Testing your AI Agent

The Test AI Agent feature allows you to validate how well the agent understands and responds to queries based on the topics derived from your connected content sources. These topics and subtopics are automatically generated and define the knowledge scope of your AI Agent. You cannot edit or remove them, but they provide the foundation for how the agent interprets and responds to user questions.

Testing can be carried out in two ways:

  1. Integrated Mode – Test how the AI Agent performs in your live environment alongside your existing systems.

  2. Standalone Mode – Evaluate the AI Agent independently to review its topic coverage, accuracy, and response quality.

To test an AI Agent:

  1. Create an AI Agent.

  2. In the Test AI Agent step, the Intents section lists the intents derived from the content sources you provided. A working preview of the AI Agent is available on the same page.

  3. You can edit or delete intents that are derived from source topics. To edit an intent, click the edit icon next to it.

    Note: If the AI Agent is created using an eGain Knowledge Base, source-derived intents cannot be edited. You can only delete them.

  4. To add a custom intent, click Add Intent. In the Add Intent dialog, provide the following details and click Save:

    • Name (required): Enter a name for the intent.
    • Description: Enter a description. The AI Agent uses this description to identify the intent.
    • Assign Workflow: Select a process workflow to associate with this intent. When a customer query matches this intent, the AI Agent routes it directly to the assigned workflow, bypassing KB search. Workflows are created in the Workflows tab. For more information, see Creating Process Tools Workflows.

    Add Intent dialog

     

  5. Use the test panel on the right to interact with the AI Agent and evaluate its responses. You can test the AI Agent in two modes using the Integrated and Standalone tabs.

Integrated

The Integrated tab simulates the AI Agent as it would behave within a live agent-assisted contact center environment. Use this view to test how the AI Agent handles customer queries in the context of an ongoing conversation, including how it routes intents to assigned workflows.

Test AI Agent

To test in Integrated mode:

  1. Click the Integrated tab in the test panel.

  2. Type a query in the message box and click Send. The AI Agent responds based on the configured intents and data sources.

  3. Review the responses to verify that the AI Agent correctly identifies intents and applies the expected workflows and knowledge sources.

  4. When you are satisfied with the results, click the Proceed to Next Step button.

Standalone

The Standalone tab simulates the AI Agent operating independently, without live agent involvement. Use this view to test the AI Agent's self-service behavior, including its greeting, intent handling, and responses when no agent escalation is involved.

Test AI Agent

To test in Standalone mode:

  1. Click the Standalone tab in the test panel.

  2. Type a query in the message box and click Send. The AI Agent responds as it would in a self-service deployment.

  3. Review the responses to verify that the AI Agent handles queries correctly without agent assistance.

  4. When you are satisfied with the results, click the Proceed to Next Step button.