As of 13 August 2026, AI can only partly extract data from email attachments.

What to actually do

1. Hand it to a person

   The route this page recommends

   A person who owns the outcome does this end to end, worth it when the failure is dear.

2. Use a tool built for this

   Second choice

3. Do it yourself

   The distant third

   A chat interface, chat-fluent skill, and roughly 30 minutes until you can act on the result.

   How to actually do it

   1. Open the mailbox folder containing the relevant messages and download the original emails and every attachment, keeping the original filenames.
   2. Create a field list for the extraction, such as invoice number, supplier, date, amount and VAT, and define the exact format required for each field.
   3. Remove unrelated messages and attachments, then upload the selected files to a chatbot that supports document and image uploads.
   4. Paste the prompt with your field list and ask for one separate row per document, including the source filename, location, confidence and notes.
   5. Open each source attachment and compare every extracted value marked medium or low confidence with the original page, cell or image.
   6. Check a sample of high-confidence rows against the originals, then correct the table without accepting values that the source does not contain.
   7. Export the checked table to the approved spreadsheet or business system, retaining the source filenames and the original files for audit purposes.

   Prompt

   Extract the requested data from the attached email files and their attachments.
   
   Requested fields:
   [LIST EACH FIELD, FOR EXAMPLE: SUPPLIER NAME, INVOICE NUMBER, INVOICE DATE, DUE DATE, NET TOTAL, VAT TOTAL, GROSS TOTAL]
   
   For each attachment, return one row in a table with these columns:
   - source email or attachment filename
   - page or cell reference where available
   - each requested field
   - confidence: high, medium or low
   - notes
   
   Rules:
   1. Copy values exactly as shown. Do not calculate, correct, infer or invent anything.
   2. Preserve dates, reference numbers, currencies and decimal places as displayed.
   3. If a field is missing, write "not found". If it is unreadable or ambiguous, write "unclear" and quote the relevant text.
   4. Keep separate rows for separate documents. Do not merge values from different attachments.
   5. Identify duplicate-looking documents but do not delete any row.
   6. For scanned or handwritten content, mark the affected field as low confidence unless it is clearly legible.
   7. After the table, list every field that needs a human check and explain exactly why.
   8. Do not include personal or confidential information that is not needed for the requested fields.
   
   Attachments and email files:
   [PASTE OR UPLOAD THE FILES HERE]

   Open it prefilled in ChatGPT or Claude, or copy it into Gemini, which takes no prefill link.