How to Test an AI 2.0 Agent in BoldDesk
Testing AI Agents in BoldDesk ensures they respond accurately and consistently before going live. After creating or opening an AI Agent from the AI Agents page, it loads in the AI Agent configuration page, where you can configure and test the agent without leaving the screen.
The configuration menu on the left—Instructions, Libraries, AI Tools, and Settings—allows you to update different aspects of the active agent during testing without changing context. At any point, you can use the Playground (Preview panel) to simulate customer queries, validate responses, and verify tool or API behavior in real time.
Prerequisites
Before testing an AI Agent, ensure the following requirements are met:
- You have access to the AI Centre in BoldDesk.
- You have the AI Agent Builder Full Access permission.
- An AI Agent has been created, or you have permission to create one.
- Any required Libraries, AI Tools, or external integrations have been configured before testing.
Open or Create the AI 2.0 Agent You Want to Test
-
Go to AI Center in BoldDesk.
-
Click AI Agents.
-
On the AI Agents page, choose one of the following:
-
Create a New AI Agent
Click Create AI Agent, enter required details such as Name, Description, and Model, then click Create.
-
Edit an Existing AI Agent
Locate the AI Agent you want to test, click the More options icon on the agent card, and select Edit.
-
The selected or newly created AI Agent opens in the AI Agent configuration page workspace. This AI Agent becomes the active AI Agent for the AI Agent configuration page and for the Playground.
You can test the AI 2.0 Agent while creating or managing it. Explore more on How to Create and Manage AI 2.0 in BoldDesk
Open the Playground Test Panel
When the AI Agent opens in the AI Agent configuration page workspace, the Playground may appear in a collapsed state on the right side.
To display it:
-
Locate the Playground tab on the right side of the page.
-
Click Playground.
This opens the Playground panel, where you can test the currently opened AI Agent.
Test the AI Agent
Once the Playground is open:
-
Enter a sample customer question.
-
Click Send.
-
Review the generated response.
-
Verify that the response aligns with the configured instructions.
-
Confirm that the correct libraries, tools, and data sources were used.
-
Repeat testing with different types of customer queries.
Test both common and edge-case scenarios to evaluate response quality and behavior.
How the Playground Works
The Playground is always tied to the AI Agent currently open in the AI Agent configuration page. This means:
-
The Playground displays the active AI Agent.
-
Switching between Instructions, Libraries, AI Tools, or Settings does not change the agent being tested.
-
Only the configuration view changes when selecting different modules.
Benefits of Using the Playground
- Test the AI Agent without leaving the configuration page.
- Validate responses immediately after making changes.
- Move between configuration sections while testing the same agent.
- Confirm the impact of updates to instructions, tools, libraries, or settings.
-
Testing an AI Agent in the Playground helps validate its responses, instructions, knowledge sources, tools, and integrations before deployment. However, successful test results do not guarantee that every AI-generated response will be accurate in a production environment.
-
Response quality can vary based on the selected AI model, configured instructions, available knowledge sources, connected tools, customer query, and the context provided during the conversation.
-
Before deployment, test the AI Agent with representative customer questions, including ambiguous, incomplete, unsupported, and edge-case queries. After deployment, regularly review production conversations and update the AI Agent’s instructions, knowledge sources, tools, and escalation behavior as needed.
Troubleshooting
The Playground does not generate a response
- Verify that an AI model has been assigned to the AI Agent.
- Confirm AI credits are available.
- Check that the AI Agent configuration was saved successfully.
The response does not use library content
- Verify that the required libraries are linked to the AI Agent.
- Confirm the library contains relevant content.
- Test with a query that matches available knowledge.
AI Tools are not being triggered
- Confirm the AI Tool is linked to the AI Agent.
- Verify the AI Tool configuration is valid.
- Test using a query that should invoke the tool.
Frequently Asked Questions
-
Why is the Playground not visible when I open an AI Agent?
The Playground may be collapsed by default. To display it, click the Playground tab on the right side of the AI Agent configuration page. -
Does switching between Instructions, Libraries, AI Tools, or Settings change the AI Agent in the Playground?
No. These options only change the configuration section you are viewing or editing. The Playground continues to show the same AI Agent that is opened in the AI Agent configuration page workspace. -
How do I test a different AI Agent in the Playground?
To test a different AI Agent, return to the AI Agents page and open another AI Agent, or create a new one. The AI Agent that you open or create becomes the active AI Agent in the AI Agent configuration page and in the Playground. -
Can I use the Playground while updating the AI Agent configuration?
Yes. The Playground is designed for testing the currently opened AI Agent while you update its instructions, libraries, tools, or settings. -
Is it expected that the same AI Agent remains in the Playground across all configuration modules?
Yes. This is expected behavior. The AI Agent that is opened in the AI Agent configuration page workspace remains active across the entire configuration page, and the Playground continues to test that same AI Agent.
