Join us at PolyAI VOX 2023
⚠️ 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

Automating Identity Verification in the Contact Center: What You Should Know

Image of Georgios Spithourakis
Georgios Spithourakis
1 Jun 2022 - 8 minutes read

Contents

    Many contact centers require identity verification steps in order to accurately identify callers and verify that they are who they say they are. 

    While they’re often lumped together, identification and verification are two separate tasks with two distinct purposes. 

    Identification is the process of matching a caller with an entry in your CRM to check their identity

    Verification is the process of checking the caller is who they claim to be

    In this post, we’ll look at the different methods contact centers can use to identify and verify callers, and explore which work best in terms of customer experience, accuracy and implementation. 

    Methods of Identification and Verification

    There are three common ways to identify and verify callers:

    1. Automatic Number Identification (ANI) – identifying callers based on their phone numbers
    2. Knowledge-Based Authentication (KBA) – identifying and verifying callers based on their responses to a question/set of questions
    3. Voice Biometrics – verifying callers based on their voice 

    Automatic Number Identification (ANI)

    ANI is the process of identifying a caller based on their phone number. It’s typically provided by CCaaS providers and uses phone numbers to discern the caller’s identity based on CRM records.  

    ANI provides a seamless experience for the caller, who can jump straight into their query without needing to go through an identification process. 

    While ANI is easy for callers, it’s not particularly secure (phone numbers are relatively easy to spoof) and it only works if the caller phones from the number that is stored in your CRM. In some cases, multiple customers will have the same phone number and the system will not be able to identify which one is calling. 

    ANI is great for enabling contact center agents to offer personalized service as soon as they answer the phone but it’s not a reliable or secure method of identifying or verifying callers, and will require a fallback option. 

    Knowledge-Based Authentication (KBA)

    Knowledge-Based Authentication (KBA) is the process of asking callers a series of questions in order to ascertain their identity. KBA is prevalent in contact centers. Opus Research found that 99% of firms use knowledge-based methods for authentication and fraud detection. 

    Callers can be identified against any piece of information the company has in their CRMs, such as name, date-of-birth, customer number, answers to a secret question, mother’s maiden name etc.

    KBA is naturally conversational, and although it requires a little effort from the customer, it is generally a simple process that callers are already familiar with. KBA is popular in contact centers as a conversational method of both identifying and verifying customers. It is secure in that multiple customers are unlikely to match against a combination of personal details (e.g. name and date of birth), but is susceptible to social engineering

    The challenge with automating KBA is speech recognition. In order to identify a caller by name, order number or ZIP code, those pieces of information need to be accurately taken down by the automated system. 

    PolyAI voice assistants have attained a reliable level of accuracy to perform KBA over the phone. We’re able to identify callers based on any piece(s) of information the company stores in their CRM including ZIP codes, order numbers, names, addresses and many more. 

    Voice Biometrics

    Voice biometrics solutions are used to verify callers’ identities based on the sound of their voice. Voice biometrics solutions require customers to provide a number of voice samples to make up a voiceprint. Some solutions require customers to state a phrase like ‘My voice is my password.’ Others can identify customers based on their voice within a natural conversation, so seamlessly that customers don’t even notice it happening. 

    There are a couple of issues with voice biometrics that you should be aware of.

    Voiceprints are not unique. There is a 0.5% (or 1 in 200) chance of getting a false match. Using voice biometrics in the contact center means that 1 in every 200 calls will include a caller being misidentified. 

    Secondly, while it is difficult to fool voice biometrics solutions, it is not impossible

    For these reasons, it would be unwise to use voice biometrics to identify customers. Voice biometrics is best suited to verifying callers who have already been identified via other means (such as KBA). 

    Finally, it is worth considering the investment that voice biometrics requires from customers. Opus Research found that when customers are invited to opt-in for voice biometrics, only around 35% will take up the offer. Companies considering investing in voice biometrics should consider not only the cost of implementation but also the cost of optimizing performance over a period of many months once the solution is deployed. 

    Methods of ID&V: A Comparison

    Automatic Number Identification

    Knowledge-Based Authentication

    Voice Biometrics

    User Experience

    Good.


    Seamless CX, but will fail if the customer calls from a different number.

    Good. 


    Naturally conversational and easy if verification questions are well-designed.

    High.


    Set-up may be difficult to explain to some customers, but once they are set up, experience should be easy.

    Accuracy of identification

    Medium - high.


    Accuracy depends on callers consistently calling from the same, personal number

    Medium-high. 


    Accuracy depends on how many questions are asked and how difficult the answers are to obtain by other parties via e.g. social engineering

    Medium-Low


    Accuracy becomes worse as customer base increases. 1/200 people will be wrongly identified. 

    Accuracy of verification

    Low. 


    Multiple customers may be registered with the same phone number & phone numbers are easy to spoof.

    Medium-high. 


    Likelihood of multiple customers having the same information is low. Susceptible to social engineering.

    High.


    Hard to impersonate but not impossible

    Implementation

    Easy. 


    Offered by most CCaaS providers.

    Medium.


    Requires sophisticated AI technology as speech recognition alone is unlikely to capture details accurately.

    Hard. 


    Requires collecting voice samples from customers and implementing voiceprint look-up.

    Cost

    Low.

    Medium.

    Medium-high.

    Creating a Customer-Centric ID&V Process 

    How companies choose to identify and verify customers will depend on the level of security necessary to provide customers with access to their accounts, and the level of customer experience the company wishes to provide. 

    PolyAI recommends following:

    Identification:

    • Use knowledge-based authentication to identify callers where ANI doesn’t work 
    • Where possible, identify the caller based on their phone number using automatic number identification.

    Verification:

    • Use knowledge-based authentication to verify callers based on personal information
    • Use voice biometrics as an additional layer of authentication

    There’s a reason call centers have relied on knowledge-based authentication for so long. It is a secure and customer-friendly means of both identifying and verifying callers. 

    Automatic number identification can help to reduce average handle time in call centers.

    Voice biometrics is a nascent technology that can create smooth customer journeys, but voiceprints are not unique and voice biometrics should only be used in conjunction with other methods of identifying callers. 


    Featured

    Conversational AI for Business | January 2021

    3 Common Voice Assistant Errors, and How to Avoid Them

    Conversational AI for Business | September 2019

    Choosing Customer Goals to Automate