High-quality metadata makes a dataset easy to understand and safe to use. It explains what fields mean (names, types, descriptions), so users can quickly find the right columns, interpret them correctly, and build consistent dashboards without guessing.
It also helps AI work better. When metadata is clear and consistent, AI can match questions to the right fields and produce more accurate suggestions—because it relies on schema meaning, not “tribal knowledge.”
You can improve your dataset metadata in two ways. Open a dataset, click Edit details, then use the Improve Metadata menu in the top-right. From there, you can either generate missing column descriptions or get AI suggestions to improve your metadata (such as clearer column names and better column types).
Column descriptions give your users more context than column names can provide alone. Column descriptions are available to your users by hovering over the (i) icon beside any column in the dashboard editor's data tab.
These descriptions can be created and edited by opening a dataset and clicking the "edit details" button. In the "Columns" section on the left, click on any description to edit it, or click on the top right the "Improve Metadata" menu and then "Create missing descriptions" to use AI to infer column descriptions from the dataset name, column names and types.
Run Metadata Suggestions to quickly improve your dataset schema by receiving clear, actionable recommendations for column names and column types. You can request suggestions by opening a dataset and clicking "Edit details". In the top-right, open the "Improve Metadata" menu and select "Suggest metadata improvements".
Metadata Suggestions uses AI to reason from the dataset title, column names and column descriptions (it does not inspect cell values) and proposes metadata-only changes.
For example: order → order_id, or price → unit_price with a CURRENCY type. For currency fields, you can choose which currency to apply to the column.
Note the limits: the feature does not inspect values or modify data. Review each recommendation and Accept or Ignore it to update the metadata. You can also Accept All suggestions at once.
Even after applying a suggestion, you can always edit the column name and/or the type.
How to get access? Reach out to your CSM or our support team to access the beta.