Conversation Logs are an essential part of the Ultimate platform. This is where you can review all messages that have been imported during your onboarding, this is where you will perform your Content Coverage Analysis, and then review which conversations the Virtual Agent Handles post-launch for ongoing monitoring and analysis.
Understanding Conversation Logs
Conversation Logs are a powerful tool that enables you to perform a lot of critical actions. In this article, you will review all of the sections within the Conversation Logs and what they can be used for.
On the top-right hand-side, we have the date picker, which allows you to choose the date range of conversations you would like to review. By default this is set to the last 7 days, however, you can use the buttons to make quick changes or manually choose the dates you want to select.
Filters are available for you to drill into a specific type of conversation based on different criteria. The filter options are detailed in the table below.
|Message Text||Search for a specific word or phrase|
|Language||Use the checkboxes to filter for a language(s) that you would like to review|
You can filter on 2 options:
Conversation not understood messages - conversations that have at least one message that was not understood
Actions applied - conversations with actions applied
|Resolutions||Based on the type of automation your Virtual Agent is supporting you will see different resolutions for Ticket and Chat, but these are the same resolutions you will see on the Analytics Dashboard.|
|Labels||Labels are useful for a lot of circumstances, for quick categorization. Find out all about them here|
|Intents||Review all messages that went through the flow of one of your intents.|
|Duration||How long was the user in the chat with the virtual agent?|
|Platform conversation id||Search for a specific conversation by the ID. You may use this when it comes to comparing logs between your CRM and Ultimate.|
From the conversation table, you have the possibility to very quickly assign a label by hovering over the respective row and the label icon will appear. This is especially helpful during the content coverage analysis phase and ongoing maintenance to mark which have been reviewed.
Within the table, you will also which resolution state the message went to such as bot handled or escalated to quickly see how the conversation is categorized.
Upon clicking a message row, the conversation transcript will show and as such you will be able to dive into the conversation details using the details button on the top right of the transcript.
This will show:
On an individual message level, you can click the settings icon to open the Expression Overview drawer.
Within the Expression Overview, you can see specific details about the AI Intent Recognition and Confidence Thresholds
First, you see the message/expression you opened, what Intent it matched with, which confidence percentage, and then which intent it is trained to.
The status will indicate whether that expression is trained to an Intent or not.
In the training section, you are able to manage the Intent training, by reviewing which Intent it is trained to, and the possibility to review which other Intent it could be matched to instead under predicted Intents, if you find an expression is being matched to the wrong Intent. You can search all Intents in the Search Intents tab. If you need to add a new Intent you can do this under the Create Intent tab.