There’s Untapped Value in Voice Biometrics

While voice biometrics can serve as a valuable addition to any company’s contact center technology stack, many service organizations are still wary of the tools, according to a recent report from Opus Research.

Organizations that use voice biometrics see improved customer experiences, thanks to increased ease of use, reduced costs per call, and stronger overall security, according to Ravin Sanjith, intelligent authentication program director at Opus and author of the report. Such applications also tend to “strengthen [agents’] relationships with callers,” because they no longer have to interrogate them with security questions, Sanjith says. Instead, voice biometrics compares current utterances with stored voiceprints and often runs in the background without the customer ever knowing that such activity is taking place.

“In the case of human-assisted calls, the dialogue with the agent becomes a lot more cordial and focused on service fulfillment rather than security,” he adds, thus eliminating a common frustration for customers when they have to repeat details they’ve already provided. 

Voice biometric technology has already proven its worth in the financial sector, with companies like Barclays, Wells Fargo, Citi, and Vanguard already boasting successful deployments, according to Sanjith, who notes that the same applies to telecommunications companies Vodafone and Swisscom. Millions of people have voluntarily created voiceprints for these companies, which also saw investment returns within their forecasted periods. 

However, many company leaders—and customers—still haven’t overcome fears regarding the use of the technology for contact center authentication. 

“For many years, voice biometrics had been associated with speech recognition, which had a somewhat jaded performance history,” before the advent of Siri, Cortana, Alexa, and Google Voice, Sanjith says. There is a difference, he adds: Voice biometrics serves to identify speakers, not recognize their speech, and people have had trouble distinguishing between the two.

Many of the holdouts are also still afraid that recorded voice audio files can be collected and used for identity theft. This concern is not totally unfounded, but it has been exacerbated by media coverage that fails to illustrate the full story, according to Sanjith. 

Another concern about voice biometrics is that it operates according to probability, unlike PINs, passwords, or tokens, which require 100 percent matches. That, too, is a legitimate concern. Even today, systems are still not totally foolproof: False rejects, where the system denies access to the right people, and false accepts, where the system grants admission to the wrong people, still happen, though they are far less likely than when the technology was first introduced. 

Biometric engines have been operating with accuracy rates of greater than 97 percent since 2014, Sanjith maintains.

Adoption rates are expected to grow in coming years. Opus has already charted hundreds of implementations and expects adoption to grow at a 15 percent compound annual rate through 2022. 

However, adoption is still not where it could be, despite the fact that the technology “has been improving in quantum leaps,” thanks to advancements in deep neural networks and machine learning, Sanjith says. 

Voice biometrics adoption will also benefit from an expanding base of use cases, including text-dependent self-service, text-independent agent-assisted service, and fraud watch lists, according to Sanjith. 

The technology’s weaknesses, he says, can be easily overcome by multi-level solution architectures, proactive internal awareness, and efforts to educate customers. Senior executives must also be able to distinguish between “high-profile” and “high-impact” risks. 

Then, too, solutions must be designed within the context of the overall customer fulfilment use case and not as “point solutions,” Sanjith argues, and should be aligned with overall know-your-customer identification and verification processes. 

Companies will also need to pair the technology with their interactive voice response, audio streaming, and CRM systems. Organizational planning must have an owner who will embrace the solution once it moves out of project mode, with regular reporting and optimization efforts, Sanjith says.