Configuring Custom Vault AI Metadata

Vault AI metadata allows you to provide an additional layer of detail on your Vault data model to agents. With Vault AI metadata, you can define details related to the purpose of specific entities, such as objects, documents, and fields, which can include the business context, synonyms, and thorough descriptions. For example, a user may include a Security Tree-class object in a Vault AI Chat query. The agent may not understand the purpose of the object, when and when not to use it, and how to use it in the query. In this case, Vault AI metadata can help agents accurately interpret the object when it is included in a query and generate an appropriate response.

You can define Vault AI metadata for entities and Picklists. Each entity (objects and document types) can also include metadata for fields, object relationships, and metrics. Each Picklist can include metadata for a Picklist, object lifecycle values, and Picklist values. When these components are included in agent context, instructions, or tool configurations, the agent uses the metadata to execute queries properly. Without metadata, agents typically can only query Name and Label field values, which can lead to misunderstandings and inaccurate responses. Vault AI metadata ensures agents can see the whole picture of your Vault data model and not just surface level details.

Agent permission to entities and Picklists with metadata is determined by the assigned agent user’s permissions.

Synonyms

You can assign synonyms to Vault AI metadata, which allows agents to query different terms and phrases related to the entities, fields, relationships, Picklists, and metrics. Synonyms do not allow for exact matches, meaning agents must still interpret the synonym to determine which data it refers to. For example, synonyms for a Region field could include District, Territory, and Zone. When a user enters one of the synonyms in Vault AI Chat, the agent knows it is related to the Region field and continues to generate an appropriate response. Agents can also verify typos and route to the appropriate data, such as Regoin for Region.

When writing a synonym, avoid obvious phrases, such as marketing brochures, marketing brochure record, and marketing brochure item for a Marketing Brochure object. Instead, use jargon, abbreviations, or terms an agent cannot use to map to data in your Vault. For example, you could use synonyms such as collateral, leave-behinds, leaflets, and GTM for a Marketing Brochure object.

How to Configure Entity Metadata

To define an entity’s metadata:

1. Navigate to Admin > Configuration > Vault AI Metadata > Entities.
2. Click Create.
3. Select Object or Doctype from the Type drop-down.
4. Based on your Type selection, select an active object or document type from the Target drop-down. You cannot change this selection after saving the entity. You can only select objects and document types not already in use by another entity.
5. The Target Label and Label are populated based on your Target selection. Update the Label value as needed. The Target Label value is updated if the selected object or document type’s Label is updated.
6. Enter an API Name.
7. Enter a Description. The Description should include details on how an agent should interpret the entity. For example, you can include when the agent can and cannot use the entity, how the agent should use the entity, and its overall purpose. Ensure the description is clear and concise. A clear and accurate description ensures agents can properly interpret the entity. If there is already a description in place, you can override it by entering a new one.
8. Optional: Enter up to ten Synonyms for the entity. Type out the synonym and press Enter to add it.
9. Click Save.

The entity is now configured with metadata. You can begin configuring metadata for fields, object relationships, and metrics related to the entity.

How to Configure Field Metadata

To configure an entity field’s metadata:

1. Select Fields from the entity configuration page. This drop-down is only available after you’ve created and saved the entity metadata.
2. Click Create.
3. Select an active field from the entity in the Target drop-down.
4. The Target Label and Label are populated based on the Target selection. Update the Label as needed. The Target Label value is updated if the selected field’s Label is updated.
5. Enter an API Name for the field.
6. Enter a Description. The Description should include details on how an agent should interpret the field. For example, you can include when the agent can and cannot use the field, how the agent should use the field, and its overall purpose. Ensure the description is clear and concise. A clear and accurate description ensures agents can properly interpret the field. If there is already a description in place, you can override it by entering a new one.
7. Optional: Enter up to ten Synonyms. Type out the synonym and press Enter to add it.
8. Optional: Enter up to ten Example Values. You can enter any term or phrase a user may enter in the selected Target field. The Example Values field is available only if the Target is a Text field.
9. Optional: If the Target field is a Picklist with metadata configured, you can override the metadata settings for it or any of its values. Enter a new Picklist Description and Picklist Synonyms as needed. The Picklist’s defined metadata settings are used instead if you do not override them.
10. Click Save.

The entity field’s metadata is now configured. When an agent queries this field, it uses this metadata to generate an appropriate response.

How to Configure Object Relationship Metadata

Relationship configuration is only available for object entities. To configure object relationship metadata:

1. Select Relationships from the entity configuration page. This drop-down is only available after you’ve created and saved the entity metadata.
2. Click Create.
3. Select an active object reference field on the entity’s object from the Target drop-down.
4. The Target Label and Label are populated based on the Target selection. Update the Label as needed. The Target Label value is updated if the selected object reference field’s Label is updated.
5. Enter a Description. The Description should include details on how an agent should interpret the object relationship. For example, you can include when the agent can and cannot use the object relationship, how the agent should use it, and its overall purpose. Ensure the description is clear and concise. A clear and accurate description ensures agents can properly interpret the object relationship. If there is already a description in place, you can override it by entering a new one.
6. Enter up to ten Synonyms. Type out the synonym and press Enter to add it.
7. Click Save.

The object relationship metadata is now configured. When an agent queries this relationship, it uses this metadata to generate an appropriate response. The Cardinality field displays the type of relationship: ONE-TO-ONE, MANY-TO-ONE, ONE-TO-MANY, MANY-TO-MANY. If the Cardinality is MANY-TO-MANY, the Name of the join object used in that relationship is populated in the Join Object field.

