Turn email into a table with Collections
Create a collection from a template or your own fields, choose which mail it reads, and let new messages become rows.
Collections turns the email that matches a rule into rows of structured data, such as supplier bills, receipts, orders or sales enquiries. Instead of opening each message, you get a table with the amount, date and sender pulled out for you. An assistant reads each matching message to fill the fields, so it uses your workspace's monthly AI allowance.
Create a collection
- Open Collections and choose New collection. The page says Turn email into structured data.
- Pick a template, or Custom for your own fields and rule:
- Invoices: bills from suppliers: who, how much and when it is due.
- Receipts: what you paid, where and when.
- Orders: customer orders as they come in.
- Leads: new enquiries from potential customers.
- Payment alerts: bank credit and debit alerts.
- Set a Name and Description. You can change everything afterwards.
- Under Fields, list what to pull out of each message. Give each a name and a type: Text, Number, Money, Date, Email, Link, Yes / no or Choice (with comma-separated choices). Mark the ones that are required. A collection can have up to 30 fields. Required fields that cannot be found send the row to review.
- Under Which mail, set the rule. Only messages matching every condition you set are read:
- Mailboxes: All my mailboxes, Some of my mailboxes, or Every mailbox in the workspace (administrators).
- From (addresses or domains, comma-separated), for example
[email protected], stripe.com. - Subject contains any of and Message contains any of.
- Has an attachment. A live preview shows which mail would match as you add conditions.
- Under Review, set the review threshold. Rows the assistant is less sure about than this percentage wait in a review queue. The default is 70%.
- Under Shared with, choose who can see, edit and review rows: just you, Everyone in the workspace or Specific people. Only you can change the collection itself.
- Choose Create collection.
New matching mail appears as rows from then on. To bring in older mail, use Look at older mail. It runs over your most recent matching messages, each 10 messages use one AI request, and messages that already have a row are skipped. A run covers up to 200 messages.
Work with the rows
Each row shows the extracted fields and a Received date. Open a row to see the message it came from. Double-click a cell to edit it. Filter by status: All but rejected, Saved, Needs review, Waiting for AI, Rejected or Everything, or search with Search rows. Use Export CSV to download the table.
See Review rows and add automations.
Good to know
- You can have up to 50 collections.
- A row marked Waiting for AI will be filled in automatically once the allowance allows.
- Deleting a row does not bring it back from the same message unless you run a backfill.
- Collections only read mail you can read. API keys limited by an agent policy cannot use them.
Tips
- Start from a template and delete the fields you do not need.
- Narrow the rule by sender first. It saves AI requests.