Testing AI

Overview

After you've trained your AI Agent with knowledge sources, scenarios, and a unique personality, the next crucial step is to test it. The AI testing page provides a safe sandbox environment where you can interact with your AI just like a customer would, without affecting any of your real conversations.

This allows you to check the AI's accuracy, tone, and behavior before deploying it to handle live customer chats.

How to Test Your AI Agent

1. Navigate to the Test Page: From the side menu, go to AI Agent > Test.

2. Select the Account to Test: Your AI's knowledge and skills are tied to specific integrations (e.g., your Facebook Page, Instagram account, etc.). Use the dropdown menu at the top of the chat window to select which specific account you want to test. This ensures you are testing the correct set of knowledge sources, scenarios, and personality.

3. Start Chatting: Type a message in the input box at the bottom, just as a customer would. You can ask questions, describe problems, or use phrases that you expect would trigger a specific scenario.

4. Review the AI's Response: The AI will generate a response based on its training. You can see the AI thinking and then delivering its answer in the chat window.

Understanding the AI's Thought Process

Sometimes, the AI might give an unexpected answer. To understand why it responded in a certain way, you can use the "Show thinking" feature.

After the AI responds, a small button will appear below its message. Clicking this reveals the AI's reasoning. It will show you exactly which Knowledge Source or Scenario it used to generate the answer.

This is an incredibly powerful tool for troubleshooting:

• Incorrect Information? The "Show thinking" feature might reveal that the AI is pulling from an outdated or incorrect section of your knowledge source.

• Wrong Scenario Triggered? You might find that a customer's phrase is triggering a different scenario than you intended.

• No Answer? If the AI can't answer, this feature will help you see that no relevant knowledge or scenario was found, indicating a gap in your training.

What to Check For When Testing

Use this checklist to guide your testing process:

• ✅ Accuracy of Information: Ask specific questions from your knowledge source. Does the AI provide the correct details about your products, policies, or business hours?

• ✅ Scenario Handling: Use trigger phrases you set up in your scenarios. Does the AI follow the instructions correctly? Does it escalate to a human agent when it's supposed to?

• ✅ Personality and Tone: Does the AI's language match the personality you defined? Is it formal, friendly, witty, or exactly as you specified in the custom guidelines?

• ✅ Handling of Vague Questions: What happens when you ask an unclear question? Does the AI ask for clarification, or does it make a wrong guess?

• ✅ Escalation: Test the boundaries. Try to confuse the AI or ask questions you know it can't answer to ensure it escalates the chat when necessary.

Refining Your Training

Testing is an iterative process. It's normal to discover areas for improvement. Based on your test results, go back to the Train tab to:

• Update your knowledge source with clearer headings or more detailed information.

• Refine your scenario triggers and instructions.

• Tweak your AI's personality and custom rules.

By regularly testing and refining, you can build a highly effective AI Agent that serves your customers accurately and aligns perfectly with your brand.