Auto-generate field descriptions within a dataset - Precisely Data Integrity Suite

This feature uses AI to automatically create meaningful descriptions for data fields in bulk, based on available metadata such as field names, data types, sample values, and classifications. It is designed to save time during dataset onboarding, bulk metadata cleanup, and governance initiatives. All generated descriptions are tracked in the dataset's Change Log for audit and compliance purposes.

1. Go to Catalog > Datasets.
2. Locate and click the dataset for which you want to generate field descriptions.
3. There are two ways to generate the field descriptions.
   1. When you click a dataset card, the side panel opens. Scroll to the Fields section to view all fields in the dataset.
   2. Alternatively, open the dataset Details page, then scroll to the Fields section to view all fields in the dataset.
4. In the Fields section, locate and click the AI Action button. You can choose between two options.
   1. Generate All Descriptions: generates descriptions for all fields in the dataset.
   2. Generate Missing Descriptions: generates descriptions only for fields that currently lack descriptions.
5. A new dialog opens, displaying field cards with a description box that will receive the generated descriptions.
   Note: By default, the override option is disabled, so only blank or missing descriptions are populated. Manually curated descriptions are preserved.
6. If a description already exists, you can select either the existing description or the AI-generated version. You can also compare both versions to inform your decision. By default, the existing description is used.
7. When a card is selected, its details are displayed in the side panel. From here, you can select all fields on the current page or choose specific fields, and save your selection by clicking the Save Selected button.
8. Alternatively, you can continue browsing across pages, select items from multiple pages, and save all selected items at once. The item count shows the total number of selections across all pages.
   Note: If the Close button is used before saving, all selections are discarded and the dialog closes. If it is used after saving, only the saved items are retained and any unsaved selections are discarded.
9. Now, the system will:
   • Analyze field metadata: names, data types, sample values, and classifications
   • Generate contextually relevant descriptions using AI
   • Populate the selected fields with the generated descriptions

   The process typically completes within seconds, depending on the number of fields.

10. After the descriptions are saved, the cards update automatically. The Save Selected action stores the description on the asset, disables the selection checkbox, and displays an Undo option to revert the change if needed.
11. Once generation completes, the Fields section refreshes to display the newly generated descriptions. Review the results by:
    • Scanning the description column for newly populated fields
    • Verifying that descriptions are accurate and relevant to your dataset
    • Identifying any fields that may require refinement or manual adjustment
      Note: Description generation completes in ≤3 seconds per field (for tables up to 500 columns).

    If generation fails for any fields, you will receive an error notification that specifies which fields failed and why (for example, missing metadata or unsupported field types).
12. Verify changes in the Change Log.
    • Timestamp of the auto-generation operation
    • User who initiated the generation
    • List of affected fields
    • Previous descriptions (if any) and new descriptions
    • Summary of the action (for example, "Auto-generated descriptions for 45 fields")

    This audit trail supports compliance requirements and allows you to track metadata changes over time.