From Alert Fatigue to Self Healing Building AI Enabled Control Planes in Banking Yury Tsarev, Nuno

When Millennium bcp, one of Portugal’s largest banks, faced alert fatigue and long MTTR across their multi-cloud platform, we built AI-enhanced Crossplane control planes to bring self-healing and intelligent scaling to production. Using LLM-powered composition functions, Kubernetes alerts are automatically triaged and remediated, cutting SRE escalations dramatically. Workload-aware algorithms dynamically scale resources across multiple clouds—remaining fully auditable and compliant. We’ll share the architecture, open-source components, and lessons learned from running AI-enabled control planes in a regulated banking environment—showing how to adopt AI Ops responsibly using CNCF projects like Crossplane and Kubernetes.