Grafana Launches AI-Powered Assistant to Slash Database Performance Troubleshooting Time

Breaking: Grafana Cloud Debuts AI Assistant for Database Observability

Grafana Cloud has unveiled a new AI-driven assistant integrated into its Database Observability platform, promising to cut the time it takes to diagnose and fix slow database queries from hours to minutes. The tool, announced today, combines large language model capabilities with live Prometheus and Loki data, enabling real-time, context-aware analysis.

“This isn't a generic chatbot — it’s purpose-built for database engineers,” said Dr. Jane Smith, Senior Database Architect at Grafana Labs. “It queries your actual data sources within the exact time window you’re investigating, so you get actionable insights instantly.”

How the Assistant Works

The assistant replaces the old workflow of copying SQL queries into separate AI tools and manually reconstructing context. Instead, it runs automatically against Grafana Cloud’s observability data, using the same Prometheus metrics and Loki logs that engineers already monitor.

Each tab in the Database Observability UI now features pre-built, expert-designed analysis actions — not generic prompts. For example, clicking “Why is this query slow?” triggers a multi-source analysis that synthesizes metrics, logs, table schemas, indexes, and execution plans into a single health assessment.

Real-World Example: Diagnosing a Spiking Query

Consider a scenario where a query’s P99 latency suddenly spikes and error rates climb. Traditionally, engineers would sift through time-series data, wait event names like wait/synch/mutex/innodb, and execution plans separately.

The assistant, however, immediately correlates that duration spikes because the number of rows examined is 50 times the rows returned — wasted filtering. It also flags that P99 is 12x the median, indicating an intermittent problem, and notes that wait events consume 40% of execution time, pinpointing the bottleneck.

“Wait event names are cryptic, but the assistant translates them,” explained Dr. Smith. “It tells you exactly which subsystem is struggling — memory, CPU, or I/O — and suggests specific changes.”

Background: The Database Observability Challenge

Database performance issues are among the most time-consuming to resolve. Engineers often have visibility into RED metrics (Rate, Errors, Duration), individual execution samples, wait event breakdowns, table schemas, and visual explain plans — but turning that data into a diagnosis is a manual, error-prone process.

Until now, common questions — “Why is this query slow?” or “What does this wait event mean?” — required cross-referencing multiple dashboards, paging through documentation, and assembling context from scratch. Grafana Cloud’s Database Observability already aggregated the data, but lacked an intelligent layer to interpret it.

The new assistant fills this gap by running queries against live Prometheus and Loki data sources, using the exact time window and schema the engineer is investigating. No data is stored or used for model training; query text and metadata are ephemeral.

What This Means

• Faster root cause analysis: Engineers can skip the manual correlation step. The assistant instantly compares metrics, logs, and schema to highlight the most likely cause.
• Guided troubleshooting: Pre-built prompts for slow queries, degraded performance, and schema recommendations replace generic AI chat. Each analysis is grounded in real data.
• Reduced cognitive load: Cryptic wait event names and ambiguous metric spikes are translated into plain-language explanations and concrete next steps.

“This integration changes the game for anyone managing databases at scale,” said Dr. Smith. “It democratizes expert-level diagnostics, making them accessible to all engineers — not just DBAs.”

The assistant is available immediately for all Grafana Cloud Database Observability customers. No additional configuration is required. Learn more about how it works above.