Conversations

The Conversations tab in the Insights section of Agent Builder allows you to review entire conversations your Mpower Agent has with contacts. From here, you can make adjustments based on how your Mpower Agent handled the situation or how well it understood the contact's intents The meaning or purpose behind what a contact says/types; what the contact wants to communicate or accomplish.. This is how you make your Mpower Agent smarter and more helpful over time.

The right side of the Conversations tab displays the following information related to the conversation:

• Customer Information: Information about the contact.
• Conversation Information: Information about the interaction, such as how long ago the conversation was and whether a Agent Builder administrator has reviewed it.
• Tags: All tags that have already been added to the conversation.

The conversations visible on this page include test conversations you have had while talking to the Mpower Agent within Agent Builder.

Review a Conversation

1. In CXone Mpower, click the app selector and select Agent Builder.

2. Click the Mpower Agent you want to work with.
3. Click Insights in the left icon menu.
4. On the Conversations tab, locate the conversation you want to review. You can search using natural language or write queries in a query language for more detailed results.
5. Click the conversation in the list on the left side of the Conversations tab. The messages it contains appear in the center of the page. The right side of the page displays information about the contact and the conversation.

6. For each message from the contact The person interacting with an agent, IVR, or bot in your contact center. or the Mpower Agent, you can view helpful information.

7. On the right side of the page, click Add Tag if you want to tag this conversation. If the tag you want doesn't exist, enter the name and click Create Tag.

8. To use the entire conversation as training data to improve the Mpower Agent's future responses, click Convert to Story . If you convert the conversation to a story, you need to add intents and modify the story to make it good training data.

   Converting a conversation to a story can create conflicts in your Mpower Agent. For example, if the conversation contains intents that are already used in other stories, it can cause intent confusion. After converting a conversation, review the new story and test it to ensure it doesn't cause a problem in your Mpower Agent's configuration. You can find the new story in the From Learning folder on the Stories tab.

9. When you're finished reviewing the conversation, click Mark Reviewed .

Conversation Information

On the Conversation tab, you can view the following information about a conversation.

Type	Details
Contact Message	You can see the exact message the contact sent. Under each is the intent the Mpower Agent selected. Next to the intent is a number indicating the Mpower Agent's confidence in its intent prediction. Click the intent to expand a list of all intents and the Mpower Agent's confidence level for each one, for that message. Actions you can take to improve your Mpower Agent:  If the intent is correct but the confidence score is low: Add more training examples or stories that use this intent to increase the Mpower Agent's confidence. You can convert conversations to stories, if they are good training examples. If the intent is incorrect: Add more training data to the intent the Mpower Agent should have selected. This will increase the Mpower Agent's ability to choose that intent.
Mpower Agent Response	You can see the Mpower Agent's responses to each contact message. The type of Mpower Agent action the Mpower Agent took is displayed under each response. Click the action drop-down under the message to see information about that response. Included in this information is the model ID. This can be helpful if you need to research a problem with the Mpower Agent. That way you can look at that model version, if necessary. The conversation includes Mpower Agent responses such as when fallback is triggered. Beneath this response is the line Threshold: 0.7. This is the confidence level the Mpower Agent must have to choose an intent. You can adjust the threshold if it's too high or too low.
Customer Information	The right side of the page displays the following information: The contact's name. The channel the conversation took place on. If the conversation is a test conversation that took place within Agent Builder, the channel is Talk to your bot.
Conversation Information	The right side of the page displays the following information:  What number conversation this is between the contact and the Mpower Agent. How long ago the conversation took place. Hover for the exact date and time. The number of messages in the conversation. Whether the conversation has been marked as reviewed.
Tags	Any tags previously assigned to this conversation.