Try it yourself
Three documents. Three mistakes.
Can you find them before udX does?
Purchase order, delivery note and invoice for a single purchase. Click anything that looks off. The clock runs while the documents are on screen.
IBAN
What it does
Four things udX does well.
Extraction
Every field, every layout, every format: PDF, scan, email, Word, image. Extracted into structured data with a confidence score per field.
Comparison
Two contracts, ten policies, a notice against the declaration. udX surfaces every meaningful diff and flags what matters.
Automation
Triggered on receipt: extract, validate, route, post. Humans only on the rows the model isn't sure about.
Monitoring
Deadlines, anomalies, missing fields, expiring contracts. udX watches and alerts before things slip.
How it works
From inbox to system of record.
Four stages. Each one inspectable, each one swappable, each one already production-tested.
Ingest
Documents arrive via email, upload, OneDrive, S3, or your existing inbox. udX accepts every common format with no preprocessing.
Understand
OCR and entity detection extract every relevant field. Domain-tuned models score confidence per field, so the engine knows when it doesn't know.
Verify
Anything below threshold lands in the HITL queue: a human approves or corrects it in seconds, and the model learns from that correction.
Deliver
Structured data flows into your ERP, CRM, DMS, or storage of choice: REST, CSV, DATEV, Lexware, SAP, whatever you run.
Integrations
Plays well with what you already run.
Don't see your stack? udX exposes a clean REST API and webhook layer, so anything you can call, you can connect.
Why it works
Built for production, not for screenshots.
Human-in-the-loop, by design
udX never silently approves what it isn't sure about. Every uncertain field is flagged for review, and every correction trains the next decision.
EU hosting, GDPR-first
Hosted in Germany. No data leaves the EU. Encryption at rest and in transit. Full audit trail per document.
API-first, no lock-in
Every action udX takes is reachable via API. Export your data and your processed results in one call, any time.
The udX family
Two products build directly on the engine.
Every document analysed by udX is immediately usable in both: same entity model, same hosting, same human-in-the-loop principle.
Ready to ship it?
45 minutes. We look at three of your real documents. You walk away knowing exactly what udX would do with them, and what it would cost.
Frequently asked questions about udX
What is udX?
udX (Universal Document Intelligence) is Koeniq’s platform that automatically reads, understands and processes any document. Four modules cover extraction, comparison, automation and monitoring, with no templates and no OCR setup. Processing in under 5 seconds at 99.2% field accuracy.
How is udX different from traditional OCR?
OCR only recognises characters and needs a fixed template per document type. udX understands the content regardless of layout: it extracts the right fields even from unknown formats, maps them correctly and detects relationships, without any predefined templates.
How is udX different from RPA?
RPA replays rigid, predefined click paths and breaks the moment a UI or document changes. udX works content-based and stays robust against layout changes, and hands uncertain cases to a human instead of running on incorrectly.
Which documents can udX process?
Invoices, receipts, contracts, insurance claims, tax notices, annual reports, doctor’s letters and application forms, among others. Typical fields are legal, tax and finance, insurance and healthcare, including integration with systems such as DATEV or Lexware.
Where is udX data hosted?
Exclusively in the EU and GDPR-compliant. For uncertain results, udX flags the case for manual review (human-in-the-loop), so no incorrect data enters your systems unnoticed.