Our client is a Southern African organization that provides call center services for the banking sector.

*The name of our client is not disclosed due to the privacy and confidentiality agreement.

The challenge

Part of our client’s call center services is customer care, onboarding, and technical support. With the goal of increasing customer satisfaction, while still fighting the rising fraud trend, our client is facing several challenges. 

Firstly, when verifying the caller’s identity, the call center is relying on knowledge-based authentication tools (KBA). However, skilled professional fraudsters find ways of getting the needed information from public records or social networks, social engineering techniques, or the dark web’s identities market. 

Secondly, with any additional security measures taken by the company, the customer’s experience is affected and the average handle time is increasing, while fraudsters find new ways of deceiving.

The cost trade-off of staffing performing additional checks on each individual caller is prohibitive due to the large volume of clients attended. The impact of frauds is a burden for the call center as they learn that they were fraud only after the act occurred. This is why the call center needs to act as quickly as possible to cut losses and protect the brand’s reputation.

The solution

In collaboration with Polygon, the call center implemented a text-independent voice enrollment system that can detect and prevent fraud and verify the caller’s true identity within seconds.

Polygon’s best-in-class voice recognition technology can capture tens of thousands of voice and speech unique characteristics enabling a highly secure and fast speaker identification and verification.

1. Voice enrollment

The caller’s voice is enrolled by creating a voice template of 20 seconds of speech (using a passphrase or a string of letters and numbers).

2. Live call

Voice recognition is seamlessly performed during a live call using 10-20 seconds of normal conversation or from a pre-recorded audio sample.

3. Biometric data analysis

In the background, the AI algorithm is analyzing unique vocal qualities (duration, intensity, dynamics, pitch).

4. Authorization

The technology is comparing the voice with the enrolled sample and deliver the result.

The text-independent voice recognition determines the identity based on conversational voice, regardless of what the person is saying. The voice recognition technology is highly accurate as it supports different languages/alphabets and accents, it can check age and gender, and works even with noisy data.

During the first step of the process, a caller’s voice is recorded to create a voice template that will be used by the technology to record the voice’s unique print. 

Next, when the same user calls the customer service team, his identity is verified within the first 10 seconds of speech, without the need for the knowledge-based question verification process.

Meanwhile, the speech recognition engine is analyzing the voice’s unique characteristics, matching them with the enrolled voice sample.

Lastly, if the match is successful, the technology authorizes the voice as genuine. 

In order to protect the company from fraud, Polygon’s software can handle both streaming real-time audio and batch audio file processing (matching known fraudsters, scanning for the same voice associated with different customer IDs, verifying age and gender conform to the expected results, and confirming all customer order calls match the voice on the customer confirmation calls).

The results

With the help of Polygon’s voice recognition software, the call center improved considerably its success metrics:

Callers are identified with maximum security in a matter of seconds, allowing the agent to focus on meeting the client’s needs and expectations. 

More than this, fraud losses were reduced by blocking all frauds before they even happened, making it possible for the company to quickly enhance its fraud prevention methods.