Supplier invoices are OCR'd the moment they arrive and matched line-by-line against purchase orders and goods-received notes. 94% post to the ERP without a single keystroke; only true mismatches reach a human, with the discrepancy already highlighted.
Typical results
36
Exceptions/month, down from 600
92 hrs
Saved every month
1 in 20
Invoices caught overbilling
Figures reflect typical results for this kind of automation.
The challenge
Around 600 supplier invoices arrived every month — email PDFs, paper, the occasional fax. Two bookkeepers keyed each one and hunted down the matching PO and delivery note, burning roughly 100 hours a month. Price creep and short deliveries slipped through, and month-end close always ran days late.
Before and after
Invoices touched by hand per month
−94%Matching workload per month
−92%Invoices matched untouched
The automation
We built an intake that OCRs every invoice on arrival — email, scan or photo — and extracts supplier, lines, quantities and prices. A Make scenario then runs a three-way match against the purchase order and the goods-received note in the ERP, line by line, with tolerance rules the CFO controls. Clean invoices post automatically and queue for payment; mismatches open an exception card showing exactly which line disagrees and why.
How it works
- 01
Invoice lands
Email, scan or photo — every invoice is captured and OCR'd into structured lines.
- 02
Documents pulled
The matching purchase order and goods-received note are fetched from the ERP.
- 03
Line-by-line match
Quantities and prices are compared under tolerance rules the CFO controls.
- 04
Clean invoices post
Matched invoices post to the ERP and queue for payment with no human touch.
- 05
Exceptions surfaced
Mismatches open a card pinpointing the exact line and the exact difference.
Starts from
- Supplier invoices
- POs & goods-received notes
Delivers
- Posted to ERP
- Exception cards
- Payment queue
The impact
94% of invoices now match and post untouched, leaving about 36 exceptions a month instead of 600 manual entries. Matching workload dropped from roughly 100 hours to 8, the match catches overcharges on about one invoice in twenty, and month-end closes on time.
Questions about this automation
How does the three-way match decide an invoice is clean?
Each invoice line is compared with the purchase order line and the goods-received note for the same item. Quantity and unit price must agree within tolerance rules the finance lead sets, such as a small percentage or a fixed amount. Only when every line passes does the invoice post to the ERP and join the payment queue.
What happens when an invoice does not match?
The invoice is held and an exception card opens for the bookkeeper. The card shows the supplier, the line that disagrees and the size of the gap, for example three units short or a price above the agreed rate. The bookkeeper approves, disputes or rejects it, and that decision is logged before the invoice moves on.
Which systems and invoice formats does it work with?
It reads emailed PDFs, scanned paper and phone photos through OCR, and connects to ERPs and accounting systems that expose purchase orders and goods receipts through an API. The flow itself runs in Make. This is a multi-step build, usually rolled out in stages, often starting with the highest-volume suppliers.