AI Agents settings page (in preview)
In this topic
On the AI Agents page of the Collibra settings, you can activate AI agents and generally configure Collibra AI Copilot.
Note You need a global role with the Manage all agents global permission to access the AI Agents settings. For more information, go to Collibra AI Copilot configuration.
AI Agents tab page
From the AI Agents tab page in the AI Agents settings, you can check the status of AI agents in Collibra AI Copilot and activate them.
- The status of an AI agent can be Preview or Generally available.
- By default, AI agents are not active when Collibra AI Copilot is enabled. To activate an AI agent, set the Active column to Yes.
- To open a dedicated page and configure an AI agent, click the agent's name.
- Business Definitions agent
- Data and Analytics Discovery agent
- Collibra Documentation agent
For the Business Definitions agent, you can add example questions and customize the content scope.
Example questions
Collibra AI Copilot can show up to 3 clickable example questions. These questions help users understand the types of queries they can ask and allow them to click to see how the AI responds.
By default, no example questions are available for any AI agent. You can add up to 3 example questions per AI agent, with each question limited to 100 characters.
Note If more than 3 example questions exist for the active AI agents, Collibra AI Copilot selects up to 3 to show when a chat starts.
Content scope
The content scope of the Business Definitions agent is based on asset type, status, and organization. The AI agent considers all 3 content scope configurations.
The default, out-of-the-box content scope includes the following:
- Asset types: The following asset types are included in the content scope. Child asset types aren't automatically included.
- Business Term
- Acronym
- KPI
- Measure
- Statuses: All statuses are included.
On the page, this is indicated with no default selection. - Organization: All communities and domains in your organization are included.
On the page, this is indicated with no default selection.
Note In addition to the content scope, Collibra AI Copilot shows an asset only if the asset has a description, description from source, or definition, and if the user has access to the asset.
As an administrator, you can customize the out-of-the-box content scope to match your organization's needs. To change the content scope:
- Manually change the scope by selecting or clearing items.
The Business Definitions agent allows you to select from all asset types in the Business Glossary product.
- Click Save.
Tip You can revert to the out-of-the-box content scope by clicking Reset to default. When you do so, the default content scope is applied and automatically saved.
For the Data and Analytics Discovery agent, you can add example questions and customize the content scope.
Example questions
Collibra AI Copilot can show up to 3 clickable example questions. These questions help users understand the types of queries they can ask and allow them to click to see how the AI responds.
By default, no example questions are available for any AI agent. You can add up to 3 example questions per AI agent, with each question limited to 100 characters.
Note If more than 3 example questions exist for the active AI agents, Collibra AI Copilot selects up to 3 to show when a chat starts.
Content scope
The content scope of the Data and Analytics Discovery agent is based on asset type, status, and organization. The AI agent considers all 3 content scope configurations.
The default, out-of-the-box content scope includes the following:
- Asset types: The following asset types are included in the content scope. Child asset types aren't automatically included.
- Report
- Data Notebook
- BI Report and all its child asset types
- Data Product and all its child asset types
- AI Model and all its child asset types
- Data Set and all its child asset types
- Statuses: All statuses are included.
On the page, this is indicated with no default selection. - Organization: All communities and domains in your organization are included.
On the page, this is indicated with no default selection.
Note In addition to the content scope, Collibra AI Copilot shows an asset only if the asset has a description, description from source, or definition, and if the user has access to the asset.
As an administrator, you can customize the out-of-the-box content scope to match your organization's needs. To change the content scope:
- Manually change the scope by selecting or clearing items.
The Data and Analytics Discovery agent allows you to select from all asset types, except Columns and the asset types associated with the Business Glossary product.
- Click Save.
- You can also set the content scope to the same scope as Data Marketplace by clicking Load Data Marketplace scope.
When you do this, part of the Data Marketplace scope is copied and automatically saved in the AI agent content scope. It includes statuses, organization, and asset types, but excludes the Column asset type and asset types associated with the Business Glossary product. - You can revert to the out-of-the-box content scope by clicking Reset to default. When you do so, the default content scope is applied and automatically saved.
For the Collibra Documentation agent, you can add example questions.
Collibra AI Copilot can show up to 3 clickable example questions. These questions help users understand the types of queries they can ask and allow them to click to see how the AI responds.
By default, no example questions are available for any AI agent. You can add up to 3 example questions per AI agent, with each question limited to 100 characters.
Note If more than 3 example questions exist for the active AI agents, Collibra AI Copilot selects up to 3 to show when a chat starts.
AI Copilot tab page
From the AI Copilot tab page in the AI Agents settings, you can create a welcome message, which can include a maximum of 300 characters. This welcome message is shown when a user starts a conversation with Collibra AI Copilot.
Tip Use the welcome message to set user expectations based on the agents you have activated.