Post-call analysis tells your AI voice agent what structured data to pull from each conversation, from a call disposition to a custom "asking price" number. You configure these fields directly in the agent editor.
Open the Post Call Analysis section
- Go to the AI Voice surface and open the agent you want to edit (or create a new one).
- In the agent editor, find the Post Call Analysis accordion and expand it.
- Pick an Analysis Model from the dropdown. This is the model that reads the transcript and fills your fields. It defaults to GPT-4.1 Mini (Free); Free-tier options are labeled
(Free).
The header reminds you that results are extracted automatically and delivered via webhooks after each call ends.
System fields you cannot change
Every agent ships with locked System fields (marked with a lock icon) that you cannot edit or remove. They keep data consistent across all your agents:
| Field | Type | Purpose |
|---|---|---|
call_disposition |
Selector | One of 9 outcomes (Qualified – Interested, Do Not Call (DNC), No Answer / Voicemail, etc.) |
interest_level |
Selector | Hot, Warm, Cold, or na |
call_summary |
Text | Structured summary of motivation, timeline, and price |
spoke_with_owner |
Yes/No | Whether the owner actually joined the call |
gatekeeper_screen |
Yes/No | Whether a screener answered instead |
price_node_reached |
Yes/No | Whether pricing was discussed |
extracted_first_name, extracted_last_name, extracted_property_address, plus phone/email/mailing fields |
Text | Feed contact auto-updates from what the lead disclosed |
Add a custom extraction field
- Click the dashed Add button to open the Add Extraction Field dialog.
- Choose a Type:
- Text — names, summaries, free-form values (supports optional Examples)
- Yes/No — a true or false determination
- Number — a numeric value
- Selector — a fixed choice list (at least 2 options required)
- Enter a Name. It is auto-formatted to snake_case (for example,
Asking Pricebecomesasking_price). - Write a clear Description of what to extract. Be specific — the model follows this instruction literally.
- For Selector fields, add each option; for Text fields, add Examples if helpful.
- Click Add (or Update when editing). Hover any custom row to reveal the edit (pencil) and remove (trash) buttons.
Fields save when you publish or update the agent. Reopen the editor anytime to see your saved items.
How fields surface later
After a call, the analysis model populates every field and the results post to your webhook. The extracted_* system fields also drive automatic contact updates from disclosed names, phones, and addresses.
When you build automation rules across multiple agents, only custom fields that match by name, type, and option list on all selected agents appear as filters. To act on the captured outcomes, set a contact's status from the result — see Set Lead Status and Temperature.