Automating DATEV Processes in 2026: Interfaces, AI & GDPR

Quick answer

In 2026 you automate DATEV processes in a tax firm by connecting document flow, client communication and booking preparation through DATEV interfaces (DATEVconnect, DATEV XML-online, DUO/"Unternehmen online") with a workflow layer (Make or n8n) and a client portal (for example Softr). AI-based document pre-capture delivers suggestions (OCR + account assignment logic) that a specialist approves — fully automatic posting only happens in clear, rule-based cases. It stays GDPR-compliant when personal and tax data are processed in the EU, a data processing agreement exists with every tool provider, and AI models do not use client data for training.

Which DATEV interfaces matter for automation

DATEV is not an open API ecosystem like a modern SaaS, but there are clearly defined docking points. Anyone who wants to automate should know these four:

• DATEV Unternehmen online (DUO): The central document store between client and firm. Documents arrive here via upload, app photo or incoming invoice and are taken over into Rechnungswesen pro. The ideal place for automation to start upstream (collect, check and tag documents).
• DATEVconnect: A local REST interface on the firm's workstation or server. It gives programmatic access to master data, documents and accounting records — technically the cleanest basis for real integrations.
• DATEV XML-online / Rechnungsdatenservice 1.0 & 2.0: Standardised import of structured booking and document data (XML in DATEV format). Perfect for feeding in data from third-party systems (ERP, POS system, your own portal).
• DATEV format (ASCII/CSV) & booking batch import: The classic, robust route — generate a booking batch as a DATEV-compliant CSV and import it. Less elegant, but sufficient for many standard cases and quick to implement.

Rule of thumb: For data delivery (into DATEV only), XML-online or the DATEV format is usually enough. For bidirectional automation with status reconciliation you need DATEVconnect or a certified partner connector.

AI document pre-capture: what realistically works in 2026

AI document pre-capture is the lever with the biggest time savings — but also the one with the most misunderstandings. A realistic sequence looks like this:

The typical workflow

1. 1Document intake: The client uploads the invoice in the portal or forwards it to a dedicated email address.
2. 2Extraction (OCR + AI): A model reads out supplier, date, net/gross, tax rate, invoice number and service description. On standard invoices, modern models reach 90–97% field accuracy.
3. 3Account assignment suggestion: Based on supplier history and rules ("supplier X → account 6600"), the system proposes account, contra account and cost centre.
4. 4Professional approval: A staff member reviews it in the portal with a click. Uncertain fields are highlighted.
5. 5Handover to DATEV: Approved documents go into the bookkeeping as a booking batch or via DUO.

Where the line runs

Fully automatic posting (without human approval) is only defensible for recurring, unambiguous documents — telecom invoice, rent, leasing instalment. Anything involving judgement (entertainment expenses, mixed input tax, foreign transactions) stays in the approval loop. The mistake many firms make: they expect 100% automation and end up disappointed. The realistic target is 60–80% automated pre-capture with a human final check — which already saves a massive amount of data entry time.

What is GDPR-compliant — and what is not

Tax data is highly sensitive: revenues, salaries, and in payroll documents partly health data. Professional law additionally imposes the duty of confidentiality (§ 203 StGB, § 62 StBerG). Sustainable automation therefore requires hard guardrails:

• EU data residency: Processing and storage should sit in the EU/EEA. With cloud tools, check the server location and the legal basis for any third-country transfers (EU-US Data Privacy Framework).
• A data processing agreement for every link in the chain: DATEV, the automation platform, the AI provider, the portal provider — each one needs a processing agreement under Art. 28 GDPR.
• No model training with client data: When using AI APIs (e.g. for OCR/extraction), explicitly choose the enterprise/zero-retention variant where inputs are neither used for training nor stored.
• Data minimisation: Extract and pass on only the fields you need. Don't send complete document images to external services when the text content is enough.
• Access logging & roles: Who saw which document and when? A portal with granular roles (clients see only their own data, case handlers only assigned mandates) is mandatory.
• Deletion concept: Configure automation logs and intermediate storage (e.g. in Make/n8n) so that personal payloads are deleted after processing.

GDPR conclusion: The automation itself is unproblematic — what is problematic is uncontrolled data outflow. Anyone who keeps processing on EU infrastructure, manages processing agreements properly and uses AI without training acts in compliance.

