Skip to main content

Use AI-suggested rules

Atlan AI automatically suggests data quality rules based on your asset's metadata structure. Suggestions are generated from column names, data types, descriptions, and other metadata attributes. Common suggested rules include null and blank validations, uniqueness checks, duplicate detection, freshness monitoring, and volume checks. You can review, adjust thresholds, and apply the suggestions in a few clicks.

Prerequisites​

  • Atlan AI must be enabled with the Enrich metadata setting in Admin → Labs → Atlan AI. See Enable Atlan AI.
  • You need permission to create rules. See Data quality permissions for Snowflake or Databricks.
  • Supported platforms: Snowflake and Databricks.

Generate AI-suggested rules​

Atlan AI analyzes your asset's structure to suggest relevant data quality rules automatically. The quality of suggestions depends on your asset metadata—column names, data types, descriptions, database, and schema information. Make sure your catalog metadata is up to date for the best results.

  1. In Data Quality Studio, navigate to the asset where you want to add rules.

  2. Atlan AI analyzes your asset's metadata and generates a list of suggested rules. Click Add rule (or open an existing rule set), then click Suggest rules beside Add rule.

  3. Review the suggested rules. Typical suggestions include null counts, blank counts, freshness checks, duplicate detection, and volume checks.

  4. Select the rules you want to apply. Adjust thresholds and parameters as needed. Adjust thresholds before applying suggestions to match your data quality requirements.

  5. Click Apply to add the selected rules to your rule set.

  6. Configure any additional settings, then save your rule set.

Generate custom SQL rules with Copilot​

For custom validation logic beyond predefined rules, Copilot can generate SQL-based rules from natural-language descriptions. This helps you create complex rules without writing SQL from scratch.

  1. Click Add rule, then select Custom SQL.

  2. In the Custom SQL dialog, under SQL Query, click Ask AI to generate code.

  3. Describe the rule in natural language. For example, "flag rows where status not in ('ACTIVE','INACTIVE')." or "percent of nulls in email < 1%."

  4. Copilot generates SQL tailored to your asset's context, including columns, data types, descriptions, database, and schema. Always review and validate Copilot-generated SQL before saving to confirm it matches your requirements. Review and edit the proposed SQL. Click Test SQL Query to validate.

  5. Set Pass if query result is and select the Dimension.

  6. Click Create to save the rule.

Need help​

If you have questions or need assistance, reach out to Atlan Support by submitting a support request.

See also​