Pull Structured Fields Out of Messy Text with AI Extract Data

Many useful details arrive buried in free text: an email from a lead provider, a long SMS, a webhook payload, or an AI response. The AI Extract Data action reads that text and returns clean, named values your workflow can use. It is a premium action with a per-run usage charge.

What you configure

• Extract From (required): the text to read, chosen with the variable picker.
• Additional Context: a sentence or two explaining what the text is, for example "This is a lead notification email; phone numbers may use different formats."
• Templates: starting sets for Contact info, Opportunity info, Order info and Appointment info. A template fills in fields that you can edit, add to or remove.
• Data Fields: each has a Name (the variable key), a Type (Text, Email, Phone, Number or Date) and an optional Description that guides the AI.

Set it up

1. Click + in your workflow and choose AI Extract Data under AI Actions.
2. Name the action.
3. Pick the source text in Extract From.
4. Add context so the AI knows what it is reading.
5. Choose a template or define your own fields, using precise names such as budget or order_id.
6. Click Save.

Use the extracted values

Each field shows up in the variable picker for later steps. Common follow-ups:

• Update Contact Field to store a value on the record.
• Create Contact or Create Opportunity from the extracted details.
• If/Else to branch on a value.
• Notifications, messages or an outbound webhook.

Getting accurate results

• Use the narrowest types possible; an Email field rejects things that are not emails.
• Write field descriptions like instructions: "Budget as a number, no currency symbol."
• Test with several real samples, including awkward ones.

Good to know: Extraction only returns what is actually in the text, so fields may come back empty when the source does not mention them.