Free skill pack
Pull clean data out of messy invoice PDFs
For bookkeepers, office managers, and ops leads who type invoice details into spreadsheets by hand. Upload invoices to ChatGPT, Claude, or Gemini, run two prompts, and get vendor, amounts, PO numbers, and due dates in a table you can check — because the second prompt audits the first one's work.
Free download · Markdown recipe
Invoice data extraction prompt pack
The full recipe: setup steps, two copy-paste prompts (extraction plus a verification pass), a field checklist, and notes for scanned documents and multi-invoice files. Works with any capable AI assistant.
- Download the recipe and open it next to your AI assistant
- Upload one to five invoice PDFs and paste the extraction prompt
- Run the verification prompt in the same chat
- Review flagged fields, then paste the table into your spreadsheet
What this handles
The extraction prompt pulls the fields most accounting workflows need: vendor name, invoice number, invoice date, due date, PO number, currency, subtotal, tax, and total. It converts dates to a single format, uses the legal billing entity when a logo and a legal name disagree, and refuses to guess — missing fields come back marked MISSING, ambiguous dates get flagged instead of silently reformatted.
It works on native PDFs and on decent scans or photos of paper invoices. It is deliberately tool-agnostic: the same prompts run in ChatGPT, Claude, or Gemini, so you are not locked into whichever assistant your team happens to use this quarter.
How to run it
Open a fresh chat and upload one to five invoices — small batches keep accuracy up, especially with scans. Paste the extraction prompt and you get one table row per invoice.
Then run the second prompt in the same chat. It re-reads each document field by field, checks that subtotal plus tax equals total, confirms the due date follows the invoice date, and scans the batch for duplicate invoice numbers. Anything it is not certain about lands in a NEEDS HUMAN REVIEW list with a reason.
Finish with the field checklist in the recipe before anything touches your books. Spot-check the first few batches against the original PDFs; once you know where your vendors' invoices trip the model, add a rule for it to the extraction prompt.
When to upgrade
This recipe is the right tool for a handful of invoices a day. It is the wrong tool when someone spends real hours every week uploading, checking, and pasting, or when the data needs to flow into your accounting system without a human in the middle. At that point you want a pipeline that watches the inbox, validates against your vendor list and PO records, posts to your accounting software, and only surfaces exceptions for review. That is what we build — and a free AI opportunity audit will tell you whether your volume justifies it yet.
Keep going
Data cleanup prompt pack
Prompts for standardizing the messy spreadsheet your extracted invoices land in.
Weekly report automation recipe
Turn the data you just extracted into a recurring report nobody has to assemble.
Invoice processing automation
What the fully automated version of this workflow looks like when we build it for you.
Want invoices to process themselves?
We build extraction pipelines that watch the inbox, validate against your records, and post straight to your accounting system — your team only touches exceptions. A free AI opportunity audit maps where that fits in your operation.
We take on companies ready to invest $5,000+/month. Not there yet? Our free resources are genuinely free.