Clear communication is crucial in today’s global workforce, but for some it is easier than for others.
On Tuesday, Krisp, an Audio AI software developer, introduced an accent conversion tool to split language barriers by changing how the voice of a person comes to listeners.
Although the technology promises inclusiveness, it also evokes concern about cultural prejudices, cultural identity and what it means to ‘sound professionally’.
Krisp’s AI Accent Converter wants to enable more inclusive and just recruitment practices by eliminating perception bias and ensuring that the accent of a person has no influence on how professional or reliable customers perceive them.
According to co-founder of Krisp Arto Minasyan, a recent study revealed an important perception gap between customers and contact centers.
“In our studies, 79% of American customers said that they ask non-Native English-speaking agents to repeat themselves at least once per call, but only 33% of the contact centers believe this is happening,” Minasyan said Decrypt. “This gap leads to lower satisfaction and frustration for both sides of the meeting.”
Minasyan noted that companies reject candidates with strong accents, disqualify on average 64% of agents or have to invest in expensive accent training.
From that moment until now
Krisp was launched in 2017 by Minasyan and Davit Baghdasaryan and is an AI-based software company for audio processing located in Berkeley, California. It specializes in real-time noise and speech oppression technology.
Minasyan explained that the tool does not make a distinction between accents, but converts speech instead to agree with a specific goal accent. During the conversion process, Krisp records characteristics of the user’s voice and records them in the final output.
“This approach helps to maintain a more natural sounding voice and retains the emotional tone of the original input,” he said.
Although users can switch the AI -accent converter on and off in real time, Minasyan said that the option to adjust the conversion level is currently not available.
Although AI -accent tools improve employment opportunities, they also have risks of abuse.
The FBI led these concerns to warn in May 2024 that scammers with AI-driven speech software use to perform phishing-scam.
In the run-up to the 2024 presidential election, the FCC Robocalls prohibited the voices of AI-generated candidates after a depth of the then President Joe Biden was used to trying to cheat voters in New Hampshire so as not to vote.
Asked for the potential for abuse by cyber criminals, recognized Minasyan the risks and noted that Krisp’s tools, such as mobile phones, laptops or internet, can be used despite their intended goal.
“Although we acknowledge that abuse is a concern, it is not feasible to fully check how the tool is used without jeopardizing users’ privacy,” he said.
Nevertheless, there is hope that the technology will mainly be used forever instead of for other purposes.
“We did not immediately consult with socioling lists, but we tested the technology in real-life settings with customers before we have made it available on a large scale,” Minasyan said.
The founder pointed to various internal statistics, including an improvement of 99% in a Net Promoter score, an increase in the sales conversion rates with more than 26% and an incentive of 12.2% in the score of the Agent Promoter, with “trust with a higher agent”.
“Seeing these improvements to both the user and the customer has strengthened our confidence that the technology does not introduce great cultural or business disadvantages,” he said.
Edited by Sebastian Sinclair
Generally intelligent Newsletter
A weekly AI trip told by Gen, a generative AI model.
