A Russian neural network analyzed more than 100 hours of recorded conversationScience and technology00:15 | December 12, 2023

The company First Bit has successfully tested the use of a neural network to analyze and transcribe more than 100 hours of audio recordings of customer conversations. The use of artificial intelligence not only saved time in transcription, but also allowed the conversion of voice into valuable data, identifying stressful situations and problems in the company’s business processes.

Pyotr Ivanov, head of analysis and reporting at First Bit, emphasized that artificial intelligence significantly increases business efficiency. The project sparked interest and colleagues suggested a new use of the technology. One idea was to create an automatic call summary that would save managers time taking notes or listening to recordings.

Converting audio files to text also makes it possible to objectively evaluate the work of employees, including sales managers and technical support. This information can provide more objective data on how well an employee follows scripts and responds to customer feedback. In addition, the neural network can also determine the mood of the interlocutors, which is useful in identifying negative situations.

Data analysis also found that technical support employees were more likely to experience negative customer emotions, while sales employees were less likely to experience aggression. Research has shown that people often adapt their emotions to the person they are talking to, which affects their communication. This real-time analysis can help determine when employees need to take breaks to avoid transferring negativity to subsequent customers and prevent burnout.

The data also showed that some managers are uncomfortable telling customers prices, causing additional stress. This information can be used to improve employees’ skills.