Teleperformance uses AI to ‘neutralize’ Indian accents of agents

By Benoit Berthelot

Teleperformance SE, the largest call-center operator in the world, is rolling out an artificial intelligence system that softens English-speaking Indian workers’ accents in real time in a move the company claims will make them more understandable.

The technology, called accent translation, coupled with background noise cancellation, is being deployed in call centers in India, where workers provide customer support to some of Teleperformance’s international clients. Teleperformance provides outsourced customer support and content moderation to global companies including Apple Inc., ByteDance Ltd.’s TikTok and Samsung Electronics Co Ltd.

“When you have an Indian agent on the line, sometimes it’s hard to hear, to understand,” Deputy-Chief Executive Officer Thomas Mackenbrock said in an interview with Bloomberg. The technology can “neutralize the accent of the Indian speaker with zero latency,” he said. This “creates more intimacy, increases the customer satisfaction, and reduces the average handling time: it is a win-win for both parties.”

The company declined to disclose which clients are using the technology.

The roll-out of accent translation is part of a bigger push by Teleperformance to invest up to €100 million ($104 million) in AI partnerships this year, the company said in an earnings statement on Thursday. The technology was developed by Palo Alto-based startup Sanas, in which Teleperformance invested $13 million earlier this year. Under the terms of the deal, Teleperformance becomes the exclusive reseller of Sanas’ technology to its clients.

The rise of AI chatbots has led to investor jitters about the sustainability of the human call-center model. Last year, Teleperformance shares fell to their lowest since late 2016 after Swedish fintech company Klarna Bank AB said its AI assistant, powered by OpenAI, was doing the equivalent work of 700 full-time agents. Teleperformance has sought to allay shareholder concerns by using AI to enhance rather than replace employees, which totaled 490,000 as of the end of 2023.

The French company is using AI across the business, including using AI co-pilots to coach new employees and transcribing calls for quality control.

Sanas, the firm that Teleperformance is using to offer this accent-softening service to clients, is one of the AI upstarts blurring the line between where generative AI tech starts and the human ends. The firm’s software also eliminates background noise — like crowing roosters, ambulance sirens and office chatter. That might sound like an aid to agents, rather than a tool that might displace them. But it carries a risk for customer service bases such as the Philippines that built market-leading positions by cultivating a preponderance of high-quality English speakers. The tools may also be perceived as taking away from workers’ cultural identities and the authenticity of conversations.

Sanas said it developed the technology with a goal of “reducing accent-based discrimination,” according to its website. The technology is available for Indian and Filipino accents and is being tuned for other regions, including Latin America, where Teleperformance has many workers supporting US-based customers, Mackenbrock said.

“AI will be ubiquitous, it is already today,” Mackenbrock said. “But in order to build connections, customer experience, branding awareness, the human element will be incredibly important.”

Teleperformance’s approach appears to be paying off: Teleperformance is forecasting 3 per cent to 5 per cent revenue growth this year, it said in the earnings statement. Fourth-quarter revenue was €2.68 billion, up 12 per cent from last year or 4 per cent on a pro forma basis.

Earlier this month, Klarna’s CEO Sebastian Siemiatkowski posted on X that Klarna “just had an epiphany,” saying that “in a world of AI nothing will be as valuable as humans.” Klarna would continue to invest in its AI support, but use cost savings to make sure that “the human service part of Klarna becomes even better,” he added.

Teleperformance is also making a push to provide services to the tech giants building AI models, Mackenbrock said.

“Whether it is OpenAI, or any AI model in the world, it needs human support for training,” he said. This includes creating, testing and labeling data.