WEB3wallet apps / dApp interfaces / token-gated membershipExplore Web3 →
All work
AI · Hospitality

AI Invoice Processing for Restaurants

Matchbound is an AI invoice tool we designed for restaurant groups that pay many suppliers. It reads every invoice in the accounts inbox, checks each line against the order and the delivery, and leaves people only the lines that don’t add up.

AI & automationHospitalityWeb app
Matchbound: invoice review screen with recognised fields and highlighted mismatches
Industry
Hospitality, accounts payable
Platform
Web app
Our role
Product and UI design
Posts to
QuickBooks, Xero, NetSuite
Who it is for

Who Matchbound is for

Accounts payable teams in restaurant groups with several locations and dozens of suppliers, where invoices arrive every week as PDFs, scans and photos.

  • The AP clerk who keys in invoices and chases the differences.
  • The approver who signs off bills above a set amount.
  • Kitchen teams who receive deliveries and photograph paper invoices.
Why it is needed

Why invoices need checking before you pay

The cost is not the typing. It is the invoice that bills tomatoes at the old price, the short delivery nobody noticed and the lemons that were never ordered.

  • A price above the one agreed on the order.
  • A quantity billed that never arrived.
  • An item that was never on the order.
What we built

What we designed for Matchbound

01

An inbox that reads itself

Invoices from mailboxes, a forwarding address and kitchen photos. The queue shows what matched, what needs review and what is a duplicate.

02

Review with confidence marks

The PDF on the left, the fields the AI read on the right, each with how sure the model is.

03

Line by line

Invoice, purchase order and goods receipt side by side. Gaps in price, quantity and order are highlighted with a reason.

04

Approve and post

The amount to pay, short-pay lines, a credit note request and QuickBooks entries with the accounts filled in.

05

Ask the buyer

A question on price or quantity goes by text to the person who raised the order.

06

Rules, not code

Mailboxes, tolerances, approval limits and duplicate holds are settings.

How it works

How an invoice goes from email to an approved bill

01

Arrives

A supplier emails a PDF. Matchbound reads the lines and the totals.

02

Matches

Each line meets the order and the receipt. Clean invoices post on their own.

03

Flags

Differences above tolerance go to a person, with an amount and a reason.

04

Approves

The clerk decides the flagged lines and posts the bill; the supplier gets the reasons.

How we built it

How we designed Matchbound

We followed one produce invoice from the morning email to an approved bill, and designed each screen around the decision it asks for. The screens are laid out in HTML and CSS.

Stack
HTMLCSS
01

One invoice, end to end

A sample invoice with three classic problems: a price rise, a short delivery, an item not ordered.

02

Exceptions first

The inbox puts the invoices that need a person on top.

03

Show the arithmetic

Every difference carries its amount and reason, so the total to pay is never a guess.

04

Six screens

Landing, inbox, review, line by line, approval and settings.

Screens

A closer look at the screens

What it does

What Matchbound does

  • Matches every line three ways: invoice, purchase order, goods receipt.
  • Names and prices each difference, and lets small ones pass within set tolerances.
  • Holds duplicates instead of paying twice.
  • Posts approved bills to the books and sends the supplier the short-pay reasons.
FAQ

Questions about this project

Comparing each invoice line with the purchase order and the goods receipt, so a wrong price, a short delivery or an unordered item is caught before payment.

Got an idea? It’s one message away.

Scope agreed before development, fixed price, six-month warranty.