Improve model clarity by proactively resolving similar definitions that confuse AI agents. The Similar Meanings (Beta) feature uses AI to detect synonyms, acronyms, and semantic overlaps to prevent ambiguity. By analyzing field labels, descriptions, and parent object semantics, the tool identifies when a model contains "two names for one thing" or "the same name for different things"—concepts that often lead to inaccurate agent reasoning. This deep scan allows you to distinguish between closely related fields, helping agents correctly interpret your data and provide more reliable answers.
Required Editions
Generate a Semantic Model is a pilot or beta service that is subject to the Beta Services Terms at Agreements - Salesforce.com or a written Unified Pilot Agreement if executed by Customer, and applicable terms in the Product Terms Directory. Use of this pilot or beta service is at the customer's sole discretion.
| View supported editions. |
Interact with the Optimize Model panel to manage identified overlaps and improve your model's AI-readiness strength.
Run an Initial Scan
- From the assets list, select your semantic model.
- On the Model tab, click AI Optimization and then Optimize Model.
- Follow the UI guidance and click the scan button to begin the analysis.
- View the Similar Meanings card to see the initial summary of the AI scan results.
Resolve Similar Meanings
- On the Similar Meanings card, click Review to see the list of identified similar meanings.
-
Review each identified similar pair.
- Examine the flagged fields to determine if they represent "two names for one thing" (such as ARR vs. Annual Recurring Revenue) or "the same name for different things".
- Review the Classification Labels and Reason on each list item for further details on the identified similarity and why the system identified the pair as similar.
Tip Use the provided reason as an indicator for potential issues and their resolutions. -
Distinguish the fields by taking one of these actions.
Clarify Field Label Provide a unique and clear display name. Clarify Field Description Provide a distinguishing description, or use the Description Generation feature. Hide Field Set Field Visibility to False to hide the field from analytics dashboards and AI agents. Remove Field Navigate to the full builder properties view to remove the field or update the table definition. -
Manage your progress by updating the Status of each item.
Unresolved Pairings you haven’t resolved. Tip Filter for unresolved items to focus on pending similar meanings instead of the full list.Pending Evaluation Pairings you have taken action on.
- Limitations
- Keep these considerations in mind when using the Similar Meanings feature.
- Supported Metadata
-
- Field and Object Metadata: Scans are limited to field and object Display Names and Descriptions. While other properties provide context, the identification process focuses on these two properties due to their significant weight in search algorithms.
- Logical View Definitions: System logic does not evaluate implementation details within logical views. The system treats a logical view as a flat object, analyzing only its display name, description, and field definitions.
- Parameters: These semantic definitions are not currently scanned for similar meanings.
- Model Compatibility
-
- PDS-Based Models: The tool scans PDS-based models for similar meanings, but users can only edit the Description and Field Visibility to distinguish between fields.
- Extended Models: Similar meanings are not supported for extended models. Neither base nor extended objects are included in the scan.
- Results and Reasoning
-
- Clusters: If more than two fields are similar, the system may surface multiple pairs of similar meanings within the same cluster.
- Hierarchies: Fields representing a hierarchy (such as Category Level 1 and Category Level 2) can sometimes be flagged as similar meanings. To help the agent distinguish between these levels, provide clear descriptions, including examples of the values each level represents.