Intelligent Network Summit

Five minutes. That is all the downtime a 99.999% availability target allows each year. One misconfigured BGP route, a failed cloud interconnect or an incomplete firmware rollout can exhaust that margin in seconds. Hardware redundancy cannot guarantee uptime on its own. Sustained availability depends on how quickly teams identify issues, implement changes and confirm results across complex hybrid infrastructure. AI agents offer a path forward, but only if they act on actual network state, not assumptions.

This session presents a practical, architecture-driven framework for integrating reasoning agents into enterprise network operations. Eugina Jordan will outline five actionable steps to embed these agents directly into operational pipelines.
Through real-world case studies, including a multi-site SD-WAN degradation and a failed cloud edge migration, the session will detail system blueprints for building agents that diagnose faults, validate fixes and execute changes with human oversight. The focus is on maintaining production stability while accelerating response and resolution.

Key Takeaways
- Connect agent reasoning to live telemetry, topology data, and controller interfaces using APIs, NETCONF, and RESTCONF. Build in policy checks and require human approval for critical actions.
- Ground agent diagnostics in live RIB and FIB tables, firewall policies, and synthetic probe data to avoid unsupported conclusions and confirm proposed fixes.
- Apply execution patterns that combine scoped permissions, canary rollouts, and state-aware rollback to safeguard application performance and availability during changes.
- Use an operational scorecard to measure MTTR, change failure rates, and maintenance downtime before scaling AI-assisted network automation.