Supports Cisco Live DEVNET-3171 session: Fault Intelligence as Code: AI Agents, RAG, and MCP for Network Ops
Jason Shoemaker, CCIE #24255
AI & Automation Architect
Customer Experience Delivery
Cisco Systems, Inc.
Steve Holl, CCIE #22739
Principal Architect
Customer Experience Product Management
Cisco Systems, Inc.
Tuesday, Jun 2
12:30 PM - 1:15 PM PDT
WoS - DevNet Theater
https://www.ciscolive.com/global/learn/session-catalog.html?search=DEVNET-3171#/
As networks scale, the knowledge required to detect, diagnose, and repair incidents becomes scattered across support cases, vendor advisories, telemetry analysis, and tribal expertise. This session shows how agentic AI can transform that fragmented knowledge into actionable fault intelligence that your tools can consume via simple data models and APIs.
- You’ll learn how AI agents, custom knowledge bases, and context-engineering techniques translate historical incidents and vendor recommendations into precise detection logic and automated repair action workflows.
- We’ll walk through an architecture and schema for representing “fault intelligence as code,” and how to keep agents grounded using RAG and Model Context Protocol (MCP).
- Through practical examples and sample code, you’ll learn how to build a lightweight AI agent that constructs its own fault-intelligence knowledge base and emits machine-ready rules and repair workflows.
- You’ll also see how these patterns plug into existing observability, ticketing, or automation tooling, and how Cisco CX is using the same ideas to continuously deliver fault intelligence that reduces MTTR and support escalations.