Reviewing conversation logs for advanced AI agents

What's my plan?

Add-on

AI agents - Advanced

Conversation logs allow you to review all messages that have been exchanged between an AI agent and end users during automated conversations. From here, you can review details for specific messages and whole conversations as part of your ongoing AI agent monitoring and analysis.

This article contains the following topics:

• Viewing the conversation logs for an advanced AI agent
• Filtering the list of conversations
• Reviewing details for a whole conversation
• Reviewing details for a single message

Related articles:

• Using labels to tag conversation content for advanced AI agents

Viewing the conversation logs for an advanced AI agent

You can view the conversation logs for an advanced AI agent to see the conversations they’ve had with users and drill deeper into specific responses.

To view the conversation logs for an advanced AI agent

1. In AI agents - Advanced, use the AI agent drop-down field to select the AI agent you want to view conversation logs for.
2. In the left sidebar, select Conversation logs.

   A list of conversations appears for the default time frame (the past 7 days). If you don’t see any conversations, adjust the time frame.

3. Toggle the Status / Custom resolutions setting at the bottom to switch between the following information:
   • Status: Shows one of the following statuses for each conversation:
     • Agent escalation: The conversation was successfully transferred to a human agent.
     • AI agent handled: The conversation had a meaningful intent, did not attempt escalation, and ended without the AI agent misunderstanding a message.
     • Custom escalation: The conversation had a custom escalation attempt.
     • Email escalation: The conversation was successfully emailed to the support team.
     • Escalation failed: The conversation had an unsuccessful escalation attempt.
   • Custom resolutions: Shows one of the following custom resolutions for each conversation:
     • Resolved: A meaningful custom resolution was provided, and there are no further questions.
     • Informed: Instructions or guidance was provided.
     • Escalated to agent: The conversation was successfully transferred to a human agent.
     • Escalated via email: The conversation was successfully converted into an email ticket for the support team.
     • Unresolved: The AI agent didn’t answer or resolve the issue.
     • Escalation failed: The conversation wasn’t escalated successfully because no agent was available, or there was a technical error.
     • Undefined: No custom resolution is associated with the conversation.

Filtering the list of conversations

You can filter the list of conversations by time frame or other filters, such as language, status, labels, and more.

To filter the list of conversations by time frame

1. In the conversation logs, in the upper-right corner, click Time frame.
2. Using one of the following methods, select the dates you want to view conversations for:
   • On the right, select one of the predefined time frames: Today, Yesterday, Last 7 days, Last 30 days, This month, Last month
   • Select specific beginning and end dates on the calendar.

     

To filter the list of conversations by other filters

1. In the conversation logs, in the upper-left corner, click Add filter.

   

2. Select and configure one or more of the following filters:

   Filter	Configuration details
   Message text	In the Message text field, enter a specific word or phrase you want to search for within the conversations.
   Language	Select the languages you want to see conversations for.
   Conversation type	Select one or more of the following options: Conversations with messages not understood: Conversations with at least one message that was not understood below the threshold. Actions applied: Conversations with applied actions. Display only test conversations: Show only conversations created through the test widget. Hide all test conversations: Hide conversations created through the test widget. Conversations with interrupted dialogues: Conversations where visitors interrupted the conversation (for example, by triggering another intent or loops). Conversations with drop-offs: Conversations where users left before completing a dialogue. Hide in-progress sessions: Hide conversations with sessions still within their duration threshold. Show only automated resolution conversations: Show only AI agent-handled conversations that have passed verification. Hide automated resolution conversations: Hide all conversations that have passed verification. Conversations with knowledge answers: Conversations in which the AI agent responded with a generated knowledge source–based answer. This excludes UltimateGPT responses. Conversations without knowledge answers: Conversations in which none of the AI agent messages were created with a generated knowledge source–based answer. This excludes UltimateGPT responses. Conversations with procedures used: Conversations in which the AI agent responded with a generated message based on a procedure at least once. Conversations without procedures used: Conversations in which the AI agent has not responded with a generated message based on a procedure.
   Conversation status	Select one or more of the following statuses: Agent escalation: The conversation was successfully transferred to a human agent. AI agent handled: The conversation had a meaningful intent, did not attempt escalation, and ended without the AI agent misunderstanding a message. Custom escalation: The conversation had a custom escalation attempt. Email escalation: The conversation was successfully emailed to the support team. Escalation failed: The conversation had an unsuccessful escalation attempt. No status
   BSAT	Select one or more of the last BSAT rating responses, or select Not now.
   Custom resolutions	Select one or more of the following custom resolution states: Resolved: A meaningful custom resolution was provided, and there are no further questions. Informed: Instructions or guidance was provided. Escalated to agent: The conversation was successfully transferred to a human agent. Escalated via email: The conversation was successfully converted into an email ticket for the support team. Unresolved: The AI agent didn’t answer or resolve the issue. Escalation failed: The conversation wasn’t escalated successfully because no agent was available, or there was a technical error. Undefined: No custom resolution is associated with the conversation.
   Labels	Select a specific label. See Using labels to tag conversation content for advanced AI agents.
   Intents	Select a specific intent. See About intents in advanced AI agents.
   Duration	Select an operator and a number of seconds to determine the duration of the conversations you want to see.
   Platform conversation ID	In the Platform conversation ID field, enter an ID from your CRM to see a specific conversation. This is useful when comparing logs between your CRM and your AI agent.
   UltimateGPT	Select one or more of the following options: UltimateGPT used UltimateGPT not used Feedback shared Feedback not shared
   UltimateGPT status	Select one or more of the following options: Response generated Not understood Escalation needed An error occurred

