How to Ask Questions About Your Data with Copilot in Power BI

João Barros 06 de July de 2026 4 min read

Asking questions of your data in natural language is one of the fastest ways to get value from a Power BI report. With Copilot in Power BI, you type your question as if you were talking to a colleague and get back an answer with a visual and references to the report pages, without needing to know DAX. This guide shows you, step by step, how to open Copilot, ask good questions, and confirm that the answer is correct before you use it.

Prerequisites

A paid Microsoft Fabric capacity (F2 or higher) or Power BI Premium (P1 or higher). Since April 2025 you no longer need a large capacity: any paid SKU from F2 upward works.
Copilot enabled by your administrator in the Fabric admin portal.
A Power BI Pro (or PPU) license to work in the service; the license alone does not replace the capacity.
A report already published to a workspace on that capacity.
A well-prepared semantic model: tables and columns with clear names and correct relationships.

Step 1: Open the report and the Copilot pane

Start by opening the report in the Power BI service, that is, Power BI in the browser (not the Power BI Desktop application). With the report open, find the Copilot button in the top bar and click it. A chat pane opens on the right side of the screen: this is where you will type your questions. If you do not see the button, it is almost always because Copilot is not enabled or the workspace is not on a paid capacity; in that case, talk to your Fabric administrator.

Step 2: Ask your first question about the data

Type a specific question about data that exists in the report model. The more specific the question, and the more you use the real field names, the better the answer. A good starting point is something like:

What was the total sales by region last quarter?

Copilot reads the question, queries the semantic model behind the report, and returns the answer as a visual (for example, a bar chart) instead of forcing you to build it field by field. Note that it answers based only on your model's data, and not on outside knowledge.

Step 3: Refine with follow-up questions

You rarely get everything right on the first try, and that is fine. Copilot keeps the conversation context, so you can adjust your request with a follow-up question instead of starting over. For example:

Now show only the five regions with the highest growth versus the previous year.

You can also ask for a different format: "show as a table", "use a line chart", or "summarize in bullet points". Breaking a complex question into several short steps usually gives more reliable answers than one very long question.

Step 4: Write questions Copilot understands well

The quality of the answer depends a lot on how you ask. A few simple good practices:

Use table and column names exactly as they appear in the report (for example "Gross Margin" rather than "profit").
State the period and the level of detail: "by month", "in 2025", "by customer".
Avoid ambiguous questions; if "sales" could mean several things, clarify which one.
Ask one thing at a time and build on the previous answer.

Step 5: Validate the answer through its references

Every Copilot answer includes references to the visuals or report pages that were used to build it. Click those references to see exactly where the numbers come from. This step is essential: Copilot is an assistant that speeds up your work, not an absolute source of truth. Always confirm that the value matches what is in the report before you share it in a meeting or an email.

Verify the result

You know it went well when three signs come together: the answer includes a visual that matches the question, the references point to real report data, and the numbers agree with what you already knew. If the answer comes back empty or strange, the problem is almost always in the model (ambiguous names, poorly identified columns, or missing relationships) and not in the question itself.

Conclusion

Asking questions of your data with Copilot in Power BI saves time and opens analysis to people who do not master DAX or visual building. The next step, to improve the answers even further, is to look after the semantic model: give clear names and add descriptions and synonyms to tables and columns so that Copilot better understands your business vocabulary. What will be the first question you ask your data?