Tax Firm Automation: DATEV Implementation for Mid-Market

Yes. We work with DATEV Unternehmen online, DATEV DMS and the import of booking batches in DATEV format. Documents, journal entries and master data move along those paths. For clarity: we are not a certified DATEV partner and we do not build an official DATEV interface. We use the documented import and export routes and the access your firm already has – and we do not work around restrictions DATEV sets.

They go into a clarification queue, not into the booking batch. A document counts as unclear when the creditor is unknown, when one invoice carries several tax rates, when it is a reverse-charge case, or when the invoice number has appeared before. That queue is the daily task of a named person, not a silent background retry. Nobody should end up with a wrong tax key because the automation guessed.

Your firm does – deliberately. The mapping table is professional logic, not technology: which creditor posts to which account, when a deviating tax key applies, how intra-EU acquisitions are handled. We build the table with you, make it maintainable without us and log every change. Who owns it is decided before we build – otherwise it goes stale within three months and the clarification rate climbs back up.

Processing runs on EU servers, typically on self-hosted n8n or on Make.com with EU data residency. Client data does not leave the European legal area. Before we build, we agree which fields actually need to enter the pipeline, which travel only as a reference, how long logs are kept and who has access. A data processing agreement is part of that. Responsibility for your client data under GDPR stays with you.

The engagement is set up as a six-week effort, and a first automation is typically in production after 4 to 8 weeks. Week one captures the current process, then the first pipeline is built and tested on real documents from two or three client accounts. Only once the clarification rate is stable do further clients follow. What drives the timeline is almost always master data quality, rarely the build.

No. The pipelines run in the background and the work still happens in DATEV and the DMS. What changes is the role: instead of retyping fields, your team works the clarification queue and maintains the mapping table. We train two or three people for that and leave documentation covering how to spot a stalled pipeline, how to pause it and how to onboard a new client account.

Costs depend on the scope. A starter package for document processing starts at €5,000, projects with multiple system integrations at €15,000. Ongoing tool costs are EUR 50-200/month. The price covers a scope defined up front. If additional systems or client-specific edge cases appear along the way, they are quoted as a separate step rather than absorbed quietly.

It does not replace professional judgement. The pipeline proposes an account and a tax key according to your rules; it does not assess a case. It files nothing with the tax office and signs no return. And it does not repair poor master data: if a client keeps four creditor numbers for the same supplier, that gets cleaned up first or stays a permanent clarification case. The duty to review remains entirely with you.

Failures do not disappear silently. Technical outages – target system unreachable, mailbox refusing access, export format changed – are reported by email or Slack to a named person, with the affected item and a timestamp. Professional edge cases go to the clarification queue. On top of that a daily reconciliation reports how many documents arrived, how many were processed and how many are open. Who acts on a failure is agreed before go-live.

That is the usual route. We take one client account with high document volume and a clean chart of accounts, build the pipeline there and watch the clarification rate over two accounting periods. If it holds, we roll out client by client – each further onboarding then needs mapping work, not build work. If it does not hold, you see that after two weeks on real numbers rather than after six months.

No, that is explicitly not the model. After go-live the workflow, the credentials, the mapping tables and the documentation belong to your firm. We prefer to build on n8n, which you can host yourself, so operation does not hang on our contract. If you never call again, the pipeline keeps running. We are happy to take on extensions selectively, but they are not a condition for running it.