How to Configure Metrics Metadata

Vault AI metadata for metrics specifies which fields to use and how to calculate them when a user asks a metric related question in Vault AI Chat. For example, a user may ask how many days a document review is overdue or how much time they have until a review period closes. In such scenarios, if the agent cannot map the data to a single field, it attempts to map and combine data across multiple fields using existing field Label values, which could lead to inaccurate results.

To configure metadata for an entity’s metrics:

1. Select Metrics from the entity configuration page. This drop-down is only available after you’ve created and saved the entity metadata.
2. Click Create.
3. Enter a Label.
4. Enter an API Name.
5. Enter a Description. The Description should include details on exactly what the metric represents. For example, you can include when the agent can and cannot use the metric, how the agent should use the metric, and the metric’s overall purpose. Ensure the description is clear and concise. A clear and accurate description ensures agents can properly interpret the metric. If there is already a description in place, you can override it by entering a new one.
6. Enter up to ten Synonyms. Type out the synonym and press Enter to add it.
7. Select up to ten Target Fields. Agents can use these fields when querying the metrics of your metadata.
8. Enter Instructions using a Vault formula or rule. Do not restate the Description here. The Instructions should tell the agent how to calculate the metric, which fields to use and combine, and any applicable conditions.
9. Click Save.

The metric metadata is now saved. When an agent queries the metric, it uses this Vault AI metadata to generate a response.

How to Configure Picklist Metadata

To configure a Picklist’s metadata:

1. Navigate to Admin > Configuration > Vault AI Metadata > Picklists.
2. Click Create.
3. Select Picklist or Objectlifecycle from the Type drop-down.
4. Based on your Type selection, select a Picklist or object lifecycle from the Target drop-down.
5. The Target Label and Label are populated based on the Target selection. Update the Label as needed. The Target Label value is updated if the selected Picklist’s or object lifecycle’s Label is updated.
6. Enter a Description. The Description should include details on how an agent should interpret the Picklist. For example, you can include when the agent can and cannot use the Picklist, how the agent should use the Picklist, and its overall purpose. Ensure the description is clear and concise. A clear and accurate description ensures agents can properly interpret the Picklist. If there is already a description in place, you can override it by entering a new one.
7. Enter up to ten Synonyms. Type out the synonym and press Enter to add it.
8. Click Save.

The Picklist metadata is now configured. You can begin configuring metadata for each value in the Picklist or state in the object lifecycle.

How to Configure Picklist Value & Lifecycle State Metadata

To configure metadata for a Picklist value or an object lifecycle state:

1. Select Picklist Values on the Picklist configuration page. This drop-down is only available after you’ve created and saved the Picklist metadata.
2. Click Create.
3. Select a Target. If the parent entity is a Picklist, select a value from the Picklist. If the parent entity is an object lifecycle, select a state.
4. The Target Label and Label are populated based on the Target selection. Update the Label as needed. The Target Label value is updated if the selected value’s or state’s Label is updated.
5. Enter an API Name.
6. Set the Status if necessary. Active is selected by default.
7. Enter a Description. The Description should include details on how an agent should interpret the Picklist value or object lifecycle state. For example, you can include when the agent can and cannot use the values, how the agent should use the values and their overall purpose. Ensure the description is clear and concise. A clear and accurate description ensures agents can properly interpret the value. If there is already a description in place, you can override it by entering a new one.
8. Optional: Enter up to ten Synonyms. Type out the synonym and press Enter to add it.
9. Click Save.

The Picklist value or object lifecycle state metadata is now configured. When an agent queries a value or state, the metadata is used to generate an accurate response.

Deleting a Metadata Entity or Picklist

If an entity is referenced in an agent context, Vault displays a warning before it is deleted. If deleted, the agent context can no longer use the entity.

If an entity is deleted, all of its related field, object relationship, and metric metadata are deleted as well. The same behavior applies to Picklists and any related Picklist values. If a Target referenced by an entity, field, object relationship, metric, Picklist, or Picklist value is deleted, its associated metadata is also deleted.

Best Practices

The following best practices apply to creating Vault AI metadata:

• It is recommended to create Vault AI metadata for the following types of objects and document types:
  • High-traffic objects and document types that appear often in Vault AI Chat inquiries
  • Objects and document types with names that do not clearly convey their purpose
  • Objects and document types with similar related objects or document types, such as any object that an agent could confuse with another
  • Picklists on key filter fields, if a Picklist determines how users phrase queries, define Picklist and Picklist value metadata
• Only add metric metadata when a query requires combining data across multiple fields, such as calculating a duration between two dates, or aggregating data across multiple objects, such as Categories, Sites, and Studies.
• Use Example Values when a Text field may contain common terms or phrases users reference in Vault AI Chat queries. If there are not predictable terms or phrases a user may enter in the field, leave Example Values blank.
• Add object relationship metadata if the relationship Name does not clearly convey its purpose, the context is not clear, or you want to specifically guide the agent on how to use and not use the relationship in queries.

Limits

The following limits apply to creating Vault AI metadata for Entities and Picklists:

• Up to 50 custom Entities per Vault
• Up to 50 fields per entity
• Up to 30 object relationships per entity
• Up to 20 metrics per entity
• Up to 45 Target Fields per metric
• Up to 100 Picklist per Vault
• Up to 50 values per Picklist
• Up to 10 values per synonym
• Up to 10 example values per field
• Multiple Entities cannot reference the same object, document type, field, object relationship, or metric.

Related Permissions

Type	Permission Label	Controls
Security Profile	Admin: Configuration: AI Metadata: Read	Controls your ability to view Vault AI metadata
Security Profile	Admin: Configuration: AI Metadata: Create	Controls your ability to create Vault AI metadata
Security Profile	Admin: Configuration: AI Metadata: Edit	Controls your ability to edit Vault AI metadata
Security Profile	Admin: Configuration: AI Metadata: Delete	Controls your ability to delete Vault AI metadata