Skip to main content
custom_extractions → Rebuild, one per item — 18 September 2026. custom_extractions returns null. The key stays in the payload.
One extraction per configured item. The item type determines the answer type you pick.

​
Type mapping

An extraction returns one value, not a nested structure. Pre-defined options accept sub_options, but these fold into the same flat set of allowed answers instead of nesting. That is why a multi-part item becomes several extractions in a shared category.

​
What to do

1

List your configured items and their types

2

Create a category per logical group

For example Loan Eligibility.
3

Add one extraction per item

Use the mapping above. Set name to the existing key to keep result keys stable.
4

For eligibility_analysis, name the verdict

Give the verdict a real name instead of the fixed eligibility_key.
5

Test against a real past transcript

See Verifying a migration.
6

Remove the Custom Analytics items from the agent

​
Before and after

Note that advanced_list and eligibility items were flattened up to the same level as everything else, rather than nested. That structure is unchanged in principle: related fields now sit in a shared category instead.
Numbers become strings. loan_amount was the JSON number 250000.0 and is now the string "250000". This is the change most likely to break a downstream numeric field silently. validation.is_valid confirms the value parses as a number, but the parsing is yours. There is also no equivalent of the min/max clamping Custom Analytics applied, so out-of-range values now reach you unchanged.

​
What changes in your code

Full detail on the result object is in What the payload looks like.

​
Other features

Extraction

extracted_data · flat keys

Agent Extraction

agent_extraction

Call Summary

summary
Email support@bolna.ai to remove a legacy summarisation task, or with an execution_id for a payload that does not match this guide.

​
Next Steps

Using Extractions

Configure extractions in the dashboard

Dispositions API

Manage extractions programmatically

Webhooks

Receive extraction data in real-time