Capturing data from conversations
With the Bhomy Context Extractor, bhomy pulls the name, request, preferred appointment or your own fields from every conversation in a structured way.
After every conversation, the Bhomy Context Extractor picks out the details you need and stores them as variables: the caller's name, their request, a preferred appointment or a field that only exists in your industry. You use these variables in emails, SMS and webhooks.
Defining variables
- Open your agent and stay on the Personality tab.
- Open the Bhomy Context Extractor section.
- Under Quick add, add standard fields or create a Custom variable.
- Choose a type for each variable and write a Description for AI.
- Drag the variables into the order you want to see them in, and save.
Two variables are already set up: the caller's name and a short description of the request.
Types
| Type | Example |
|---|---|
| Text | request, vehicle model, type of damage |
| Number | number of people, square metres |
| the caller's email address | |
| Phone | callback number |
| Date | preferred appointment, date of birth |
| Yes/No | existing customer, emergency, callback requested |
Quick add offers email, phone number, preferred date, preferred time, company and address.
The description for AI
The description tells bhomy exactly what belongs in the variable. A good description is unambiguous:
| Good | Too vague |
|---|---|
| "The caller's full name, first and last name" | "Name" |
| "Yes if the caller has been a customer before, otherwise No" | "Customer?" |
| "Requested date of the appointment, not today's date" | "Date" |
If a detail doesn't come up in the conversation, bhomy sets the variable to null. To have the value as often as possible, tell the agent in the prompt what to ask for.
Where the values end up
| Destination | How |
|---|---|
| Call history | in the call details on the Extraction tab |
| Email and SMS | via Insert variable, such as {{anliegen}} |
| Webhook after the call | in the fixed fields context and extractionData, if your variables have the same names as these fields, such as email, firma, anliegen or prioritaet. The exact structure is shown under What bhomy sends after the call |
bhomy stores captured data for 90 days and then deletes it automatically. If a caller declines the recording and Automatic transcript deletion on refusal is switched on, bhomy doesn't extract any data from that conversation.
Example: damage reports for a trades business
| Variable | Type | Description for AI |
|---|---|---|
damage_type | Text | Type of damage, such as a burst water pipe, heating failure or blocked drain |
emergency | Yes/No | Yes if water is leaking or the heating has failed in winter |
adresse | Text | Street, house number and town where the damage is |
callback_number | Phone | Number on which the caller can be reached today |
With an email with the condition "The caller reported an emergency", every emergency report then goes straight to your on-call service.
Was this page helpful?