Skip to main content

Topics: Purpose, Types, and Access

Topics group conversations into recurring semantic categories. They help you analyze, filter, and compare conversations based on context.

What is a Topic?

A Topic is a category used to identify the key issue, subject, product, or process discussed in a conversation. Examples include “Refund request,” “ATM issue,” or “Verification completed.”

A Topic can be used more broadly than simply identifying the subject of a conversation. For example, a single Topic can include all conversations related to a specific product.

What information does a Topic contain?

  • a name;

  • a required description;

  • a categorization type: Use Gen AI, Use Rules, or Human Only;

  • a priority position used when a conversation matches multiple Topics.

Where are Topics used?

  • in filters on the conversations page;

  • in Dashboards as separate groups;

  • in Charts to create samples;

  • in Funnels as separate steps;

  • in selection lists during manual conversation review.

How to open the Topics list

  1. Open Analytics configuration from the main menu.

  2. Select Topics.

The list of active Topics opens by default.

Topics list in the Topics section

Who can manage Topics?

The Topics section is displayed only to users with permission to manage analytics.

Users with this permission can:

  • create and edit Topics;

  • change their order;

  • archive, restore, and delete Topics;

  • import and export Topics.

Users without the appropriate permission cannot see the Topics section or change its settings. Active Topics continue to be applied to conversations and remain available in filters and analytics.

Categorization types

Use Gen AI

Generative AI determines whether a conversation matches a Topic based on its description. This type is suitable for new conversations that are difficult to describe using exact words or rules.

The Use Gen AI option is available only when a generative AI provider has been configured for the workspace.

Use Gen AI applies to new conversations and is not used to categorize historical conversations.

Use Rules

The Topic is identified using predefined conditions. Rules can use tags, words, phrases, metadata, and the logical operators AND and OR. This type is suitable for stable and recurring signals.

Use Rules can be applied to new and historical conversations. After rules are created or changed, the system automatically rematches conversations from the last 30 days only; older conversations must be matched manually.

Human Only

The Topic is not assigned automatically. A user selects it manually while reviewing a conversation. This type is suitable for internal labels, training samples, and decisions that must be made by a person.

Did this answer your question?