Product · BankingAI
BankingAI automatically reconciles your bank statements (CAMT.053) with the open invoices in your FileMaker ERP: invoice references are detected, payments matched, and bookings happen only on an exact match — differences come to you for review.
Deterministic and auditable — same input, same result.
The problem
In manual reconciliation, accounting opens every bank line, reads the payment reference, finds the invoice in the ERP, checks the amount and books by hand. With batch payments, credit notes and discounts, that quickly turns into detective work.
BankingAI takes over the routine — and only asks when something is off.
Features
BankingAI automates the entire reconciliation — and keeps a human in the loop exactly where judgement is needed.
Bank statements in the Austrian STUZZA format are imported and checked — balance continuity, gaps, duplicates. Every payment is matched to its invoice.
One transfer, many invoices? BankingAI splits batch payments proportionally, nets credit notes and detects discounts and bank fees — the cases where automation usually fails.
BankingAI reads open outgoing and incoming invoices from FileMaker and books payments straight back — including the paid status.
Payment advice PDFs are read out — including scans, via local text recognition. And outgoing SEPA files from FileMaker are validated by the QA tool before the banking upload.
Invoice prefixes, colours and the instance name are configurable per company — BankingAI runs in production at an Austrian commercial-vehicle dealer today and can be adapted to further companies.
Trust & privacy
When it comes to money: traceable or not at all. BankingAI works deterministically — same input, same result — and entirely on your own server.
Automatic booking happens exclusively on an exact amount match. Everything else comes to a human with a difference analysis.
Rules instead of a model: every match can be recalculated, every booking justified — including towards your tax advisor.
Banking data, database and text recognition run in-house. No cloud, no external service — account data never leaves the company.
Login with roles, encrypted connections, access limits and logging — operated on your own server.
Deviation reasons such as discount, bank fees, rounding or currency difference are documented on the item.
Included in the system
BankingAI covers the whole journey — from the bank statement to the closed invoice. These building blocks are included as standard.
STUZZA profile, balance and gap checks
prefix variants configurable per company
proportional split with credit-note netting
detected automatically, handled correctly
payments written back, status set
read PDF advices, scans via local OCR
validate pain.001 files before upload
for reconciliation, review and tax advisor
In production: BankingAI continuously processes the statements of two bank accounts of an Austrian commercial-vehicle dealer — including a year-end change of invoice numbering.
Technology
BankingAI is a lean web application (Python/Flask) with PostgreSQL, operated on your own server. Amounts are calculated decimal-exact throughout — no floating-point surprises with money.
Detection is deliberately rule-based: reference patterns, tolerances and thresholds are configured, not trained. Scanned advices are read by local text recognition — here too, no document leaves the building. BankingAI itself is developed AI-assisted throughout.
No. Automatic booking happens only on an exact match between payment and invoice (tolerance €0.01). Every difference comes with an analysis for review — you decide, BankingAI books.
Bank statements in the CAMT.053 format following the STUZZA profile of Austrian banks — the standard your online banking exports. Outgoing SEPA files (pain.001) are validated by the QA tool. Further formats are feasible; talk to us.
With you. Database, processing and text recognition run entirely on your own server — no cloud, no external services, no transfer to third parties.
That is exactly what BankingAI was built for: batch payments are split proportionally across the referenced invoices, credit notes are netted, discounts and bank fees detected — all documented traceably.
The matching is deliberately rule-based and deterministic — same input, same result, no hallucinations. With money, that is a feature. Scanned advices are read by local text recognition; BankingAI itself is developed AI-assisted throughout.
The ready-made, bidirectional integration exists for FileMaker (Data API). The reconciliation itself is built separately from it — other ERP connections are feasible as a project.
Reference & contact
BankingAI continuously reconciles the bank accounts of an Austrian commercial-vehicle dealer with its FileMaker ERP. If your accounting still matches every bank line by hand: tell us about your process — we will tell you honestly how much can be automated.
BankingAI is a software product by Pabeschitz Computer Consulting GmbH. It runs entirely on-premise; automatic bookings happen exclusively on exact matches, and all other cases are presented for review. No tax or legal advice.
