TechTalk #3: AI Call Evaluation Using ChatGPT

Welcome to TechTalk, friends. We continue our conversation about Webitel's numerous capabilities for contact center automation. In the third edition, our CTO Vitalii Kovalyshyn shares his team’s experience on AI call evaluation using platform tools and services. Let’s get started!

Call transcription

Did you know the ability to transcribe conversation recordings into text was added to the platform services a few years ago? Moreover, we can transcribe all calls (not just by clicking) and also use our own local models. This allows us to convert all conversations into text at an acceptable price (at the cost of a simple server). By the way, if you are Webitel’s client with localization in our cloud, we invite you to a free trial!

Now, I would like to show you how it works based on the test case for our customer. So, we have a call transcribed into text:

Call visualization and transcription

Questionnaires for conversations

Our platform has an Audit module that allows you to evaluate calls and agents using various questionnaires. Let’s create a scorecard with questions you would like to receive automatic answers to with the help of AI. Here is an example of such a questionnaire:

A scorecard for call evaluation

You can set multiple answers and points for each question.

AI call evaluation

Now, we move on to the most interesting part: automatic call evaluation. In this case, we’ve connected ChatGPT, which will get call recordings, questions from the scorecard, and instructions for conducting the questionnaire. However, you can use any other language model. It all depends on your budget and tasks.

After evaluation, we will receive the result with the detailed report in the ‘Comment’ field and the explanation of why exactly such a rating was given (if necessary, this can be disabled):

Call evaluation using ChatGPT

Seems interesting? If you would also like to try AI call evaluation, please contact our managers, and we will schedule a test.