Skip to main content
An answer can combine a written explanation with charts, tables, headline metrics, and generated artifacts. Start with the conclusion, then inspect the evidence that matters to your decision.

Before you begin

Ask the Analyst a question using an agent with access to the relevant data. Include any specific definition or comparison you have in mind.

Read the answer

  1. Check the scope: the date range, filters, and groups included.
  2. Check the meaning: what counts as revenue, an active customer, or another business metric.
  3. Open the supporting chart or table to examine the numbers.
  4. Look for assumptions and missing information before acting on the result.
A saved Analyst answer describing a revenue dashboard, explaining its data definitions and assumptions, and linking to the dashboard preview

Check the explanation and its assumptions, then open the artifact card to inspect the result.

Ask follow-ups in the same chat: “Does revenue include refunds?” or “Show the same comparison for returning customers only.”

Choose the right depth

The current modes are Flash, Pro, and Max. Flash suits everyday questions; Pro supports deeper analysis; Max plans before investigating complex tasks. Older guides refer to Concise and Deep Dive. Request the detail you need in any mode: “Give me three key findings” or “Explain the drivers and show the supporting breakdowns.” A shorter explanation still needs the right data and definitions.

Work with the result

  • Open a visual or artifact for a closer look.
  • Use a widget’s ⋯ menu for actions such as downloading an image or exporting a table as CSV.
  • Use Pin to reuse a widget in a dashboard or Datasite.
  • Ask for a change in the same chat: “Use a line chart” or “Add a breakdown by channel.”

Correct a result

Use Dislike on the affected widget and explain what was wrong when prompted—for example, “Exclude cancelled orders.” Use Like for a useful result. If the issue is a business definition everyone should use, update the agent’s rules, knowledge, or verified queries to provide shared context for future questions.

Next steps

You should now know what the answer covers and which data supports it. Improve your next prompt, create a report, or save the work in a project.
Last modified on October 2, 2026