These components have additional usage costs. Callers will only be enrolled if they explicitly opt into having their voice profile captured. At any time, during an interaction with an agent, callers can ask to trigger an opt-out, which will delete their respective voice profiles. Talkdesk recommends reaching out to your legal counsel about consent-based voice biometrics requirements, before setting up this functionality.
When a caller chooses to opt into the feature, the “Voice Authentication” and “Enroll voice” components play an important role regarding voice collection and imprinting, and secure authentication.
Talkdesk Identity offers two types of voice authentication - active and passive. When configuring Studio, you can choose if you want to use both or just one of them.
The main differences between these authentication types are highlighted in the next table:
Active voice authentication |
Passive voice authentication |
Both the enrollment and authentication are done in the IVR |
Both the enrollment and authentication are done during a live call and no passphrase is necessary |
Requires the callers to repeat a passphrase to enroll and authenticate |
No passphrase is necessary |
When building the flow in Studio you need to configure the “Enroll voice” and “Authenticate voice” components |
When building the flow in Studio you only need to configure the “Authenticate voice” component |
To know more about the difference between both, check our FAQ.
The recommended approach is to use both authentication types since combined they provide more secure authentication and higher flexibility for agents to detect and prevent fraudulent attacks during a call.
Caller and agent experience with active and passive
For the caller, the experience of having to authenticate with only active or with both active and passive combined is the same.
After the caller consents to use voice authentication, the passive enrollment and authentication will begin. The caller just continues to talk with the agent, and the system captures the caller’s speech and displays the result to the agent. The agent needs to inform the caller that they need to stay on the line to complete the active authentication, aka Passphrase. When the agent hangs up, the caller is instructed by the IVR to repeat a passphrase and the enrollment is done.
In the next calls, before the agent picks up, the IVR instructs the caller to repeat the same passphrase used during enrollment, and the authentication is done. When the call starts, passive authentication is done when the caller starts talking with the agent.
In all this process, the caller just needs to engage directly with voice authentication when repeating the passphrase and is mostly oblivious of when passive is being performed. So, in a sense, having both authentication types configured for the caller is the same as having just the active authentication.
On the other hand, for the agent, having both authentication types configured gives the agent extra power to ensure that the person that authenticated with the IVR continues to be the same during the call. Besides the passive authentication that is done as soon as the call starts, the agent can at any point reauthenticate the caller and see the result instantly.
This article shows an overview of how agents interact with the enrollment and authentication and what they see in real time during these processes.
Main settings in Studio for voice authentication
Both the Authenticate voice and the Enroll voice have a set of configurations that are going to be explained in the flow presented in this article and set in an illustrative and sometimes recommended way for the flow to work.
Amongst these settings are:
- The type of authentication the callers will use;
- What is the language the callers will hear when receiving the instructions concerning the passphrase;
- The choice of which passphrase to challenge the caller with;
- Whether to play a “beep” so that the caller is made aware when they should start speaking the passphrase;
- Whether the caller is required to press a key (any key) to move forward after they have spoken the passphrase, or if the system is to detect silence and move forward 3 seconds after the caller has spoken the passphrase;
- Which unique identifier the system will use to match a voiceprint with.
Also, and as a general rule for any flow, it is necessary to configure:
- To whom the flow should be assigned - a particular agent or group of agents, a ring group, or based on an attribute.
- The phone number that is assigned to this flow to which callers will dial.
Configuring Authenticate voice and Enroll voice in Studio
To ensure that callers are securely authenticated, please configure your Studio flow as detailed below:
1. After the initial step, you can choose to add an Input IVR component that will be used to collect the contact’s unique identifier, i.e., the Contact ID.
The Contact ID is a personal and non-transferable number of which the caller is the sole holder and that will be used to perform a secure authentication every time the user calls in. Examples of Contact IDs that can be used by callers include:
- Bank account number.
- Credit card number.
- Social Security Number.
- ID Card number.
- Personal Phone Number. Any other set of digits that uniquely identifies the user calling in.
The purpose of this Contact ID is to have a single voiceprint matching a single contact. If you choose not to set any “Input IVR” component to collect a unique identifier, the caller’s phone number can be used as the identifier. However, please bear in mind that a phone number is not a unique identifier, as it can be shared by many people (in the context of a company or a household, for example), thus compromising the efficiency of the authentication.
In the example, the “Input IVR” is being used to collect the contact’s ID. Configure the Input IVR component according to your preferences and, in the end, set up the output variable from this component. In our example, we’ve named it “Contact_ID_Variable”.
2. Add a "Voice Authentication" component to the Input IVR exit for successful outcomes (exit “OK”) and fill in the remaining fields per your preferences.
