What Power BI Copilot requires before you can use it
By SchemaChat (Data BI LLC) · Published 30 October 2026 · Last reviewed 30 October 2026 · 6 min read
According to Microsoft Learn (checked 30 October 2026), Copilot in Power BI needs a paid Fabric capacity of F2 or higher, or a Power BI Premium capacity of P1 or higher, plus an administrator who has enabled it and a supported region. A Power BI Pro or Premium Per User license on its own is not enough, and trial capacities are not supported.
Older articles and forum answers give different numbers (some say you need a much larger Fabric capacity). This page reports what Microsoft's own documentation said on the date above. Microsoft changes these requirements, so treat the Microsoft Learn page as the source of truth and re-check it before you buy.
The requirements at a glance
| Requirement | Details (Microsoft Learn, 30 Oct 2026) |
|---|---|
| Capacity | Paid Fabric capacity (F2 or higher) or Power BI Premium (P1 or higher). Trial capacities and free SKUs aren't supported. |
| Admin setting | “Users can use Copilot and other features powered by Azure OpenAI” must be on. It is on by default, but an admin might turn it off. |
| Region | The Fabric capacity must be in a supported region. Sovereign clouds aren't supported. |
| Licenses | A Power BI Pro or Premium Per User license alone is not sufficient; Copilot requires organizational capacity. |
What changes by experience
- The Copilot pane inside a report is described as generally available; the standalone full-screen Copilot and Copilot in apps are described as preview.
- The standalone Copilot needs access to a Copilot-supported capacity (a dedicated Fabric Copilot capacity, or a workspace on an F2-or-higher capacity in a supported region). Without one, you can't use it.
- In Power BI Desktop, you need write access to a workspace on a paid Fabric or Premium capacity where you plan to publish the report.
- If your tenant is outside the US or EU data boundary, Copilot is disabled by default until a Fabric tenant admin enables a setting that lets data sent to Azure OpenAI be processed outside the region.
- After you buy or scale a capacity, Microsoft says it can take up to 24 hours for Copilot to recognise the change.
Things worth knowing before you rely on it
- Preparing your data: Microsoft advises preparing the semantic model for AI before use, and says that without preparation Copilot can misinterpret the data and return generic or inaccurate results.
- Where answers come from: when a question relates to data in the semantic model, Copilot uses it; otherwise Microsoft says it answers from the language model's general knowledge.
- Verified answers: authors of Power BI apps can prepare verified answers for common questions, which Copilot can surface to improve consistency.
- Cost: token consumption drives billing for Copilot in Fabric and appears as capacity usage; downstream actions it triggers (queries, refreshes, email subscriptions) are billed separately through the normal Fabric paths.
What this means for a small team
If you already run on Microsoft 365 and Fabric, and have people who build and maintain semantic models, Copilot is a natural extension and benefits from governance features a small tool does not have. If you do not already have the capacity and the modelling skills, the first-answer path involves buying capacity, preparing a model and enabling a setting before the first useful answer.
When a simpler tool fits
- You want to ask questions of Postgres, MySQL, BigQuery, Snowflake or a sheet today, without a modelling project.
- You want every number to come with its SQL so a colleague can check it.
- You do not need SSO or a compliance certification yet. (SchemaChat has neither today.)
That is a description of where SchemaChat fits, not a claim that it is better. For governed enterprise reporting, a Microsoft shop is usually better served by Power BI. See the side-by-side comparison for both.