Solution
AI invoice capture and bank reconciliation
Incoming invoices are read automatically, pre-coded and, once approved, posted to your ERP. Incoming payments from the bank are matched to open items. The solution is aimed at businesses and fiduciaries whose standard software cannot cope with multiple entities, their own coding rules or the existing ERP.
Who it is for: For SMEs and fiduciary firms handling around 200 invoices a month or more, with Abacus, Business Central, Sage or BMD.
Starting point
How the process runs today
Invoices arrive as PDFs by email, on paper and through portals. Someone opens them, keys in supplier, amount, VAT and reference, looks up the right account and cost centre, and files the document. At month-end, bank statements are compared with open items line by line. It works, but it ties qualified people to routine work.
What it costs
- Four to six minutes per invoice: at 500 invoices, that is about a working week every month.
- Typing errors in amounts and references that only come to light in the dunning run.
- Early payment discounts are lost because invoices wait too long for approval.
- The close is delayed because reconciliation is bunched up at month-end.
Process
What the process looks like afterwards
01
Intake
Invoices from the mailbox, the scanner and supplier portals come together in one place.
02
Extraction
The AI reads supplier, amounts, VAT, QR reference and line items, and rates its own confidence.
03
Pre-coding
Account, cost centre and tax code are proposed based on your rules and previous postings.
04
Approval
Confident invoices go straight through. Deviations are shown to the person responsible, with the reason.
05
Posting
The invoice and its image are posted to the ERP and archived.
06
Bank reconciliation
Bank statements in camt.053 format are imported daily and matched to open items.
Integrations
Which systems are connected
The posting logic stays in your ERP. We deliver pre-coded invoices through the existing interface and collect bank statements via EBICS or your bank's file export.
- ERP and accounting
- Abacus
- Microsoft Dynamics 365 Business Central
- Sage
- BMD
- bexio
- Banking and payments
- camt.053 and camt.054
- Bank connection via EBICS
- QR-bill
- Documents and archive
- DMS and archive
- AI and automation
- Azure OpenAI
Make or buy
Off-the-shelf or custom?
Off-the-shelf software is enough if
- You run a single entity with a common chart of accounts.
- Your accounting software has its own invoice recognition that covers your cases.
- Volume is below 200 invoices a month.
Finished products for this case: bexio, AbaNinja, KLARA and Abacus DeepBox offer invoice recognition as a ready-made product.
Custom is worth it if
- Several entities or clients with different coding rules.
- Cost centres, projects or orders have to be assigned per line item.
- Approvals pass through several people with amount limits.
- Your ERP has no usable recognition and switching is not up for discussion.
Effort
Price range and duration
- Price range
- CHF 25'000 to 60'000
- Duration
- 8 to 14 weeks
Non-binding estimate, excl. VAT.
What drives the effort
The effort is driven by the number of entities, the coding rules and your ERP's interface. On top of that come model costs of a few centimes per invoice.
The price becomes binding after the fixed-price discovery phase.
Your case in numbers
- Hours per year tied up
- 500
- Cost per year today
- CHF 35'000
- Possible saving per year
- CHF 26'300
375 hours freed up for other work
Rough calculation based on your figures. Not a quote.
AI invoice capture and bank reconciliation
Frequently asked questions
How reliably does the AI read our invoices?
Header data such as supplier, amount, date and reference is read very reliably from clean PDF invoices, less so from scans and handwriting. That is why, in the discovery phase, we measure the accuracy rate on a few hundred of your real invoices. Anything below an agreed confidence score goes to your accounts team for checking.
Does this work with our accounting software?
Abacus, bexio, Business Central, Sage and BMD offer documented interfaces through which postings and document images are handed over. For other systems, we check in the discovery phase whether postings can be transferred by API, by import file or through the vendor. The account coding logic stays in your hands.
Where does our data go when AI is involved?
We set that out in writing for each project. The default is processing in a European region, under contracts that rule out any use of your data for training. For sensitive data, self-hosted models are an option. Hosting and database are located in the EEA, and processing is designed around the GDPR and the Swiss Federal Act on Data Protection.
Let us talk about your case.
20 minutes are enough to find out whether a custom solution is worth it for you.