Deploying Agentic AI in Reliability-Critical Grid Operations: A CAISO Case Study

Gopakumar Gopinathan – CAISO

This case study presents California ISO’s beta deployment of an agentic AI system designed to augment real-time outage management and reliability operations. Developed in collaboration with OATI, the platform integrates large language models with structured outage data, operating procedures, and market information to provide contextualized decision support. The system detects anomalies in the setup of planned and forced outages, synthesizes historical patterns, and generates operator-facing insights within a governed, human-in-the-loop framework.
We will discuss architectural considerations for reliability-critical AI, including data provenance, model guardrails, explainability, and regulatory alignment. Early results and operational lessons will be shared, along with insights into scaling AI within an ISO environment. The session will also explore ongoing research into the use of time-series foundation models for forecasting balancing authority capacity margins across the RC footprint.