In modern cloud environments, log volumes are exploding, and with them, the challenge of spotting anomalies that matter. Traditional alerting often leads to fatigue, while real threats get buried. In this session, I’ll demonstrate how I built a multi-agent workflow using LangGraph to automatically detect, analyze, and respond to anomalies in Datadog logs.

The solution uses AI agents working together:

One agent extracts and parses log events.

Another applies anomaly detection (Isolation Forest) to identify outliers.

A third agent contextualizes the anomalies, grouping them by error patterns and impacted remote IDs.

A final agent pushes insights into downstream workflows (alerts via Webex, ticket creation in Jira).

By orchestrating these specialized agents, the system transforms Datadog from a passive log collector into an active anomaly detection and response engine.

Attendees will leave with practical insights into:

How to design and orchestrate multi-agent systems with LangGraph

Applying ML-based anomaly detection on real-world logs

Integrating detection with collaboration and ticketing workflows

This session is designed to be hands-on, showing how to combine AI, automation, and observability to move from reactive monitoring to proactive resilience.

Audience Takeaways

Understand multi-agent orchestration with LangGraph

Learn how to apply anomaly detection techniques to logs

See how AI can automate noisy monitoring workflows

Get inspired to extend existing observability platforms with AI