Skip to main content

Exporting VoiceAgent Call Results to CRM Integrations

Learn how to configure CloudTalk VoiceAgent to export AI-analyzed call data into your CRM

V
Written by Valeriia Volobrinskaia
Updated today

CloudTalk VoiceAgent helps you extract structured data from inbound and outbound calls using AI. This data can then be sent directly to your connected CRM or helpdesk system, enriching contact profiles with insights from real conversations.

User Level:

  • Admin


Supported integrations

Exporting VoiceAgent data is currently available for:

What gets exported

You fine-tune your VoiceAgent by defining a Call Analysis Prompt, that tells it what to extract from the call. This prompt analyzes the full call transcript and returns a structured JSON object with the fields you’ve defined.

You control exactly what the AI should extract. If you want to include contact-related details, like the caller’s number, company name, or CRM ID, you need to reference those values directly in your prompt using variables like {{caller_number}},{{contact.name}} or {{contact.custom.property_1}}

These values are available to the prompt at runtime but will only be included if you specify them.

Data included in the export

The exported data includes only the fields returned by your prompt. These can be:

  • Insights pulled from the conversation

  • Contact information you referenced in the prompt

VoiceAgent has access to contact context, which is the data CloudTalk already knows about the caller. This information is not added automatically, but you can include it by referencing the fields in your prompt.

Here is an example of contact context:

{
"id": 123,
"name": "John Doe",
"company": "Example Corp",
"title": "Software Engineer",
"phoneNumbers": ["+15550123", "+15550124"],
"emails": ["[email protected]", "[email protected]"],
"customFields": [
{
"title": "Department",
"value": "Engineering"
}
],
"integrationLinks": {
"salesforce": "https://salesforce.com/contact/123",
"linkedin": "https://linkedin.com/in/johndoe"
}
}

You can use this data inside your prompt or when setting up integration parameters. For example:

  • {{contact.name}}

  • {{contact.integrationLinks.hubspot}}

  • {{contact.custom.property_1}}

How Your CRM Gets the Data

When a VoiceAgent handles a call, it processes the transcript after the call ends. Your prompt defines what information the AI should extract. The result is a JSON object with the values you asked for.

CloudTalk then sends this JSON result to your connected CRM. The integration uses a contact ID or URL to update the correct record.

For example, in a HubSpot setup:

Request URL:

PUT https://api.hubspot.com/contact/{{id}}

Request body:

{
"name": "John Doe",
"phone_number": "+15550123",
"leadSource": string
}

Prompt Setup Example

Here is a basic example of a JSON structure you might define in your prompt:

{
"leadSource": string,
"inboundNumber": string,
"contactSegment": string
}

You can prefill known fields using variables like {{caller_number}} or {{contact.custom.property_1}}:

{
"inboundNumber": "{{caller_number}}",
"contactSegment": "{{contact.custom.property_1}}"
}

When setting up the integration mapping:

  • Use contact.name for the "name" field

  • Use contact.phoneNumbers[0] as the primary phone number

  • Use contact.integrationLinks.hubspot for the contact URL

For values that are missing or unknown, return "" for text, 0 for numbers, and false for booleans.
The output must be a clean JSON object with no extra formatting, labels, or explanations.

Exporting via Tools

If you're not using one of CloudTalk’s supported native integrations, or you need more flexibility, you can export VoiceAgent data using Tools. This method works by mapping values from the VoiceAgent output or contact context into Path and Body parameters.

For example, you can:

  • Pass contact.name as a body parameter to update a CRM record

  • Use {{contact.integrationLinks.hubspot}} as a path parameter to define the target URL

You can also write LLM prompts in parameter descriptions to help the AI extract specific values from the transcript, such as an ID from a link.

Example: Exported Call Data in HubSpot

Once everything is set up, the data extracted by VoiceAgent is sent to your connected system and automatically logged in the relevant record.

Here’s what the exported call data can look like in HubSpot:

The VoiceAgent analysis appears as part of the call note, including key fields such as call_result, caller_number, issue, and summary. This makes it easy for your team to review the outcome of the call without listening to the full recording.

Note that in order to export calls to HubSpot for numbers that are not yet saved as contacts, the setting "Create new contacts in HubSpot for unassociated call numbers" must be enabled in your CloudTalk integration settings.


If you need further assistance or have any questions, you can contact our Support team. We are always here to help you!

Did this answer your question?