Automated Invoice Processing

01The challenge:

2,400 supplier invoices a month, each one typed in by hand

FinanceFlow runs accounts payable for about forty construction subcontractors and restaurant groups. Supplier invoices arrived as email attachments in a dozen layouts, and three staff typed each one into the right client's QuickBooks file, then checked it against the purchase order line by line. A clean invoice took twelve minutes. A messy one took half an hour.

Month-end was the bottleneck. Invoices piled up in the shared inbox, early-payment discounts lapsed, and two clients paid the same lumber invoice twice because it arrived by email and by mail. FinanceFlow could not take on new clients without hiring another clerk for every four it signed.

02What we built:

Every invoice read, matched to its PO and posted without retyping

An n8n workflow watches each client's AP inbox. When a PDF lands, it goes to an AWS Lambda function in Python that runs OCR and pulls the supplier, invoice number, dates, line items and totals. Fields read with less than 85% confidence are flagged, and the supplier is matched to the vendor list in that client's QuickBooks file.

A duplicate check on supplier, number and amount runs first. The workflow then pulls the open purchase order through the QuickBooks API and compares quantities, unit prices within 2%, and totals. Clean invoices post as bills awaiting payment approval. Anything else lands on one exception list with the PDF and the mismatch highlighted.

03What changed:

Processing time fell by nine tenths, and staff only work exceptions

Invoice processing time dropped 90%, from twelve minutes to 1.2 on average across 2,400 invoices a month. Of those, 82% post without anyone touching them. The duplicate check stopped eleven repeat invoices in the first quarter, and early-payment discounts are now taken on time.

FinanceFlow onboarded nine new clients over the next two quarters with the same three staff. The workflows, the Lambda code and the QuickBooks connections sit in FinanceFlow's own AWS and n8n accounts, and its operations manager adds each new client from a setup checklist written with the team.