Ender Turing's AI-powered quality evaluation tool is designed to help businesses automate and streamline the quality assurance process for contact center conversations (calls, chats, support tickets, etc).

By leveraging Generative AI, Ender Turing offers a scalable solution that automatically detects topics for post-conversation analytics and helps identify and categorize the key issues or themes discussed during customer interactions. This allows contact centers to gain insights into common problems, identify trends and products, and optimize agent training and business processes.

Generative AI for conversation analytics involves leveraging LLMs (Large Language Models) to analyze and assess customer-agent interactions. The AI leverages AI models to automatically generate and categorize topics based on the call's content. The AI identifies and categorizes key themes without predefined rules, adapting to new topics over time. It’s highly flexible and works well in dynamic environments where new trends emerge frequently. Generative categorization is a fast and efficient way to categorize new or evolving topics but can't be used to categorize through historical conversation.

This guide will walk you through how to maximize Ender Turing's Automated Categorisation for your business.

Using a generative AI model for conversation categorization begins by anonymizing the customer-agent interactions to remove personal data, ensuring compliance with privacy regulations. The anonymized conversation is then sent to the AI with a predefined list of Categories (Topics). The AI processes the conversation and suggests the closest matched categories for each conversation. Managers and other teams (marketing, product) then review these results to help refine the performance and decision-making of their operations.

To use AutoQA in Ender Turing, you need to create:

In most cases, your organization already has categories (topics) from previous manual categorization processes made by Agents after the conversation. You can use these topics.

! Important - If you have more than 50 topics, please use any generative chat tool (ChatGPT, Anthropic, Gemini, etc.) to reduce the number of topics to 50. Using more than 50 topics to categorize a single conversation may lead to picking up a very close but incorrect topic. Make sure Topics are different enough and do not overlap in meaning.

Please use any generative chat tool (ChatGPT, Anthropic, Gemini, etc.) to create such a list using prompts like:

Or more specifically, for example, for a financial company:

The process of creating a Category is described here.

While creating a Category, consider adding a detailed description (explanation) of a Category if a name is not self-explanatory enough. This description will be passed on to AI as instructions for categorization with the Category name. The more accurate the description (explanation) you provide, the more accurate the automated categorization can be.

Read our detailed guide on improving category description (explanation).

Here is an example:

Category name: Credit Card Reissue

Category description: A customer asking about a credit card reissue due to loss, theft, or expiration.

To start automated categorization, you need to provide rules on which conversations to categorize and which to skip. You can also define multiple rules, where, for example, calls/tickets with Issues will be categorized using Procuct-related categories, and all others will be categorized using the main (high-level) list of Categories.

Such rules are created using Enders. The process of creating Ender is described here.

! Important—We recommend using less than 50 categories for single conversation categorization. You can create multiple Enders with corresponding rules to use different category lists for various situations, e.g., categorizing Products for issue-related conversations in addition to the main topic.

Below, you can see an example of defined rules:

Which conversations to categorize: conversations that have more than 50 words inside

Which Categories to use: All Generative Categories

The logic above is straightforward: use all Gen AI categories created at step 2 in categorization.

Select specific categories if you need to customize logic and use different categories (topics) for different conversations (based on rules defined in filters).

Below, you can see an example of defined rules:

Which conversations to categorize: conversations with the sales process inside

Which Categories to use: Use only a subset of categories

And after saving, It'll look like this.

By following these practices, you can ensure the categorization tool delivers accurate, actionable insights and drives efficiency in customer service operations.