Voice AI Custom Actions let your Voice AI agent trigger real-time webhook integrations during a live call. Instead of waiting until the call ends, your agent can interact with external systems mid-conversation—pulling data, executing processes, and personalizing the call based on what the caller says.
Note on PHI: Don't enter protected health information (PHI) into Voice AI. AI features aren't approved for processing PHI — use them for productivity, communication, scheduling, and engagement, not to store or process medical records. See the HIPAA Compliance guide for details.
Voice AI Custom Actions allow your AI agent to trigger custom POST webhook calls to external APIs during a live conversation. These actions can include authentication, headers, and dynamic parameters collected in real time from the call, so your agent can retrieve or send information instantly based on what the caller mentions.
Custom Actions connect your Voice AI agent to external systems directly within a live call, helping it personalize conversations and resolve requests faster.
Perform real-time API calls triggered by conversation cues.
Automate data lookups or submissions mid-call.
For example: If a caller asks, "What's the status of my recent order?", the agent can instantly call your order management system and retrieve the real-time status—without putting the caller on hold.
Configure POST requests with authentication and headers.
Dynamically pass call data, like phone numbers or order IDs.
Test webhook responses before going live.
Reduce follow-up tasks by resolving needs during the call.
Triggers define the conditions under which a Custom Action runs during a call. You can create simple phrase-based triggers or configure more complex logic. Triggers can also be layered with conditions, such as "only run if parameter X is present."
Example triggers:
When a caller says: "I want to check my appointment."
When an email is mentioned.
When a string of digits (such as an order number) is spoken.
Webhook integration is the core of Voice AI Custom Actions. It lets your agent interact with any external system that supports APIs—your CRM, scheduling tools, databases, and more.
Each Custom Action is defined by a POST request, which may include:
A webhook endpoint URL.
Headers (such as API keys or tokens).
A request body with dynamic parameters.
Authentication (Bearer token, Basic Auth, and similar).
Example: If you're integrating with an external system, your webhook URL might target an endpoint like /api/v1/lookupContact and include parameters such as the contact's email or phone number.
Voice AI can extract and label relevant data in real time during a conversation. These values are automatically assigned to the parameters used in the webhook request. This is useful when, for example, the agent needs to capture both an order number and an email address before triggering a shipping status lookup. You can assign these extracted values to your webhook payload so the integration stays context-aware and personalized.
Supported data types:
Text (String)
Number (Numeric)
Phone Number
Date
You can configure Custom Actions from the Voice AI agent interface in just a few steps.
Go to AI Agents > Voice AI > Agent Goals > Switch to Advanced Mode (if not already enabled) > Custom Actions.
Click + New Action to open the Custom Action configuration window.
Enter the details for your Custom Action. Not all fields are required.
Name
Set the conversation trigger conditions.
For example: If a caller asks to "check my account balance," you can set a trigger with the phrase "check my balance" and define a parameter for their phone number to pass into your external API.
Add your Webhook URL and select POST as the method.
Add any custom headers as needed.
Enter any required authentication details (such as a Bearer token).
Define parameters that are dynamically pulled from the conversation.
Use the Test Webhook feature to validate the setup.
Before saving a Custom Action, use the built-in Test Webhook tool. This lets you simulate a call scenario, pass test data, and view the response from your external system in real time. For example, you can mimic a caller asking to "reschedule an appointment" and check whether the webhook correctly pulls and sends the provided date and time to your scheduling system.
With the test tool you can:
See the full request (headers and body).
View the raw response (200 OK, 404 Not Found, and so on). For example, if your webhook response includes an estimated delivery date, your agent can immediately tell the caller: "Your package is expected to arrive by Thursday."
Identify and fix misconfigurations before saving.
Can I use GET or other request types?
No. Only POST requests are currently supported for Custom Actions.
Where do I access Voice AI Custom Actions?
Go to AI Agents > Voice AI > Agent Goals > Switch to Advanced Mode (if not already enabled) > Custom Actions.
Is authentication supported in webhooks?
Yes. You can use Bearer tokens, Basic Auth, or pass keys in the headers.
What if my webhook fails during the call?
The system logs the failure, and you can define fallback behavior for when no data is returned or the webhook times out.
Can I trigger multiple webhooks during a single call?
Yes. Each Custom Action can be triggered independently based on its own conditions.
Do I need a developer to set up these actions?
Not necessarily. As long as you have access to the API documentation for the external system, you can set this up with minimal technical skills.
