Test Conversations give you a controlled space to interact with your AI Agent before it handles real leads. You can replay scenarios from your Test Cases, start new conversations manually, and validate how the agent responds across different inquiry types and edge cases.
In this article:
Review conversations generated from Test Cases
Start a manual Test Conversation
Interact with the agent and validate its responses
Update guidance and regenerate a response without restarting
Overview
When something looks off in how your agent responds, such as a canned answer that misses the point, an incomplete follow-up, or a response that ignores an edge case, test conversations give you a controlled environment to reproduce and fix the behavior.
Every time you run Test Cases, the results show up here as individual conversations. Each one simulates a real interaction between a lead and your AI Agent, so you can review the full message flow, check responses, and confirm the logic holds up before it reaches anyone real.
Review conversations from test cases
After running your Test Cases, the results are displayed automatically in Test Conversations.
Open the Flex Admin Portal and go to AI Hub > Training > Test Conversations.
Select a conversation from the list and click View to open the full message thread.
Review the messages, agent responses, and overall conversation flow.
Start a manual test conversation
You can also test a specific scenario that isn't covered by your existing Test Cases. To do that:
Open the Flex Admin Portal and go to AI Hub > Training > Test Conversations.
Click Start New Conversation.
Select the conversation type:
Form submission — simulates a website inquiry
Email — simulates a direct email from a lead
Fill in the details:
Subject (for email scenarios)
Message — include the full inquiry or form submission details you want to test
Click Create Conversation.
Send messages in a test conversation
Once a conversation is open, you can exchange messages with the agent in real time. You can:
Send follow-up questions
Test how the agent handles incomplete or ambiguous inputs
Reproduce edge cases that have come up with real leads
To send messages in a test conversation:
Open the Flex Admin Portal and go to AI Hub > Training > Test Conversations.
Select a conversation from the list and click View to open the full message thread.
Enter a message in the text field and click Send Message.
Update guidance and regenerate a response
If a response looks wrong, you do not need to restart the conversation from scratch.
Identify the message that needs to change.
Go to your agent's Guidance or Training section (for example, FAQs, Inventory, or other configured sources).
Update the relevant content.
Return to the conversation and select the regenerate button (↻) on the message.
The agent generates a new response based on the updated configuration. The rest of the conversation stays intact.