DATEV

AI invoice processing for DATEV in 2026: what actually works

A practical guide to AI invoice processing with DATEV: capture, SKR03/SKR04 coding, approvals, e-invoices, and GoBD-aligned posting without CSV exports.

Bruno & Seb, co-founders, Cortena.ai9 min read

German finance teams searching for KI Rechnungsverarbeitung DATEV are usually not looking for another OCR demo. They need invoices to land in DATEV Unternehmen online or Rechnungswesen with the right SKR account, cost centre, and BU key, with a trail their Steuerberater can trust. This guide explains what works in 2026, and where Cortena.ai fits.

Why DATEV invoice processing still eats the week

Capture is rarely the bottleneck anymore. The slow part is finishing the job: missing supplier details, wrong cost centres, BU keys that fail validation, and re-keying the same invoice into DATEV after someone already touched it in email or a shared drive. At a few hundred invoices a month, that work compounds into days of productivity lost every week.

What good AI invoice processing for DATEV must cover

  • Capture from email, PDF, and e-invoices (ZUGFeRD / XRechnung), with duplicates caught early.
  • Line-level coding to SKR03 or SKR04, cost centre, and BU key, with a readable reason on each field.
  • Approval routing that matches how your company actually signs off.
  • Official DATEV posting (Rechnungsdatenservice), document attached, no CSV shuttle.
  • GoBD-aligned audit trail, with the archive staying in DATEV.

E-Rechnung and DATEV in the same flow

Since 2025, German B2B buyers must be able to receive e-invoices. Structured formats help extraction, but they do not finish bookkeeping. Cortena.ai reads XRechnung and ZUGFeRD field by field, then continues into coding, approvals, and DATEV booking so the mandate does not become a second inbox problem.

How Cortena.ai approaches DATEV

Cortena.ai is the AI operator for accounts payable. You brief it like a bookkeeper. It codes each line, shows why, asks when unsure, and exports booking-ready batches into DATEV. After a short learning period, roughly 90% of volume is handled end to end, about 30 seconds per reviewed invoice for the rest, with white-glove setup and a typical go-live in 15 days.

What to ask any DATEV AI vendor

  • Do you post via the official DATEV service, or via file export?
  • Can we see the reason on every coded field, not only a confidence score?
  • Where is data hosted, and is customer data used to train models?
  • Will you run a pilot on our real supplier mix, including multi-line invoices?

Frequently asked questions

The short answers below match the FAQ schema on this page. For a deeper product walkthrough, open the DATEV integration page or book a demo with a week of real invoices.

  1. It means invoices are captured, coded to your chart of accounts (SKR03/SKR04), routed for approval, and posted as booking-ready entries into DATEV Unternehmen online or Rechnungswesen, without manual re-typing or CSV hand-offs.

  2. No. Cortena.ai automates upstream bookkeeping work. Your tax advisor still receives clean, pre-coded data and focuses on closing, tax, and advice.

  3. Cortena.ai keeps an audit trail of capture, coding, human changes, and export. The archive stays in DATEV. Hosting is in Germany. We describe this as GoBD-aligned, never as a certification.

  4. White-glove setup is included. Most teams connect DATEV and are live in about 15 days, with roughly 90% of volume coded after a short learning period.

  5. Still have a question?

See Cortena on your own invoices.

The fastest way to understand what Cortena does is to watch it run on your real books.

Book a demo