PolyAI named to Forbes AI 50 2023 list
⚠️ Unsupported Browser

Your browser is not supported.

The latest version of Safari, Chrome, Firefox, Internet Explorer or Microsoft Edge is required to use this website.

Click the button below to update and we look forward to seeing you soon.

Update now

Conversational ID&V vs Voice Biometrics: A New Approach to Authenticating Customers on Customer Service Calls

Image of Michael Chen
Michael Chen
11 Jan 2021 - 6 minutes read

Language is full of characters and words that sound frustratingly similar. Sorry, was that M or N? Did you say three, free, or tree? In normal conversation amongst friends, this is easily overcome but customers are less forgiving when they have an issue with a product or service.

Before customers even get the opportunity to unload the burden of their problems, they are usually waiting on hold longer than they would like, only to then constantly repeat themselves just to get past the caller verification step. Research by the Harvard Business School has shown that added effort can reduce customer loyalty over time.

Caller authentication is a prerequisite for introducing more personalised customer service and better self-service options. Countless hours are currently spent by customers waiting on hold for the next available agent to verify their identity. 

Similar phonetic sounds, combined with background noise, accents, and call quality make caller authentication a difficult task for even the most skilled call centre agents. It’s a repetitive and thankless part of their job: no-one ever left a positive review for being correctly authenticated, but customers won’t hesitate to leave a negative review if asked to repeat themselves too often.

Voice biometrics and its vulnerabilities

Voice biometrics is one way to get around the repetitive nature of caller authentication. These solutions use a complex understanding of vocal tracts, pitch, and nasal passages to create a ‘voiceprint’ using samples of each person’s voice. However, as is often the case, the business risk is not neutralised, it simply gets moved around. Risk is shifted from how customers safeguard their personal information to the security of the software.

Industry experts see voiceprints as weaker than facial or fingerprint biometrics. The BBC has already demonstrated that voice biometrics can be fooled by mimicry. The need for explicit consent from customers to use voice biometrics also creates new risks for customer experience, privacy and data security. HMRC (the UK’s taxation authority) was forced to delete 5 million voice recordings after its voice biometrics solution was found to have breached data protection rules

The vulnerabilities of voice biometrics

The vulnerabilities of voice biometrics

It’s an unfortunate reality that in our brave new world no software can be 100% secure. That’s why organisations need to use multiple authentication methods.

Seamless authentication in natural conversation

At PolyAI, we have pioneered a new approach for voice assistants to authenticate customer information in conversation: phonetic matching. We combined our deep expertise in speech recognition systems with a complex understanding of vocal tracts to help our voice assistants match alphanumeric information with unparalleled precision. 

Put simply, our voice assistants can authenticate account information such as a customer’s name, address and date of birth with up to 95% accuracy in the flow of conversation, before any additional training. 

Authentication using natural conversation means our voice assistants fit seamlessly into the existing security practices of call centres. There is no requirement for individual voice samples and no additional considerations around the privacy and security of those samples. This results in an effortless authentication experience that customers and organisations can get behind.

The benefits of PolyAI's phonetic matching

The benefits of PolyAI’s phonetic matching

A focus on phonetic sounds also enables our solution to accurately match information such as the name Tsung-Hsien Wen (our CTO), overcoming previous barriers in the customer experience of non-native speakers in different languages.

Phonetic matching creates a robust and scalable new authentication method that will complement existing account security measures. It will also remove a significant but repetitive and thankless workload from the call center, freeing agents up to focus on value-added problem solving and personalising the experience for each caller. Customers will be able to effortlessly verify their identity while waiting for the next available agent, or better yet, they’ll be able to proceed directly to resolve their issue with a voice assistant without any wait time at all. 

Accurate caller verification has long been a barrier to better personalisation and self-service for customers. With PolyAI’s phonetic matching approach, we believe organisations are closer than ever to robust voice assistants that can be trusted with this important task at scale.


Customer Experience | March 2018

The Age of Personal Assistants: More Machine Learning!

Customer Experience | January 2019

Moving from Engineering to Orchestrating Conversations

Get in Touch

Learn more about voice-based conversational AI, request a demo or find out how PolyAI can help.

Request Info