You are here:
Set Up Agent Scoping for Concierge
Agent Scoping defines the list of semantic models that an analytics agent can search when a user asks a question in Concierge: Analytics Q&A with no clear context. For example, Concierge may lack context when it's used in Slack or when it's embedded in a page without Tableau Next assets such as a metric or dashboard.
Required Editions
Admins can use Agent Scoping to define a secure, approved list of semantic models for each analytics agent that uses the pre-built Data Analysis subagent with the Analyze Data action. The list of approved semantic models defined for an agent is called a fallback list.
The agent can be based on the Analytics and Visualization template or a custom agent based on the Agentforce Employee Agent template.
Analysts can view the fallback list but can't edit it.
| View supported editions. |
| User Permissions Needed | |
|---|---|
| To create and edit the fallback list: | Tableau Unmetered Admin or Tableau Next Admin permission set |
| To view the fallback list: | Tableau Unmetered Platform or Self-Service Analyst or Tableau Next Platform or Self-Service Analyst permission set |
- In Tableau Next home, click Agent Scoping.
-
Select the agent you want to configure. Only active agents in the org that use the Data
Analysis topic are available.
Agents can be based on the Analytics and Visualization agent template or a custom agent based on the Agentforce Employee Agent template.
- Click Select Models.
- In the Select Semantic Models dialog box, select each semantic model (up to 10 models allowed), and then click Select. Use the Data Space search to filter the view to semantic models in that data space.
-
Confirm that the models you selected are in the list of semantic models when you select
that agent in the Agent Scoping page.
To view details about a model, select the model name in the list, click the arrow, and then click Details. You might need to scroll to the right to see the arrow. To remove a model from the list, select the model name in the list, click the arrow, and thenTo remove a model from the list, select the model name in the list, click the arrow, and then click Remove.
Requirements
- Fallback lists of semantic models can only be created for active agents that use the Data Analysis subagent with the Analyze Data action. The active agents can be based on the Analytics and Visualization agent template or a custom agent based on the Agentforce Employee Agent template.
- The agent must be added to the Agent Access permission set to be available. For more information, see Manage Employee Agent Access.
- Semantic models added to the fallback list must belong to the same org as the agent being configured. At least one active analytics agent must be available in the same org to configure a fallback list.
Limitations and Considerations
- You can add any number of semantic models to the fallback list, however selecting 10 or fewer models improves agent accuracy.
- Each fallback list is associated with a specific agent. Adding a new analytics agent to the org doesn't inherit another agent fallback list.
- User permissions are always respected at runtime. If a user doesn't have access to a semantic model, it won't be used for their queries even if it's on the fallback list.
- When a user asks questions in Concierge with available context in a dashboard or metric, the agent uses the semantic models associated with the dashboard or metric instead of the fallback list of approved models.
- For Home and Governed orgs, one fallback list can be configured per agent. For Personal orgs, the fallback list can be configured per agent for semantic models available in the Personal org only.
