Understand Curation Automation Rules¶
Alation Cloud Service Applies to Alation Cloud Service instances of Alation
Before creating a curation automation rule, understand the various building blocks in detail.
The fundamental components that make up a rule include:
Purpose: An explanation of why the rule exists.
Scope: The specific set of data assets (schemas, tables, or columns) that the rule will target.
Fields to Populate: The metadata fields, such as Stewards, Descriptions, or classifications represented by Picker-type fields, that the rule is designed to update.
Value Sources - The method used to derive new values, which can be Deterministic (manual/fixed values) or AI-generated (using confidence thresholds).
Purpose¶
A clear Purpose defines the goal of the curation rule and provides initial context for the AI agent to follow alongside your field-level instructions. Provide a clear context through the Purpose field for the AI agent to produce the desired output. Ensure that your Purpose statement contains the right tone, level of detail, intended audience, and governance framing.
Remember to include the following while writing your curation rule’s Purpose statement:
What are you trying to achieve
Why are you doing it
Who is the audience
Your industry or domain for the AI Agent to accurately interpret industry-specific terminology (for example, aviation, pharma, manufacturing, finance)
Here’s an example of a Purpose statement:
We are a healthcare organization. This rule ensures that all tables and columns in the Clinical Gold layer have clear, business-friendly descriptions. These descriptions must avoid technical jargon and help clinicians and analysts understand the dataset’s purpose without referencing internal system codes.
Scope Selection¶
Defining the scope is a significant part of rule creation. This step determines the specific data objects your curation rule targets. We recommend starting with a small, limited scope and then expanding. Every time you run the rule, it curates only the empty fields. Previously curated objects are not reprocessed unless explicitly configured, allowing you to expand the scope and rerun the rule safely.
You can mix and match selections across different object levels to create a precise list of assets for bulk curation.
For each data source, you can define your scope using three main criteria:
Structural Selection: Choose specific data sources or hierarchical parts of a source (example, specific schemas or tables) to include in the rule.
Object Types: Filter your scope by selecting whether the rule applies to Schemas, Tables, or Columns.
Selection Logic: For each object type, you can choose:
Do not apply: Excludes this object type entirely from the rule.
Select specific: Allows you to manually pick individual objects (e.g., choosing only the Finance and Audit schemas).
All [Objects]: Automatically includes every object of that type within the selected data source.
You do not need to follow a strict hierarchy. For example, you can target specific Columns directly, even if their parent Schemas or Tables are not explicitly selected for the rule. This allows you to target properties (like specific field names) across your entire data source.
As a best practice, to improve the outcome of your rule, use a column-first approach: start with columns, then curate tables, and finally expand to schemas. This approach to building a rule helps as columns are at the granular level of data, and if column descriptions and classifications are accurate, it:
Improves Table descriptions automatically.
Provides precise PII classifications.
Provides stronger trust signals.
Fields to Populate¶
Once you have defined where the rule applies, you must choose which metadata fields the rule will populate.
By default, the rule will only populate fields that are currently empty. The rule will not overwrite the existing metadata.
You can choose one of the ways to generate the field values:
Manual vs. AI Values: You can enter values manually or allow AI assistance to suggest values based on your data context.
AI-Only Fields: Some fields are designated as AI-only, meaning they rely exclusively on high-confidence suggestions by Alation AI Agent.
AI Instructions¶
When using AI to generate values for your field, you must provide instructions for the AI to use and to produce results that enhance your curation. Providing clear, actionable guidance to the AI is essential for automating high-quality metadata. When you select a field for AI generation, you are essentially defining the logic for the Alation AI Agent to follow. As a best practice, tailor your instructions specifically to field semantics, which means you stick to the business context of the field, such as focusing on business impact for Descriptions or compliance standards for Policy fields. You can use up to 2,000 characters to frame your AI instruction.
Note
Low-confidence results (AI instructions with a confidence score below 80%) aren’t applied; confidence thresholds aren’t user-configurable.
Example AI Instructions¶
The following examples illustrate how to tailor instructions to specific field semantics to achieve higher-quality curation results.
Description¶
Generate a clear, accurate description of this data object based on its name, metadata, relationships, and available context. Explain its purpose, the type of information it represents, and how it fits within the broader dataset or domain.
Title¶
Create a clean, readable, business-friendly title for this object by interpreting its technical name, abbreviations, and conventions. Expand acronyms where appropriate and produce a meaningful title that helps users quickly understand what the object represents without altering the underlying intent.
PII Classification¶
Evaluate whether this object contains or relates to personally identifiable information by analyzing naming patterns, metadata, associated attributes, and contextual indicators. If the object depends on underlying components (such as child fields or attributes), incorporate those dependencies when determining whether PII is present. Classify only when there is clear, evidence-based justification.
Configure Agents¶
After the field selection is complete, you select an asset to review its fields and fine-tune the curation settings. You will understand how your rule updates the fields and how the AI Agent implements the curation instruction to generate the fields. You can review the results, and based on how effective the results are, you can modify your AI instructions.
Review and Audit¶
Before running the rule, review the changes the rule will apply, preview the curation summary, and get to know your AI Actions score.