These processes are worth automating first

Not everything at once. Prioritise by frequency × time spent × error-proneness:

1. 1Collecting documents from clients: Automatic reminders, portal upload instead of email chaos, completeness checks per month. The biggest time drain in almost every firm.
2. 2Document pre-capture & account assignment suggestions: As described above.
3. 3Deadline and task management: VAT return, payroll tax and annual accounts deadlines automatically into an ops dashboard with a status per client.
4. 4Onboarding new mandates: Data intake, powers of attorney, DUO activation and master data setup as a guided workflow.
5. 5Query management: Open points per document are assigned to the client in a structured way in the portal instead of in bulk emails.

No-code vs. custom: the build decision

The central strategic question is not "which tool", but how much of it the team can own and maintain itself. Three build approaches compared:

1. No-code stack (Softr + Make/n8n)

• What: Client portal in Softr, automation in Make or n8n, database in Airtable or Postgres, connection to DATEV via XML-online/booking batch.
• For whom: Firms with 3–50 employees that want their own adaptable system quickly (4–8 weeks).
• Advantage: The firm understands and changes workflows itself; no vendor lock-in to a rigid firm SaaS.
• Cost (order of magnitude in 2026): €3,000–15,000 to build, then €150–600/month in tool licences.

2. Standard firm software with add-on modules

• What: DATEV's own extensions (DUO, document transfer) plus third-party providers for AI OCR.
• For whom: Firms that want to stay as close to DATEV as possible and do not want to map their own processes.
• Drawback: Little flexibility for firm-specific workflows and client portals.

3. Custom development

• What: Individually programmed integration via DATEVconnect, your own database, your own frontend.
• For whom: Large firms/networks with very specific requirements and their own IT.
• Drawback: Higher cost (from €40,000+), longer timeline, ongoing maintenance.

Recommendation: For most DACH firms, the no-code stack is the best compromise in 2026 between speed, ownership and cost. n8n is particularly attractive because it can run self-hosted in the EU — a GDPR advantage over pure cloud automation platforms. Make scores with faster implementation and more than 2,000 connectors.

Reference architecture for an automated firm

A proven setup that a firm's team owns itself:

• Frontend / portal (Softr): Clients upload documents and see open tasks, deadlines and queries. Role-based.
• Database (Airtable or Postgres): Client master data, document status, assignments.
• Automation (n8n self-hosted in the EU or Make EU region): Orchestrates document intake, calls the OCR/AI extraction, writes suggestions back, generates the DATEV booking batch.
• AI layer (zero-retention API): Extraction and account assignment suggestion.
• DATEV handover: XML-online or booking batch import; DATEVconnect for advanced setups.

This setup can start modularly (document collection first, then pre-capture) and grows with you.

Implementation in 5 steps

1. 1Process audit (1 week): Which 3 processes cost the most time? Measure volume and error sources.
2. 2Data protection setup: Clarify processing agreements, EU hosting and AI zero-retention before any data flows.
3. 3Pilot with one process (2–4 weeks): Usually document collection + pre-capture for 5–10 mandates.
4. 4Fine-tune the approval logic: Define thresholds for when AI suggestions require approval.
5. 5Scale & train the team: Roll out to all mandates and name owners for the workflows.

FAQ

Can DATEV be fully automated without DATEV software?

No. DATEV remains the leading system for bookkeeping and financial statements. What gets automated is the preparatory work — document collection, pre-capture, status tracking — which is then handed over to DATEV in structured form via an interface.

Is using AI for document capture permissible under professional law?

Yes, as long as professional responsibility stays with the licensed professional, confidentiality is preserved (EU processing, processing agreement, no model training) and the human keeps the final approval. AI delivers suggestions, not binding postings.

Which is better for a firm: Make or n8n?

n8n suits you when EU self-hosting and maximum data control are the priority. Make is faster to set up and has more ready-made connectors. Both can be operated in a GDPR-compliant way; the choice depends on IT resources and your data protection requirements.

How long does a first productive automation building block take?

A pilot for document collection and pre-capture is typically productive within 4–8 weeks with a no-code stack, including the portal, the data protection setup and the DATEV handover.

Who owns the system in the end?

With the no-code approach, the firm itself: it holds the licences, the data and the workflows, and can adapt them without a developer. That is the decisive difference from closed standard software.

Mindflows builds these GDPR-compliant firm systems on Softr, Make and n8n — from document collection to the DATEV handover. We design, build and hand over systems your team owns itself.