"How Observability Combined with Generative AI Revolutionizes Performance and Insights"

Presented by Elastic

In our digital world, the reliable operation of key software systems and services is essential for business success.

Downtime or performance issues can lead to various negative outcomes, including lost revenue as potential customers turn to competitors, and decreased employee productivity when they cannot meet deadlines.

For site reliability engineers (SREs) and DevOps professionals, maintaining critical websites and applications can feel like an ongoing battle. However, there is promising news: Generative AI is here to enhance traditional observability methods, accelerating the resolution of reliability, security, and speed challenges.

The AI Advantage

Traditionally, monitoring and observability revolved around identifying signals amidst the noise and diagnosing unknown issues to enable swift remediation. Generative AI streamlines this process, allowing SREs and DevOps teams to respond to incidents with greater speed and confidence.

Consider a newly hired on-call engineer who lacks in-depth knowledge of the organization’s systems. If alerted in the middle of the night about an irregularity in a system they don't fully understand, they can converse with an AI assistant to quickly gather essential information. By asking questions like, “What is the purpose of this system?” or “Which other systems connect to it?” the engineer receives valuable context in seconds, thanks to the large language model (LLM) that powers the generative AI.

What’s particularly impressive is that the engineer interacts with the LLM using natural language; there’s no need to grasp complex query languages. This conversational approach allows them to quickly access the information required to troubleshoot effectively.

Empowering Collective Knowledge

Generative AI not only responds to queries but can proactively summarize relevant context for SREs. For instance, an engineer can receive a comprehensive issue summary in their Slack channel before being awakened by an alert. This includes all actions taken and involved parties, enabling immediate readiness to respond rather than wasting valuable time in catching up.

By providing a snapshot of the playbook used during similar past incidents, the LLM empowers the engineer to either execute it themselves or simply instruct the LLM to do so. This eliminates much of the guesswork and resolves potential issues efficiently, regardless of the engineer's experience level.

Companies like T-Mobile Netherlands are already harnessing this functionality, leveraging AI technology to support their network operations and ensuring improved network reliability and rapid issue resolution.

Looking Ahead

Currently, generative AI acts as an assistant that offers context and support, but its role is set to evolve. In the near future, generative AI could automate many responses on behalf of engineers. If an AI agent repeatedly recognizes a specific alert pattern, it could autonomously execute the appropriate playbook and confirm actions taken.

Moreover, combining observability data with other organizational systems—such as ERP and security—will allow engineers to pose more sophisticated, business-critical queries. They may transition from asking about past alerts to understanding the revenue impact of similar incidents or operational implications on the supply chain.

A Transformative Tool

While observability professionals have always had powerful tools at their disposal, generative AI introduces an innovative method to enhance their workflows. Importantly, it does not replace SREs or DevOps professionals; it alleviates the routine toil of their roles, freeing them to focus on higher-level problem-solving.

By facilitating access to relevant information, enhancing insights, and expediting decision-making, the integration of generative AI with observability data marks a significant breakthrough—truly a gamechanger.

Abhishek Singh is GM, Observability at Elastic.