AWS re:Invent 2025 - AIOps Revolution: How iHeart slashed incident response time by 60% with Bedrock

Transforming IT Incident Response with Agentic AI: iHeart's Journey

The Challenge of Modern IT Incidents

• Large digital organizations like iHeart Media face complex, distributed IT systems that make it difficult to quickly identify and resolve incidents
• Incident response often involves a "seven circles of on-call hell" - logging in, hunting for information, relying on tribal knowledge, manual diagnosis, and more - wasting precious time
• Traditional monitoring systems generate too much noise, making it hard to pinpoint root causes

iHeart Media's Scale and Operations

• iHeart Media is a massive media company with:
  • 850+ AM/FM radio stations
  • 250 million monthly digital users
  • 150 million monthly podcast downloads
  • 5-7 billion monthly digital requests
  • 70+ AWS services powering its architecture
• The company's digital platform is mission-critical, requiring 24/7 uptime and fast incident resolution to avoid major business impacts

Introducing Agentic AI for IT Operations (AIOps)

• iHeart built a multi-agent AI system to automate incident response and remediation
• Key components:
  • Slack bot interface for human interaction
  • Orchestrator agent to delegate tasks to specialized sub-agents
  • Sub-agents for monitoring, logs, Kubernetes, knowledge base, etc.
  • Leveraging AWS Bedrock Agent Core for secure, scalable agent deployment
• Agents work together to quickly triage incidents, diagnose root causes, and recommend remediation steps

Benefits of the Agentic AI Approach

• 60% reduction in incident response time by automating triage and diagnosis
• Improved operational efficiency and reduced toil for on-call engineers
• Increased consistency and reliability in incident response
• Preservation of institutional knowledge for faster future incident resolution

Implementing the Agentic AI Solution

• Slack bot interface allows simple, natural language interaction to trigger incident investigation
• Orchestrator agent delegates tasks to specialized sub-agents, each with their own context window
  • Prevents sub-agents from overloading the main context with unnecessary data
  • Allows parallel, targeted investigations across monitoring, logs, Kubernetes, etc.
• Bedrock Agent Core provides a secure, scalable runtime to deploy and manage the multi-agent system

Lessons and Next Steps

• Quality of context data is critical - "garbage in, garbage out" for AI agents
• Gradual adoption approach: start with read-only, low-risk tasks before expanding to high-stakes actions
• Build a robust evaluation environment to continuously test and validate agent performance
• Future goals include expanding agent capabilities, integrating more data sources, and enabling proactive incident prevention

Key Takeaways

• Agentic AI can revolutionize IT incident response by automating triage, diagnosis, and remediation
• iHeart achieved 60% faster incident resolution, reduced toil, and preserved institutional knowledge
• Implementing a multi-agent architecture with specialized sub-agents is key to managing complexity
• Careful attention to context data quality and gradual adoption are critical for success
• The agentic AI revolution is here, and AWS Bedrock provides the tools to join it