Invoice-to-bank-transaction matching is one of the most consistent drains on a finance team at month-end. The work is highly repeatable, yet most organizations still do it manually, line by line, chasing statement exports from accounting systems that closed yesterday.
Why invoice reconciliation is the quiet bottleneck
Controllers at European SMEs tell us the same thing: they know the data is there, they just cannot get to it fast enough. By the time the bank CSV arrives, matches need to be worked through from scratch. The information was never the problem. The latency was.
Live bank data across 1,000+ European banks
Through our partnership with Yapily, Cortena connects to more than 1,000 European banks. The system pulls live transaction data directly, without requiring CSV exports or batch uploads. When an invoice is approved and sent to DATEV, the matching agent already has the bank view it needs.
How agent-based reconciliation closes the loop
The workflow runs agents in sequence. One agent pulls and normalises bank transactions. Another matches them against open invoices using amount, date window, supplier reference, and IBAN. Mismatches are not dropped. They are flagged with the agent reasoning included, so a reviewer can see exactly why the match failed and resolve it in one step.
The result is that exceptions surface with context rather than as unexplained gaps. Finance teams stop chasing and start deciding.
Real numbers: 5,000+ invoices a month at close to 90% accuracy
As of March 2026, the platform processes over 5,000 invoices monthly. Close-to-90% accuracy means the remaining 10% surfaces as traceable exceptions, each with reasoning attached. That ratio holds across different customer ledger structures because the matching logic adapts to each company's supplier patterns over time.
"I've never seen a tool so well integrated with DATEV. It does do all the pre-accounting work." · Dimas, Head of Finance, Proxima Fusion
Workflow Builder: automation that matches your process
Workflow Builder, released this month, lets finance teams configure approval chains, escalation rules, and conditional logic without writing code. A team can model their existing process in the visual builder, then let the agents run it. The process does not change to fit the software. The software fits the process.
What this unlocks next
Live banking data combined with deterministic agents and configurable workflows forms an execution layer. The roadmap from here runs into spend tools and receivables automation. Each step adds more of the operational cycle to the layer that runs on its own.