Copilot can now analyze individual queries

When investigating a cost spike or performance issue, the trail often leads to a specific query. Previously, Copilot could help you narrow down the problem — identifying the warehouse, workload, or time period responsible — but once you got to the query level, you were on your own.

Now, Copilot can analyze individual Snowflake queries end-to-end. Just ask Copilot to dig into a specific query and it will find and analyze it for you.

It returns a structured analysis covering:

• What the query does. An AI-generated plain-English summary so you can quickly understand unfamiliar SQL without reading through hundreds of lines.
• Why it's slow. If the query has performance bottlenecks — like excessive data spillage, exploding joins, or full column scans — Copilot identifies them and explains what's happening.
• What it costs. Copilot calculates the per-execution cost and projects annualized costs at different execution frequencies (hourly, daily, weekly), so you can understand the real impact of a recurring query.
• How to fix it. Copilot runs query-level insights that detect optimization opportunities — warehouse right-sizing, scan efficiency improvements, redundant operators, and more — and returns actionable recommendations sorted by effort.

If you already have a query ID — from a Slack alert, a monitor, or your own investigation — you can paste it directly and Copilot will analyze it immediately.

This turns a multi-step manual investigation into a single question. Instead of bouncing between the query profile, warehouse settings, and cost dashboards, you get the full picture in seconds.

Query-level insights

Supporting this feature, we've also refined the insights that power query analysis. Copilot now evaluates queries against 8 different optimization checks:

• Excessive data spillage
• Exploding joins
• Poor scan pruning efficiency
• Warehouse oversizing
• Sub-minute billing waste
• Redundant query operators
• Full column scans
• Inefficient Cortex function usage

These insights are surfaced directly in Copilot's query analysis, prioritized by implementation effort so you know where to start.

Other things we shipped

• 🚀 Copilot can now analyze point aggregates (full bar) for stacked/multi-series charts
• 🐛 Various bug fixes and improvements