AI Invoice Processing for a Restaurant Group

Matchbound was built for a restaurant group with a handful of locations and dozens of suppliers. Every week, invoices arrive by email as PDFs and photos; someone in accounts payable types them into the books, compares them with what was ordered and what actually arrived, and chases the differences by phone.

The cost is not the typing. It is the invoice that quietly bills tomatoes at the old price, the short delivery nobody noticed and the lemons that were never ordered. We wanted a tool that catches those three things on every invoice and leaves the accountant with decisions, not data entry.

The inbox collects invoices straight from the mailbox and shows the queue by supplier and due date. Opening one puts the PDF on the left and the recognised fields on the right — supplier, number, dates, totals — with anything the model is unsure about marked for a glance rather than a retype.

"Line by line" is the heart of it: each invoice line against the purchase order and the goods receipt, with the three classic discrepancies highlighted — a price that differs, a quantity short on delivery, an item that was never ordered. The sample invoice of $1,514.90 becomes $1,421.22 to pay, and the arithmetic is shown, not hidden. "Approve and post" writes the entries to QuickBooks with the right accounts; settings hold the mailboxes, the tolerances and the supplier list.