Table of Contents

Documenting Your Semantic Model

Good documentation makes a semantic model usable. When columns and measures have clear descriptions, report builders can work confidently without guessing what fields mean. But writing descriptions by hand for every object in a large model is tedious.

This walkthrough shows how to combine the Data Dictionary and AI Model Documentation features to document your model quickly, then hand it off to business stakeholders for review.

Who This Is For

  • Data engineers and model developers who want to document their models without spending hours writing descriptions manually.
  • Business analysts who need to review and refine model documentation to make sure it reflects business terminology and context.

The Workflow

Step 1: Generate Descriptions with AI

Start by letting AI generate a first pass at your documentation.

  1. Open your model in Semantic Modeler.
  2. Go to the AI Model Documentation feature (requires an AI provider configured in Settings, or the AI Add-on license).
  3. Run the documentation generator. It analyzes your table names, column names, data types, and DAX expressions to produce descriptions for tables, columns, and measures.

The AI does a good job of explaining what things are based on naming conventions and DAX logic. It won't know your specific business context (like why a certain filter exists or what a particular KPI means to your organization), but it gives you a solid starting point.

Step 2: Export the Data Dictionary

Now that your model has AI-generated descriptions, export everything to a spreadsheet.

  1. Click Export Data Dictionary in the toolbar.
  2. Choose XLSX or ODS format.
  3. Save the file.

The exported workbook contains all your tables, columns, measures, and relationships with their current descriptions.

Step 3: Review and Refine

Share the spreadsheet with business analysts, subject matter experts, or other stakeholders. They can:

  • Correct any AI-generated descriptions that miss the mark
  • Add business context that the AI couldn't know (e.g., "This measure excludes returns processed after the 30-day window per company policy")
  • Standardize terminology to match what the organization actually uses
  • Flag columns or measures that need attention

This step is where the real value comes in. The AI saves hours of initial writing, and the human review ensures accuracy and business relevance.

Step 4: Import the Revised Spreadsheet

Once the review is complete, import the updated spreadsheet back into Semantic Modeler.

  1. Click Import Data Dictionary in the toolbar.
  2. Select the reviewed file.
  3. Semantic Modeler updates all descriptions in one step.

Only non-empty cells overwrite existing values, so reviewers don't need to worry about accidentally clearing fields they didn't touch.

Time Savings

For a model with hundreds of columns and dozens of measures, writing descriptions manually could take a full day or more. This workflow typically reduces that to:

  • A few minutes for AI generation
  • An hour or two for stakeholder review (which can happen asynchronously)
  • Seconds for the import

The result is documentation that combines AI speed with human expertise.

Edit this page