Teleperformance, the world's largest call centre operator, relies on artificial intelligence technology to “neutralize” the accents of English-speaking Indian customer service agents in real time.
A new feature known as accent translation is featured in call centres in India. There, employees provide support to some of Telepelforming's international clients. According to Bloomberg News.
Developed by Palo Alto-based startup Sanas, this AI-driven solution works alongside background noise cancellation technology to improve call quality.
Teleperformance, which serves leading companies such as Apple, Tiktok Parent Company Bytedance and Samsung, has invested $13 million in Sanas to secure exclusive rights to resell the technology.
“When Indian agents are on the line, it can be difficult to hear,” Telepelforming's assistant chief executive Thomas McKenbrock told Bloomberg News.
The technology “neutralizes the accent of Indian speakers with zero latency,” he explained.
According to Mackenbrock, the coordination promotes close connections between customers and agents, increases satisfaction, shortens call duration, and creates a “win-win for both parties.”
Teleperformance has not revealed which clients are currently using technology.
However, adoption of accent translation is part of a larger corporate strategy to expand AI integration.
The company is committed to investing up to $104 million in AI collaboration this year, as outlined in its latest revenue report.
The rapid advances in AI-powered chatbots have raised concerns about the future of human-run call centres.
Last year, Telepelforming's stock was a hit after Swedish fintech company Kralnabank announced that its Openai-powered AI assistant was doing the same job as 700 full-time employees.
In response, telepelforming emphasizes an approach of using AI to enhance it rather than replacing human employees.
As of the end of 2023, the company employs 490,000 people worldwide, using AI to train new recruits and monitor call quality.
The startup behind Accent Moderation technology, Sanas is at the forefront of AI innovation that blurs the boundaries between machine-generated enhancements and human interactions.
The software not only changes the audio pattern, but also eliminates destructive background noise such as office chatting, sirens and even cocks.
While this technology could improve efficiency, it raises concerns about its impact on global customer service hubs like the Philippines. It has built the industry's reputation based on a highly skilled English-speaking workforce.
Critics also argue that accent neutralization can reduce cultural identity and credibility in customer interactions.
Sanas argues that the technology was developed with the goal of “reducing accent-based discrimination.”
Currently available in Indian and Philippine accents, the software has been tailored to accommodate other regions, including Latin America, where Telepelforming has a critical workforce serving its US customers.
“AI will become ubiquitous. We already have it today,” Mackenbrock said.
“But the human element is extremely important to building connections, customer experiences, and branding awareness.”
Despite the skepticism surrounding the role of AI in call centres, telepelforming strategies appear to bear fruit.
The company projects revenue growth rates of 3% to 5% this year.
In the fourth quarter, it reported profit of $2.688 billion, reflecting a 12% increase from the previous year or a 4% increase on a pro-forma basis.
Meanwhile, Klarna CEO Sebastian Siemiatkowski recently focused on debates on the impact of AI on the workforce.
In a post on X, he said his company has a new appreciation for human employees, saying, “In the AI world, nothing is more valuable than humans.”
While Klarna intends to continue investing in AI-driven support, Siemiatkowski has pledged to use cost reductions to enhance the company's operational services.
Going forward, Teleperformance is expanding its AI industry footprint by providing essential services to high-tech companies developing AI models.
“Whether it's Openai or the world's AI model, we need human support for training,” Mackenbrock said, adding that this includes tasks such as creating data, testing and labeling.
