Your team reviews exceptions. Not every receipt.
Invoices against delivery notes, orders against goods receipts — your AI matches, detects deviations, and routes them. With your logic. On your infrastructure.
This is what automated 3-way matching looks like:
Today's Reality
Invoices are matched line by line against delivery notes and purchase orders — by hand, often across two different systems simultaneously.
From receipt to booking, the average turnaround is 9–12 days. Review, code, approve, forward — every step manual, early payment discounts regularly lost.
Documents arrive as scans, PDFs, and email attachments — and since January 2025, also as mandatory e-invoices (XRechnung/ZUGFeRD). Four different formats, no system handles all of them equally well.
Which suppliers have special conditions, which exceptions exist, how edge cases are decided — 2–3 people know. Nobody else.
How It Works Instead
Ingest
Invoices, delivery notes, purchase orders — as PDF, scan, email attachment, XRechnung, or ZUGFeRD. The system recognizes format, structure, and content automatically. Even with poor scan quality.
Match
The 3-way match runs automatically — purchase order, delivery note, invoice, line by line. Partial deliveries, special conditions, differing article numbers — the system learns your logic and soon knows it better than any handbook.
Route
Correct transactions are passed through. Deviations go to your team — with full context, not flying blind.
Book
Approved transactions are submitted as booking proposals to the ERP. Your accounting team confirms. They no longer enter data.
Especially relevant if:
- You process 200+ documents manually each month — incoming invoices, delivery notes, purchase orders.
- Your ERP manages the master data: SAP, ABAS, proALPHA, Microsoft Dynamics, or comparable systems.
- At least one person spends a significant part of their week on matching and routing — instead of tasks that truly matter.
Under the Hood
Infrastructure
On-premise or EU cloud — in both cases under your control. No US provider, no Cloud Act risk, GDPR-compliant by design.
Integration
Direct connection to SAP, ABAS, proALPHA, Microsoft Dynamics, and other ERP systems — via existing interfaces where possible, new ones where needed. No rebuilding your IT landscape.
Models
Open-source models for document extraction and semantic matching — specialized for your document types and exception logic. You can switch or take over the model at any time.
Your IT Provider Stays
They handle infrastructure. We build the AI application on top. Clear responsibilities from day one — no turf wars.
How It Works
Design Sprint
2 WeeksWe analyze your document process with real receipts from your operations — not demo data. You get: technical solution architecture, business case with ROI numbers, identified risks, and a clear go/no-go recommendation. Fixed price. If the project doesn't pay off, we say so here — before you invest big.
Implementation & Go-Live
4–10 WeeksWe build the system on your data, with your process logic and your exception rules. Your team is involved from the start. When the system goes live, we train your team — on real cases, not theory. Clear scope, clear price.
Operations
OngoingMonitoring, updates, optimization when suppliers, formats, or validation rules change. Either as a managed service by us — or we hand over to your IT with full documentation. No vendor lock-in.
“Omniance has made our internal administration significantly faster and simpler — with excellent return on investment.”
INOS GmbH & Co. KG
Industrial On-Site Solutions
Built by a team with 16 AI patents and 14 years of applied AI development — at Philips Research, Heidelberg University Hospital, and CERN. Our standard: systems must work 10 out of 10 times. Not 9.
Meet the team →In 30 minutes you'll know if it's worth it.
No pitch. We look at your process together — and honestly tell you whether automation makes sense for you. If not, we'll say so.