The same AI engine that runs your outreach also runs your back office. It reads every mailbox, understands each document, keeps your customer and supplier records current, issues invoices on schedule and matches bank statement lines to the right invoice — with every send and every write-off still passing through a human.
Finance work starts with hunting for documents. This part stops being work.
Company mailboxes are scanned end to end and every arriving document is classified by the AI — invoices, e-invoices, receipts, reminders, statements and credit notes among them, with the set of types open rather than fixed. Nothing is missed because it landed in the wrong folder.
The counterparty is read out of the invoice itself, so a bill forwarded by a colleague is still filed under the company that issued it — not under the person who pressed forward.
A single message carrying thirty-six PDFs becomes thirty-six separate records, each with its own number, amount and due date, instead of one attachment nobody ever opens again.
Photographed and scanned invoices are read visually, and documents can be pulled in from outside sources such as Gmail or the LinkedIn Admin Center in a single click.
Half-downloaded and corrupted files are caught rather than filed, and anything the system could not verify carries a visible "not verified" badge instead of quietly entering your numbers.
Nobody types a customer card again. The archive already contains one.
Contacts, invoice history and documents are gathered from the mail archive into one card per company — NDAs and contracts are discovered and attached on their own.
Invoicing frequency, payment terms, recurring amount, currency and recipient list are inferred from the actual history, and every derived value can be corrected by hand.
The same company written four ways is recognised across domain, near-identical names and legal suffixes — A.Ş. and Anonim Şirketi, d.o.o. and GmbH — and merged into one record.
Which side of the relationship a company sits on is inferred from the direction of the invoices, not from a field somebody has to remember to set.
From the agreed frequency to a finished PDF — with a person on the send button.
Each customer’s frequency produces a six-month forward schedule: who is invoiced when, and whether each one is ready or still waiting.
The invoice is generated on your real template, with the VAT table, the ZDDV-1 clauses, your bank details, the signature and the stamp already in place.
Every invoice is previewed before sending, and sending is always a manual confirmation — with fixed CC rules so invoice@ and management are on the record every time.
The month-end reconciliation that eats a day becomes a review of the exceptions.
Bank and card statements arrive as PDF or CSV, and each line is matched against the invoices in the system on number, amount, currency — bridging exchange rates — date and vendor name.
A document can only settle once. Certain matches are applied first and anything less than certain is labelled "probable" rather than dressed up as a fact.
An unmatched line can have a document attached, be marked as having no invoice, or be undone — three buttons instead of a spreadsheet.
Download the invoices behind selected lines as a ZIP renamed after the statement lines, and the full statement as a colour-coded Excel file.
Paid, partial, unpaid and overdue are tracked per invoice, and the invoice-payment-refund chain shows up on the customer card as debit, credit and balance.
Book a 30-minute demo. We run it against a real mailbox of yours and show you the documents it finds, classifies and matches.
Book my demoTell us about your sales team and target markets. We'll reply within one business day — usually much faster.
📧 [email protected]
☎️ +386 41 441 757
🏢 Meta Smart Factory d.o.o. — Dunajska cesta 106, 1000 Ljubljana, Slovenia
🌐 www.metasmartfactory.com
💼 LinkedIn · YouTube
🌍 Partner network in Turkey, Saudi Arabia, Pakistan, Brazil, USA, Canada, UK, France, Croatia, Serbia, South Korea, Indonesia, Mexico & more
💬 Or just ask the chat assistant in the corner →