Skip to main content

Blog consacré à Microsoft Power BI

Fabric Copilot to help write DAX queries improvements

Headshot of article author Zoe Douglas

Product Manager
8 mai, 2025

We’ve made some exciting improvements to Fabric Copilot by using the most advanced Azure Foundry models when helping you write DAX queries on semantic models. This update comes with a marked improvement when generating DAX queries.

These improvements include overall correctness of the DAX queries, deeper understanding of how to utilize DAX row and filter context, and how to use semantic model features such as inactive relationships. This upgrade also includes increased stability, meaning similar user requests phrased in different ways create similar or identical DAX queries, leading to more predictable outcomes.

These improvements are available now in these scenarios:

These improvements are available starting in May 2025 releases in these scenarios:

  • Copilot in Power BI Desktop to help write DAX queries in DAX query view
  • Copilot in Power BI Desktop to help answer data questions with a DAX query in the Report view
  • When using semantic models with Data Agent.

Using the Adventure Works sample data, we can see this in action. I am writing DAX queries in the Fabric workspace, and I want to see for online orders, what is the top product with % of total sales by year, based on ship date.

I get a clear answer back, incorporating the use of an inactive relationship and multi-step analysis to determine the top product per year.

When asking about this data when editing a report in the web, I can ask ‘Which days of the week have the most percentage of total orders shipped’ and a DAX query uses the inactive relationship again and calculates the % of total shipped orders by day.

In addition, the results are summarized as an answer to my question. I can check the DAX query to verify which fields are being used:

And even run this query in DAX query view to make changes with Copilot’s help, such as formatting the values.

This is in addition to recent updates to include more information about the semantic model, such as using descriptions, synonyms, sample data, hierarchies, and display folders.

Try it out yourself and you can find more information about how these all work together with these resources: