Document intake

Work > Document intake takes in paper and email so nobody has to sort it by hand.

Where documents come from

• Upload documents: PDFs, scans, photos of paper, Word files or text, up to 20 at a time.
• Email: attachments of messages arriving in the shared inbox, when Company settings > Branding and security > Read attachments of incoming emails is on.
• The same file sent twice is only taken in once.

What happens to each one

1. Text is read: from the PDF's text layer, or by text recognition (OCR) for scans and photos.
2. It is classified: medical invoice, repair estimate, police report, claim form, ID document, bank details or proposal form, with how sure the system is.
3. Fields are pulled out: claim and policy numbers, amounts, dates, invoice numbers, plate numbers, emails.
4. It is matched to a claim, policy or customer.
5. When the match is sure (at or above the class's threshold), it is filed on the record by itself. Otherwise it waits under To check and in the exception queue.

Checking a document

Open it to see the file, the fields found and the score per kind. File it with the right kind and the claim or policy number; each confirmation teaches the classifier what that kind looks like. Reject what does not belong. AI review asks for a short reading of the document.

Kinds of document and retention

Automation > Document classes lists each kind with its keywords, the fields to pull out, what it links to, the confidence needed to file by itself and how many years it is kept. Add the usual kinds installs the starters. Past its keep-until date, the intake copy is archived (the copy on the record follows the record's own retention), unless a legal hold covers the record.