Why this matters. Digital process automation separates process logic (what runs) from execution technology (how it runs). Without that line, you buy tools before knowing what you'll do with them. Mittelstand-Digital data shows around 75 percent of German companies use RPA — yet most pilots still miss ROI.
When RPA fits — and when re-engineering is cheaper
RPA fits when:
-
the applications expose no APIs and interfaces cannot be retrofitted;
-
the process is stable — the input masks don't change every month;
-
the volume justifies the investment — bot licences, maintenance and monitoring are not small numbers.
Re-engineering is cheaper when:
-
the application is modern and APIs are available — the bot would be an expensive workaround;
-
the process needs rework anyway because it has grown in complexity rather than value;
-
the total cost of ownership over three years exceeds a re-engineering investment.
How to structure digital process automation in the mid-market
Digital process automation follows a method that is independent of the tool. You can deliver it with RPA, a workflow engine or an ERP extension — the sequence stays the same.
-
Phase 1 — Business case: monthly volume, handling time per case, error rate, share of exceptions. Output: a defensible business case with three scenarios.
-
Phase 2 — Target process: modelled in BPMN 2.0, exceptions documented, escalation paths defined. Output: a target process that works without a bot.
-
Phase 3 — Tool choice: API audit per involved application, then a decision on RPA bot, workflow engine or re-engineering. Output: a written tool decision with rationale.
-
Phase 4 — Audit trail: every step the bot takes logs structured data for GoBD and GDPR. Output: an audit trail that withstands external review.
The typical mistake — and how to avoid it
A southern German specialty services firm with around 140 employees commissioned a consultant who arrived with RPA licences in hand. The tool was bought; the target concept was left open. Six months later a bot was running — but it stopped every time the upstream application updated its masks. The annual RPA licence cost exceeded the manual handling time the bot had been built to replace.
Two questions decide outcomes before any tool is bought: which applications are involved — and which expose APIs? When APIs are available, a workflow solution inside the ERP estate is usually cheaper than an RPA bot. Bitkom guidance: end-to-end view requires ownership, not licences.
What to do next
If you want to validate the tool decision methodologically:
-
List the applications involved and check API availability for each.
-
Build a three-year TCO — licence, maintenance, monitoring, fallback path.
-
Compare that TCO honestly to a re-engineering scenario before commissioning bot licences.
If you want to assess that tool decision vendor-neutrally, we will sit through the TCO with you in twenty minutes — methodology-led, not product-led.
FAQ
The discipline of replacing a business process with a documented technical implementation — typically through a workflow engine, BPMN model, RPA bot or ERP extension. It covers the business case, target concept, tool choice and audit trail — not just the bot.
RPA emulates human clicks on a user interface — typical when the application has no API. A workflow engine drives process steps through API calls — typical when applications are modern and integrable. RPA is tactical, workflow engines are strategic.
When process volume is measurable and handling time per case is five minutes or more. Below that threshold, maintenance and monitoring costs typically exceed the saving — at which point re-engineering or an ERP extension is the better lever.