3. Click Apply.

Reviewing details for a whole conversation

You can review the details for a conversation to get more information about the overall interaction between the user and the AI agent.

To review details for a whole conversation

1. In the conversation logs, click the conversation you want to view.

   The conversation appears on the right half of the page, and includes any messages sent as well as any actions or events that were applied.

2. In the upper-right corner, click Details.
3. Review the following details in the Conversation overview pane:
   • Date and time: The date and time the conversation took place.
   • Duration: How long the conversation lasted.
   • Test conversation: Whether the conversation was a test (as determined by whether it was created using the Test Bot, Test Dialogue, or Test Branch options).
   • Platform conversation ID: A unique identifier for the conversation in the CRM your AI agent is integrated with. This is used for finding the same conversation in your CRM. Not supported for Zendesk messaging integrations.
   • Labels: Which labels were added to the conversation. You can add additional labels by clicking Add.
   • Presumed use case: (Appears only for zero-training AI agents.) The use case that best fits the conversation. An AI-generated explanation describes why the use case was selected.
   • Automated resolution: Whether the conversation consumed an automated resolution.
4. Click Session data to show the following additional information:
   • The actions and events that occurred throughout the conversation.
   • The active language set.
   • The last detected language.

     

Reviewing details for a single message

You can review the details for a single message within a conversation to get more information about why the message was generated (in the case of an AI agent message) or how the AI agent interpreted the message (in the case of a user message).

To review details for a single message

1. In the conversation logs, click the conversation you want to view.

   The conversation appears on the right half of the page, and includes any messages sent as well as any actions or events that were applied.

2. Hover over the message you want to see more information about and click the View details icon () that appears.

   The icon appears to the left of AI agent messages and to the right of user messages.

   

3. Review the message details that appear in the pane on the right.
   • For user messages, the pane is called Message overview and shows the following information:
     • Message: The exact text of the user’s message.
     • Presumed use case: (Appears only for zero-training and agentic AI agents.) The use case that best fits the conversation. An AI-generated explanation describes why the use case was selected.

       

   • For AI agent messages associated with a dialogue, the pane is called Reply overview and shows the following details:
     • Details
       • Used reply: Dialogue the message came from.
       • Reply type: Type of reply.
       • Intent: Use case the dialogue is associated with.
     • Edit reply
       • Edit dialogue: Opens the dialogue builder for the dialogue that the message came from.

         

   • For AI agent messages associated with a generative procedure, the pane is called AI agent message overview and shows the following details:
     • Details
       • Use case: Which use case the generative procedure is associated with.
       • Plan: The AI agent’s reasoning in determining how to resolve the user’s issue.
       • Procedure step: Click to view the specific step within a generative procedure that produced the message.
       • Response before customization: The response text that the AI agent generated before the AI agent persona or any instructions were applied.
     • Active instructions: Any instructions you’ve created that are currently active. These instructions affect the AI agent’s ultimate response to the user.
     • Personalization
       • AI agent name: Name of the AI agent.
       • Company name: Name of your company.
       • Tone: Tone of voice you selected for the AI agent.
       • Answer length: Answer length you selected for the AI agent.