How AI Makes It Easier to Verify a Customer Who Calls In

An opinion piece by the vice president of product management, research and engineering at Pindrop, a voice security and intelligence company

Amit Gupta

July 26, 2022

5 Min Read

AI is having a transformative impact on a wide variety of business processes. The ability to crunch data and make connections at superhuman speeds has opened up a vast array of opportunities.   

Customer experience is one prominent area where the technology’s merits are shining through. In particular, AI is delivering powerful results when it comes to authenticating customers – the process of making sure the customer accessing an account is the owner and not a fraudster.

It has long been difficult to perform authentication without degrading the customer experience. This is particularly true in the voice channel – such as a phone call − that has relied on manual processes that are both frustrating for legitimate customers, and increasingly easy for fraudsters to bypass.

Research from Pindrop and the Call Centre Management Association (CCMA) delved into the state of play for voice authentication, highlighting the greatest challenges and exploring how the speed and power of AI can help mitigate this longstanding pain point.

Manual authentication causes too much friction

The average consumer has grown accustomed to the frictionless authentication processes that are now commonplace for online channels. It is extremely easy to log in to an online platform, and even the most secure services such as banking only take a few moments. By contrast, caller verification is still largely handled by cumbersome manual processes that waste time and create frustration for the customer.

The voice channel has traditionally relied on Knowledge Based Authentication (KBA), such as question sets that typically involve a pre-set password, code, or personal details about the account holder. (For example, a customer’s date of birth.)

This process is problematic from the beginning as it means the first several minutes of the call are focused on identification, not the customer’s need. This has always been an issue but has become more obvious as voice processes have been outclassed by more dynamic and accessible online options.

According to the CCMA survey, 25% of respondents aged between 18 to 54 said accessing accounts through voice could be simpler. Worse, 27% of all respondents have gone as far as to stop doing business with a provider due to authentication issues. The problem is a bigger deal for younger people, with 42% of those under 34 having abandoned a business due to a poor authentication process.

Digitally savvy fraudsters are outpacing legacy authentication

The frustrating friction of traditional authentication over voice is only half the problem – crucially, the process is also failing to reliably protect customer accounts from fraudulent access.

KBAs have always had security shortcomings, but this has been greatly exacerbated by the huge volume of data breaches now occurring on a regular basis. More than 5.1 million records were stolen in 2021 alone.

With so much personal data in the hands of criminals, it has never been easier for fraudsters to access the information they need to pass KBA questions through dark web forums. Pindrop’s most recent Voice Intelligence & Security Report found that a complete set of financial log-in data and personal records can easily be bought for less than $60.

As the telephony channel typically lacks additional verification methods such as multifactor authentication (MFA), it is being targeted as a weak point to access financial accounts and other high-value profiles. Indeed, 34% of respondents have experienced unauthorized access attempts.

KBAs are also ineffective in the digital age due to the habit of sharing the same verification information across multiple services and platforms. The CCMA research found that younger people are more likely to use the same handful of passwords, with 33% of those aged 18 to 34 stating they use three or fewer passwords across all their accounts.

Indeed, while legitimate customers are finding authentication in voice channels ever more frustrating, fraudsters are increasingly able to pass the system with ease. Pindrop found that the success rate for fraudsters passing KBAs is as high as 92%, while legitimate callers might forget their answers as much as 46% of the time.

The prospect of a fraudster successfully accessing a financial account through voice is a grave one. If the verification system is not strong enough, the organization may need to take the hit and compensate the customer, as well as face additional financial and reputational damage from regulatory bodies, legal action, and lost business.

Speed, accuracy of AI and ML are key

The voice channel is a prime example of how the speed and accuracy of AI can be used to support and augment human personnel. AI and ML-powered analytics can rapidly crunch through large amounts of call data far faster than the most highly trained and experienced fraud detection experts. Data points such as the caller’s voice, device and call metadata can be quickly analyzed for signs of discrepancy that indicate a possible fraudster.

For example, analysis can recognise if the caller is spoofing their number, or if their device has been used before in other known fraud instances.

Even well-prepared fraudsters, armed with all the right KBA answers and a practiced social engineering playbook, can be identified in moments, helping stop fraud attempts in their tracks. The AI-powered analytics works unobtrusively in the background so there is no discernible impact on legitimate callers.

Indeed, these same capabilities can also be applied to authentication, creating a frictionless experience for genuine customers. The pattern recognition capabilities of AI solutions also support the creation of customer profiles. Once a caller has been authenticated by the system, they can be rapidly verified in future calls simply by their voice and metadata, potentially doing away with traditional verification processes entirely.

These processes provide powerful benefits to both security and customer experience, as well as valuable operational data that would otherwise go unseen. For example, First National Bank of Omaha was able to reduce one time password usage by 75% while simultaneously reducing overall average handling time by 30 seconds on every call. Account takeover recognition rose by 59%, and there was a 47% decrease in average loss.

Replacing traditional manual verification processes with passive, AI-powered authentication enables organizations to finally bring their voice channels in line with the more secure and frictionless experience offered online.

Embracing the speed and power of AI can empower businesses to create a smoother, more welcoming voice experience for their customers, while simultaneously shutting out fraudsters attempting to exploit the system.

Get the newsletter
From automation advancements to policy announcements, stay ahead of the curve with the bi-weekly AI Business newsletter.