice president and general manager of Cisco’s enterprise connectivity and collaboration group. The data comes together in real-time dashboards to guide corrective actions.
“Team members can share dashboards with colleagues and save them to persist across sessions,” Dhingra wrote in a blog post. “AI Canvas uses advanced reasoning models to break down troubleshooting into structured steps. When teams agree on a proposed solution, AI Canvas takes action to fix the problem, from implementing a configuration change to executing steps in a runbook. This collaborative approach uses AI to bring NetOps, SecOps, and other teams together for faster problem resolution.”
It’s a difference maker, according to Neil Anderson, vice president of cloud, infra, and AI solutions with IT service and global systems integrator World Wide Technology.
“Finally, being able to work across all these different silos that an enterprise has had to work with separately for years, and easily mitigate network and security problems, is game changing for customers,” Anderson said. In addition, since the Deep Network Model is built on Cisco’s own knowledge base, the details it feeds AI Canvas will be much more accurate than the sorts of data available from public systems, Anderson said.
The Deep Network Model also feeds the new Cisco Live Protect system, which is a real-time, live patching system that applies security controls at the kernel level to defend against newly spotted vulnerabilities and exposures without requiring switch or router reboots. The Live Protect feature can be triggered or managed via Cisco’s AI assistants or Hypershield workflows.
Agentic AI
Cisco talked extensively at the event about the emerging world of agentic AI, where AI agents can share information, collaborate, reason, and take autonomous actions in real-time, Sampath said.
“Cisco’s vision integrates cutting-edge capabilities, including automated agent discovery, delegated authorization, secure zero-trust agentic access, and native support for the Model Context Protocol (MCP),” said Jeetu Patel, Cisco’s president and chief product officer (pictured above at Cisco Live keynote). MCP offers a standardized way for AI models to interact with external tools and services, such as code repositories, databases and web services.
Cisco’s agentic AI developments were influenced by the work its own advanced research group, Outshift, did in developing its “Internet of Agents,” plan, Sampath noted. The Internet of Agents proposes an open-sourced, three-layer architecture that would enable quantum-safe, agent-to-agent communication to let AI agents to collaborate autonomously and share complex reasoning.
In March, Outshift announced AGNTCY, an open-source collective building the critical infrastructure for AI agents to work together. Cisco, LangChain, and Galileo are the initial core maintainers, with Glean and LlamaIndex as contributors.
“As AI gets built into every application and service, organizations will find themselves managing hundreds or thousands of discrete agents. Without open standards and frameworks, this diversity creates chaos,” Vijoy Pandey, senior vice president of Outshift by Cisco, told Network World in January. “It’s like the early days of networking – we need common protocols and standards so these agents can discover, communicate, and collaborate with each other effectively. This standardization and interoperability will be essential for enterprises to effectively manage and scale their AI initiatives.”
Cisco will also use a number of products to manage AI agents, including Duo Identity & Access Management (IAM) to provide the authorization and Secure Access for semantic inspection so that the end user does not have to be prompted repeatedly for access permission. In addition, Cisco’s AI Defense is invoked to evaluate that agent actions align with its purpose, and Cisco Identity Intelligence monitors the actions and provides visibility, explained Raj Chopra, senior vice president and chief product officer for Cisco Security, in a blog post.