3. On the “Voice Authentication” step, configure the audio message your contacts will hear.
4. Select the language in which the callers will hear the instructions for active authentication.
Note: The same language should later be selected when configuring the Enroll voice component.
5. Choose if you want the callers to wait for a beep signal before they start saying the passphrase. By default, the “Beep” setting is disabled. If you enable it, it is recommended that you indicate in the instructions that the caller will hear a beep signal, and only after they should start saying the passphrase.
6. You can also choose if the callers should press any key after they say the passphrase or just wait until the system detects silence. By default, the “Keypress” is enabled and in that case, it is also recommended that the instructions explain the need for the caller to press any key after they say the passphrase.
7. Configure the audio message your contacts will hear.
You can use the variable {{passphrase}} in the text-to-speech audio format. This passphrase is the phrase that the contact uses to enroll, so for an accurate authentication, we advise that this passphrase be stated during the audio instructions in any of the available options (“Text-to-speech”, “Audio file upload” or “Other sources”).
Note: Please be aware that steps 4, 5, 6, and 7 are only necessary if Active authentication is selected - as in this case, where the option was “Active & Passive”. These steps are necessary since they ensure that authentication with the passphrase follows your requirements.
8. Configure “Caller unique identifier” as the Contact ID chosen in the “Input IVR” in step 1.
9. Configure the "Authenticate voice" exits:
- Success: In this section, you must select where to send the call if the voice authentication was successful, i.e. when the caller’s audio passphrase matched the contact ID voiceprint.
- No Match: Select where to send the call if the voice authentication failed, i.e. the caller’s audio did not match the contact ID voiceprint or an error occurred. You can use a “Play Audio” component to let your caller know that the authentication was not successful.
- Contact not enrolled: Select where to send the call when the caller is still not yet enrolled into voice biometrics, i.e. the contact still doesn’t have a voiceprint created. You can also add a "Play Audio" component, for instance, to let your caller know that a voice authentication feature is available, or you can choose to add an Assignment and Dial step to have the caller speak to an agent who will propose an opt-in to the feature.
- Timeout: Configure the maximum time the system will wait for the caller to speak the passphrase, before redirecting the call to the next component in the flow.
- Error handling: This section is available in most Studio components. You can use it to define the behavior for unexpected failure during the execution of a component. For more information on this section, please refer to the "Error handling" documentation.
If the contact is not enrolled, the voice authentication failed or if the call has timed out, we recommend you consider adding, to those exits, the “Assignment and Dial” component. This way, your agents can perform Knowledge Base Authentication (KBA) questions at any point.
10. After the exits have been configured, the call should be routed to an agent, so that relevant Knowledge Base questions are performed to ensure the identity of the contact person.
Furthermore, routing to an agent is crucial for Passive authentication to work since it’s during the call with the agent that the enrollment and authentication
Configuring Enroll voice
In this flow, the "Enroll voice" component was placed after the “Assignment and Dial” component, on the “Call Finished” exit. Therefore, the agent will be able to perform some KBA questions and ask for the caller’s consent to enroll using active authentication. After the agent hangs up the call, the caller will remain on the line and start the enrollment process.
11. First, you need to choose if you are either going to use a predefined passphrase or a custom one. The main differences are explained in this article here.
Then, similar to steps 4, 5, 6, and 7, you need to:
- Select the language the callers will hear the instructions to enroll. The language should be the same as the one selected in the “Authenticate voice” component;
- Choose if the callers will wait for a beep signal before saying the passphrase;
- Choose if the callers need to press a key after saying the passphrase or the system automatically detects the silence;
- Configure the audio instructions by choosing from the available options - “Text-to-speech”, “Audio file”, or “Other sources”.
12. Set up the passphrase confirmation. This second audio message is aimed at instructing the contacts that a new audio segment is needed for the enrollment to be completed. This audio message will be repeated until the number of required passphrase repetitions is enough.
Note: For successful enrollment, the number of passphrase repetitions that contacts have to state may vary between two and five, depending on some factors (audio quality, passphrase wording, background noise, etc.).
13. Configure the “Caller unique identifier” as the Contact ID variable defined in the “Input IVR” in step 1.
14. Set up the “Enroll voice” exits, depending on the flow you would like to define.
15. Click Save and then Publish, to complete the process.
Note: These steps assume that the "Voice Enrollment" component exits are configured as in our example above.
Conversations app
When an agent is using the Conversations app, they can see a tab called Identity that is shown automatically to all agents during all inbound calls.
The agents can see real-time information regarding the authentication of callers with voice biometrics and fraud-related insights concerning the caller’s phone number.
This article shows an overview of the Identity tab functionalities.