Documents and OCR

Internal chat does not accept attachments or execute OCR. There is no paperclip or drop zone in the conversation. Document capture belongs to other Frihet workflows: expense capture and invoice import.

Use expense capture

For receipts or supplier invoices, see OCR scanning and recording expenses. Use the capture workflow's file selector and check the formats and limits supported by that screen.

Extraction and data review are separate stages from completing a record. Check vendor, date, currency, amounts and taxes against the original document. Do not turn missing or uncertain information into confirmed data.

Choose the document type

An invoice received from a supplier and an invoice issued to a client are not the same operation. Expense capture does not create a sales invoice. To bring issued invoices in PDF or image format into Frihet, use the AI import in invoice import: it creates drafts that you review before issuing.

An invoice image does not establish that it has been paid either. Verify payments through the appropriate channel before changing its state.

Improve the original's readability

Use a complete, legible image with good lighting, without cropping amounts or supplier information. Retain the original to check the extraction. Review an unreadable field rather than inventing it.

A fixed processing time or error-free extraction is not guaranteed.

Privacy

Read the privacy information before uploading documents. To extract data, images are sent to a cloud AI model; PDFs are too, unless Frihet can read their text reliably in the browser. Extracted data is stored in your account and the original file stays attached to the expense.

Back to: What each AI channel can do.