š Hey dev.to Community,
A while back I shared a purchase order extractor I built for a friend with a big online store who was working through PO PDFs by hand. It's now published on the n8n template library, so you can grab it for free.
What it does: you upload one or many PO PDFs through a form, and it extracts each into a Google Sheet, one row per line item, with the source document name on every row. It handles POs in different layouts, runs them as a batch, and before writing it checks each PO number against the sheet so nothing gets entered twice. The completion screen then tells you what was added, what it skipped as already processed (with the PO number), and whether any extraction looked off.
One thing worth pulling out if you build batch extraction of your own: a batch is only worth running if you can trust it without eyeballing every row. The feature that actually made this usable wasn't extraction accuracy, it was the summary that surfaces skipped duplicates and failed fields, so you know exactly what to check instead of assuming it all worked.
Template (free to grab): https://n8n.io/workflows/18674-extract-purchase-orders-from-pdfs-with-easybits-google-sheets-and-google-drive/
The optional EDI 850 add-on (for pushing orders into SAP and other ERPs) isn't in the library version, but it sits in my repo alongside 25+ other workflows: https://github.com/felix-sattler-easybits/n8n-workflows
If you're automating PO or invoice intake too, how are you handling the fact that every supplier sends a different layout? That was the hardest part to make reliable here, and I'd be curious how others approach it.
Best,
Felix
Top comments (0)