What happens when your production environment starts generating hundreds of alerts at 3 AM?
Payment gateway latency. JVM memory pressure. Kubernetes pod restarts. Disk alarms. Application errors.
The challenge isn't simply detecting these alerts. The real challenge is understanding how they are related, identifying the underlying root cause, and deciding what action can safely be taken.
In this video, I walk through an architecture for combining AIOps, Observability, and Agentic AI to address this problem.
The architecture covers:
- OpenTelemetry for collecting application and infrastructure telemetry
- Kafka / Strimzi as the event streaming layer
- BigPanda for event normalization, correlation, and deduplication
- ServiceNow for incident management
- Agentic AI for investigation and Root Cause Analysis (RCA)
- Tool-based agents for gathering operational evidence
- Dependency and contextual analysis
- Human-in-the-loop approval
- Controlled and auditable remediation
The important architectural principle is that the LLM should not simply be given access to production systems and told to "fix the problem."
Instead, the AI operates through controlled tools, gathers evidence from multiple sources, reasons over the available context, and follows defined safety boundaries before any remediation action is performed.
The video walks through the architecture layer by layer and explores what it takes to move from traditional monitoring and alerting toward AI-assisted investigation and safe remediation.
🎥 Watch the architecture deep dive:
This is aimed at Solution Architects, AI Architects, AIOps/SRE engineers, DevOps engineers, and anyone exploring Agentic AI for enterprise operations.
Top comments (0)