Cisco Expands AgenticOps for Autonomous Network Troubleshooting
Cisco is expanding its AgenticOps capabilities across its portfolio to enable autonomous troubleshooting in campus, industrial, and data center environments. The initiative represents a shift from AI assistants to self-resolving agents that can manage network issues without human intervention. This move aims to increase network reliability and reduce operational overhead.
- The core of AgenticOps is Cisco's Deep Network Model, a specialized LLM trained on over 40 million tokens from the company's 40 years of operational data, including Cisco U courseware and CCIE-level knowledge. This purpose-built model demonstrates up to 20% greater accuracy in troubleshooting and configuration tasks compared to general-purpose LLMs. - To handle the massive scale of network telemetry, Cisco developed two key innovations: Analytics Context Engineering (ACE) to optimize context while reducing prompt size, and Lightweight Autonomous Program Synthesis and Execution (LAPSE) to manage and execute tasks based on machine data. - A central user interface for this technology is the AI Canvas, a generative and collaborative workspace where NetOps, SecOps, and DevOps teams can interact with the AI Assistant. This interface dynamically generates visualizations and provides a unified view by integrating data from tools like Meraki, ThousandEyes, and Splunk. - Practical use cases include autonomous troubleshooting for reducing mean-time-to-resolution, agentic workflow creation for production-ready automations, and proactive analysis of firewall traffic to recommend more robust zero-trust controls. - The underlying infrastructure for these AI capabilities is also a focus, with Cisco investing in its Silicon One architecture, including the new G300 chip, to power high-performance AI networking. - In the competitive landscape, while other networking vendors like Juniper are also leveraging AI with their Mist AI and Marvis assistant, Cisco's strategy centers on a domain-specific LLM and an agent-driven operational model across its entire portfolio. - According to Jeetu Patel, Cisco's President and Chief Product Officer, the goal is to address the constraints of the agentic AI era, where "agents are going to be working 7 by 24 autonomously," by providing the necessary infrastructure and building trust through security. - For observability, Cisco is integrating AI Agent Monitoring within the Splunk Observability Cloud, allowing teams to track the performance, cost, and behavior of the agentic workflows in real-world environments.
