macomore
All case studies

Customer story · Aparthotel group

Inside the group's ERP, supplier invoices are uploaded as PDFs or photos. AI reads them, someone checks them next to the original, and out comes the Excel ready to import into Wolters Kluwer's a3con.

· 2 min read

Invoice review: the fields read on the left and the original PDF on the right
Review: the fields it read, next to the original, ready to verify
Industry
Aparthotels and property management
Location
Spain
Scope
Internal group ERPAI invoice readingReview next to the originalExcel and accounting link for a3conSupplier accounts it remembersWater and power use per supply point
Benefits
  • Company, supplier, tax ID, number, dates, bases, VAT, withholding and total arrive already filled in
  • Each invoice goes to the right company through the recipient's tax ID, with no manual picking
  • Nothing reaches accounting until a person has verified it
  • Each supplier's account is remembered per company and applied to later invoices
12

a3con Excel columns filled in automatically, from date to withholding

1

Row per VAT rate on the invoice, as the importer requires

2

Output formats: importer Excel and accounting link

0

Unverified invoices that can reach a3con

Context

The group runs aparthotels through several companies and manages them in its own ERP: accounting, tax, contracts, properties and people. The official books stay in Wolters Kluwer's a3con.

Between the two sat a repetitive job: every laundry, cleaning, power, water or maintenance invoice has to end up in a3con, under the right company and account. The screenshots use a sample group, suppliers and invoices.

How it works, step by step

Upload
PDFs or photos are dropped into the app, many at a time.
Read
AI extracts issuer, recipient, number, dates, bases per VAT rate, withholding and total. On power and water bills it also picks up the supply reference, the consumption and the period.
Assign
The recipient's tax ID says which company it belongs to. The supplier's account is remembered from earlier invoices, or proposed from the reading if there is none yet.
Review
Each invoice appears next to the original. What a3con requires blocks verification; VAT or total mismatches and possible duplicates raise a warning.
Export
One button produces a file per company with the verified invoices: the a3con importer Excel or the accounting link, depending on how each company works. Invoices move to exported, and reopening one lets you fix it and export it again.
List of received invoices grouped into to review, verified and exported
Received invoices: to review, verified and exported
Export window for a3con with the selected company and the twelve Excel columns
Export: one file per company, with the twelve columns a3con asks for
Generated Excel with one row per invoice: dates, description, number, tax ID, supplier, account, base, VAT and withholding
The Excel it generates, ready for the a3con importer

The hard parts

Reading well isn't enough: the result has to pass a3con's checks. Before generating the accounting link, the app sets aside what a3con would reject, with the reason, such as a tax ID with a wrong check digit, a zero base, a mismatch or a duplicate. Then it checks the final file. Only what was included moves to exported.

The app learns, not the model. Habits come from the latest verified invoices from the same supplier, so correcting the history corrects what it learned. What's printed on the invoice always wins, and an unusual VAT rate only raises a warning.

An invoice is a document from a third party, so its content is treated as data and never as instructions for the AI.

Result

The accounting work shifts from typing to reviewing. Water and power bills also build a table of consumption and price per supply point and month, which would otherwise have to be put together separately.

Next step

Is your team typing into the accounting software what the invoice already says? Let's talk about automating it.

More case studies