Power BI Prompt Guide

DAX made approachable

Describe the calculation and your model, and ask the AI whether it should be a measure or a calculated column. Always request an explanation of the logic so you can maintain it later.

Designing measures

Ask for a base set of reusable measures (total, average, running total, YoY) rather than one-off formulas. Reuse keeps your model clean.

Data models

Prompt the AI to design a star schema with clear fact grain and dimensions. A good model makes every downstream measure simpler.

From prompt to published report

A good workflow is to prompt for the model and base measures first, then prompt for individual visuals and their DAX, and finally ask for formatting and conditional-formatting suggestions. Building in that order keeps each step simple and prevents the model from inventing fields that do not exist in your tables, which is the most common source of broken measures.

Always paste your actual table and column names, and clarify whether a calculation should be a measure or a calculated column, since the two behave differently under filter context. Ask the model to explain the filter context a measure relies on, because that is where most DAX bugs hide. Once a measure works, document its purpose in a comment so future-you, or a colleague picking up the report, can maintain it without reverse-engineering the logic from scratch.

Finally, ask the model to suggest a sensible visual type for each measure and the sort order that makes the trend clearest. A well-chosen chart communicates a measure far better than a table of numbers, and the AI can justify each choice so you learn the design principles rather than just accepting a recommendation blindly.