# An Agentic Control Plane Enables AI-Driven Self-Healing in Kong Connect

Gan from Kong’s AI platform team explains how an agentic control plane can make infrastructure more self-healing by solving the investigation problem that slows incident response. The core argument is that traditional monitoring tools are good at detecting symptoms through alerts, autoscaling, and health checks, but they fragment telemetry across logs, traces, metrics, and dashboards and rarely identify the causal chain behind an outage. The proposed approach uses existing data already flowing through Kong Connect’s control plane and data plane, plus an agentic cognitive overlay that reasons out of the request path, synthesizes live telemetry, diagnoses root causes, proposes or executes fixes, verifies outcomes, and rolls back if needed. Gan outlines a six-step closed loop of sensing, diagnosis, decision, action, monitoring, and long-term learning, along with strict safety controls such as policy guardrails, blast-radius limits, human approval for risky changes, audit trails, canary rollouts, and dead-man-switch rollbacks, with adoption progressing from copilot mode to autonomous self-healing.