Managing Credentials for AI Transforms in Power BI Desktop

The November Power BI Desktop release introduced AI Transforms in Power Query. These functions for Text Analytics, Vision, and Azure Machine Learning are available to enrich your data with basic sentiment, key phrase extraction, vision tags, and whatever you might build with custom models through Azure ML.

Premium Capacity

One of the things that users will quickly learn is that using the AI Transforms require the AI workload from a Power BI Premium capacity. Attempting to sign in with a user that is not associated with any dedicated capacity presents a message stating that you need “access to a Premium capacity with the AI workload enabled”. For many people, this may be a non-starter, but it makes sense from the consistency of refreshing reports in Desktop and then when published to Service. It also aligns with the AI Insights features (the same Power Query transforms) that are available in Dataflows, which also require the AI workload in Premium.

This is not the first set of features in Power BI Desktop that require Premium to ultimately use. For example, Incremental Refresh can be setup in Desktop and then takes effect once published to a dedicated capacity in Service. To my knowledge, however, AI Transforms are the first to require Premium to use the feature fully while still in Desktop.

AI Transforms Act Like Data Sources

For anyone that frequently switches between accounts in Power BI Desktop, using AI Transforms may be tricky if you forget which dedicated capacity you are authenticated to. Why is that?

Users are presented with an authentication request when they first attempt to use the AI functions. While trialing the AI functions, I accidentally signed in with a user that is not tied to any dedicated capacity.

This led to the message stating that I did not have access, but nowhere on the screen does it show how to authenticate with a different login.

No problem, I thought. I signed out of Power BI from the top right corner of the application and signed in using the “correct” user. WRONG! This step signs in with the user that is tied to dedicated capacity, but it does not change the credentials stored for AI Functions.

After some initial confusion, I learned that the credentials for AI Functions are managed under Data Source Settings. To login as a different user to access dedicated capacities, you have to do this from Data Source Settings rather than the AI window. This is similar to Power BI Dataflows authenticating as the stored credentials under Data Source Settings rather than the user logged into Service overall in the top right corner. If you’ve ever logged into Power BI Desktop and then went to import a dataflow only to find that they were dataflows from some other tenant…welcome to my world.

Both AI Functions and Azure Machine Learning Functions store credentials under Data Source Settings. Edit or clear them as needed from Data Source Settings rather than inside Power Query or the Service login in the top right corner.

Once I edited the credentials under Data Source Settings, I was able to use the AI Transforms as expected.

Since I switch accounts frequently, I’ll probably keep forgetting this aspect like I often do for Dataflows. Hopefully reading this post will help prevent you from making the same mistake that I made though.