SoftBank AI to soften complaining customers’ voices for BPOs
Have you ever worked as a call center representative? One of their biggest challenges is the late work hours and the incessant customer complaints.
It’s becoming a huge problem in the Land of the Rising Sun as their agents suffer from “customer harassment.” That is why internet services company Softbank is building Emotion Canceling.
READ: How to cancel online subscriptions on a smartphone
Article continues after this advertisementThis Softbank AI system softens the loud, shrill voices of irate customers to address these issues. Project leader Toshiyuki Nakatani says it will be a “mental shield” for CSRs.
How does the Softbank AI work?
Emotion Canceling is an audio processing technology that uses artificial intelligence to change a caller’s voice. It maintains the person’s words but makes their pitch and inflection gentler.
The Softbank AI learns various expressions like yelling and sarcasm to improve voice conversion. The current version has three male and female voices like OpenAI’s GPT-4o.
Article continues after this advertisementCall center agents may choose voices that suit them best. Softbank told the Japan-based publication Mainichi that Toshiyuki Nakatani started the project.
The project leader took inspiration from a news report on customer harassment. He thought, “If the yelling voice was that of Medama-Oyaji (a character in the Japanese anime “GeGeGe no Kitaro”), it would be less frightening.”
Most call centers encourage agents to handle calls in one sitting and avoid passing them to other agents. Unfortunately, some customers take advantage of these interactions to hurl social and personal grievances as complaints.
Japan’s Ministry of Health, Labor, and Welfare revealed that roughly 30% of companies received customer harassment complaints from employees in the past three years.
“I want phone operators to have a ‘mental shield’ to prevent their hearts from being worn out,” Nakatani explained.
He will expand the Softbank AI’s capabilities, such as noise reduction and additional voices. Then, the company will release the product for purchase in 2025.