AWS re:Invent 2025 - Ops in the AI age: Innovating together for faster, more efficient operations

Innovating Together for Faster, More Efficient Cloud Operations

Embracing the Age of Agentic AI

• AI has disrupted customer experiences, with end-users now expecting faster and smarter experiences powered by AI
• AI adoption varies across industries, with 88% of organizations using AI in at least one business function
• Compliance and security are critical for the adoption of generative AI

Enhancing Fan Experiences with AI at PGA Tour

• PGA Tour is a massive, data-intensive operation with over 120 cameras and 36 radar trackers capturing 32,000+ shots per week
• They use AI-powered shot commentary to provide context and relevance to fans watching the live events
• This system is built on a cloud-based architecture with real-time data ingestion, analysis, and commentary generation
• Operational dashboards and monitoring provide visibility into the health and performance of the system

Addressing the Challenges of Agentic AI

1. Trust and Safety: Visibility into AI agent behavior is critical to understand how and why decisions are made
2. Operational Complexity: Managing microservices, distributed agents, and event-driven architectures across multiple accounts and regions
3. Data Explosion: AI agents handle thousands of customer requests per hour, generating exponential amounts of telemetry data to monitor and secure

Providing Observability for Agentic AI

• Generative AI Observability in CloudWatch provides visibility into AI agent behavior, including latency, token usage, and performance
• CloudWatch Application Map automatically discovers and organizes services, correlating metrics, logs, and traces to simplify root cause analysis
• Developers can use CloudWatch MCP Servers to troubleshoot AI agents directly from their preferred IDEs and productivity tools

Empowering Smarter Operations with AI

• CloudWatch Investigations uses generative AI to automate root cause analysis and provide troubleshooting guidance
• Interactive Incident Report Generation leverages the "Five Whys" framework to capture and share operational learnings
• Application Observability for AWS GitHub Actions brings observability directly into developers' workflows, enabling them to diagnose production issues using live telemetry

Simplifying Overall Cloud Operations

• Enhancements to CloudWatch Log Analytics, including natural language support and over 35 analytical commands, provide faster and more accessible insights
• Expanded CloudWatch Real User Monitoring for mobile apps enables end-to-end visibility, from user behavior to underlying infrastructure
• CloudTrail Aggregated Events simplify security monitoring by pre-aggregating high-volume API calls and identifying anomalies

Centralizing Observability Across Hybrid and Multi-Cloud Environments

• Cross-account, cross-region log centralization in CloudWatch consolidates data from development, QA, and production environments
• Database Insights expand the cross-account, cross-region capabilities to support monitoring and troubleshooting of RDS and Aurora databases

Delivering Personalized Ads at Scale with Warner Bros. Discovery

• Warner Bros. Discovery operates a massive ad platform across streaming services like HBO Max and Discovery+
• They use a custom metric called "permits" to autoscale their infrastructure based on factors like content duration, DVR window, and number of ads
• Predictive autoscaling, based on the rate of change in permits, helps them maintain a 90% ad fill rate with low latency

Key Takeaways

• AWS is committed to simplifying cloud operations for organizations, whether they are using agentic AI or traditional microservices
• Observability is the control plane for trust, safety, and accountability in the age of agentic AI
• AI-powered tools like CloudWatch Investigations and Incident Report Generation can automate root cause analysis and capture operational learnings
• Centralized observability across hybrid and multi-cloud environments is crucial for managing complex, data-intensive workloads
• Innovative approaches to autoscaling and infrastructure management can enable the delivery of personalized, low-latency experiences at scale