AWS re:Invent 2025-AI-Powered Observability & Observability for AI: The Two Sides of the Coin-AIM206

AI-Powered Observability & Observability for AI: The Two Sides of the Coin

Introduction

• The speaker, Anand, is part of the leadership at Zoho Corporation's ManageEngine division.
• The presentation focuses on the two sides of AI in the context of modern digital business operations: AI-powered observability and observability for AI.
• The key question posed is: "If AI is the watcher, who is really watching AI?"

The Changing Landscape of Monitoring and Observability

• In the past, monitoring was relatively simple, with monolithic applications deployed in data centers.
• However, the modern digital landscape has become increasingly complex:
  • Multiple cloud providers and regions
  • Containerization with thousands of containers spawning up and down
  • Interconnected APIs calling other APIs in complex chains
  • Impatient users who abandon slow applications
• This complex "jungle" of systems requires a new approach to observability.

AI as a Co-Pilot for Observability

• Traditional monitoring approaches using dashboards, alerts, and workflows are no longer sufficient to navigate this complexity.
• AI can act as a "co-pilot" in the observability ecosystem, helping to:
  • Scan and correlate millions of events across networks, user experience, servers, and applications
  • Cluster signals and connect the dots across the application ecosystem
  • Predict potential issues before they occur

AI-Powered Observability in Action

• Example: Handling a sudden spike in traffic and checkout issues during a promotional campaign
  • Traditional monitoring would show spikes in various metrics, but the operations team may not be able to quickly identify and resolve the root cause.
  • AI-powered observability can seamlessly correlate events, provide efficient root cause analysis, and suggest capacity adjustments to handle the increased load.

The Flip Side: Observability for AI

• While AI can be a powerful tool for observability, it also introduces new challenges:
  • AI can fail quietly without obvious warning signs, and it often does so with a high degree of confidence.
  • Example: A weather app providing inaccurate forecasts due to issues with the underlying API or model updates.
• Observability for AI is crucial to ensure the reliability and accountability of AI-powered systems.

Measuring What Matters

• Key metrics for observability of AI-powered systems:
  • Drift, prediction rate, confidence levels, and cost
• These metrics help ensure the trustworthiness, reliability, and sustainability of AI-powered decision-making.

The Future of Observability

• Observability is no longer just about infrastructure and applications; it must also encompass AI as a unifying element.
• The future of observability is about humans and AI working together as active enablers, with AI helping to measure and address what's truly important.

Conclusion

• ManageEngine's approach to observability follows a holistic, integrated approach, leveraging AI to correlate signals across the entire IT ecosystem.
• The key is to embrace both sides of the AI observability coin - the power of AI-powered observability and the need for observability of AI itself.