Guided Conversation Categorization

Buzzeasy can automatically analyze and categorize conversations based on rules you create. This adds additional tools to qualify agent performance and is vital for conversations that were resolved by AI conversational bots without engaging a human agent.

You set up multi level categorization rules with detailed model prompts that Buzzeasy runs against the conversation data to be able to associate the best match.

For example:

  • Set “Technical Assistance” as first level category and create the following subcategories with AI model instructions:
    • “Software Issues”
      • Description: “The customer mentions problems related to software, such as errors, bugs, crashes, or difficulties with specific features.”
    • “Hardware Problems”
      • Description: “The conversation discusses malfunctioning devices, physical damage, or issues with hardware components.”

The AI model instruction prompts need to be unambiguous and straightforward for each subcategory so Buzzeasy could easily apply the correct category to your conversations.

Tip

Language based AI models can provide false information in a convincing manner. Please read our practical guide and best practices in writing AI prompt instructions in our Controlling Buzzeasy AI Processes for Conversational Bots document. Not all tips will be relevant for categorization purposes.

Categorization flow

The initial inbound message of the email conversations can be analyzed by a Categorization node in the workflow. Script designers link a categorization rule to this node, which selects the best matching category from that.

For voice and the various chat conversations, the relevant channel is configured to use the proper categorization rule. These are not happening in the workflow because chat and especially voice conversations need time to generate textual transcripts, which typically happen past the time the workflow ends on the conversation.

Categorization rule details

A rule consists of a name, any number of primary categories with any number of subcategories under these. All subcategories must contain a description, you can't save a rule with missing descriptions.

The Description field contains a proper AI model prompt describing the business context of the conversations that this category and subcategory can be associated with.

Tip

Be explicit in your AI prompts. Avoid ambiguity, don't create too similar subcategory descriptions.

Create a new categorization rule

  1. Navigate to the Administration > Categorization rules menu of the Buzzeasy portal. The list of existing rules is displayed.

  2. Click Create at the top of the list. The details modal opens up.

  3. Enter the name of the new rule.

  4. Add the name of the first category in the rule. You can add more categories, subcategories and descriptions in the modal.

  5. As soon as it becomes available, you can click create to save the new categorization rule with the items you added so far.

Update a categorization rule

  1. Navigate to the Administration > Categorization rules menu of the Buzzeasy portal. The list of existing rules is displayed.

  2. Click the list item you'd like to update. The details panel shows up next to the list blade.

  3. You can modify the name of the rule, of any primary and subcategories as well as their descriptions.

  4. Click the delete icon to remove any category or subcategory.

    Note that deleting a primary category will remove all subcategories with their descriptions.

  5. Click Save to store the changes.

Delete a categorization rule

  1. Navigate to the Administration > Categorization rules menu of the Buzzeasy portal. The list of existing rules is displayed.

  2. Click the list item you'd like to delete. The details panel shows up next to the list blade.

  3. Click Delete at the top of the details panel.