Why this matters. Process automation replaces repetitive work with documented, auditable flows. You cannot automate a broken process; you can only repeat it faster. The Bitkom AI Study 2026 shows 41 percent of German companies now use AI — yet most pilots still fail on missing process ownership.
When this methodology fits — and when it doesn't
This methodology fits when:
-
the process is written down before it is automated — not afterwards;
-
a process owner with deputy cover is named, with authority to approve changes;
-
the business case is measurable — handling time, error rate, monthly volume.
This methodology does not fit when:
-
the process runs differently every month because the rules live in heads rather than the system;
-
the pilot is run as an IT project, with no business-side ownership;
-
no documented manual fallback exists in case the bot fails or the interface breaks.
How to structure process automation — the four stages
Durable process automation follows a maturity arc, not a tool. The Bitkom Technology Playbook is clear: end-to-end perspective comes before tool choice. Methodology decides — not the software.
-
Stage 1 — Discover: process mining or structured shadowing. Which steps run today, where do queues build up, which variants actually exist. Output: a defensible as-is picture.
-
Stage 2 — Standardise: reduce variants, document exceptions, write the rules down. Output: a target process that works without a bot.
-
Stage 3 — Automate: RPA bot or workflow engine — driven by whether the applications expose APIs. Output: a productive bot with documented fallback.
-
Stage 4 — Orchestrate: chain multiple bots into an end-to-end flow, set up monitoring, drift detection. Output: a measurable end-to-end path.
The typical mistake — and how to avoid it
A southern German machinery builder with roughly 220 employees launched an RPA pilot for order confirmations. The bot read emails, extracted line items, posted orders — until one customer changed its PDF format. The bot sat silent for three weeks because no one owned the error queue. Clerks worked manually in parallel — and the system lost trust.
In practice, three disciplines decide success: a documented manual fallback, named process ownership with deputy cover, and monthly drift review. Mittelstand-Digital experience confirms it: repetitive admin work is the ideal entry point — only when the process stands cleanly first.
What to do next
If you want to test whether your next automation candidate will hold:
-
Write the process down in fifteen steps or fewer — without naming any software.
-
Name a process owner with deputy cover.
-
Document the manual fallback before you commission the bot.
If you want to validate these three steps without vendor lock-in, we will sit through the methodology with you in twenty minutes — the ERP selection methodology follows the same maturity arc, only at a larger scale.
FAQ
Process automation is the discipline; RPA is one tool. Process automation can run through workflow engines, BPMN models or RPA bots — the choice follows API availability. RPA emulates human clicks where no interface exists.
Start with a documented process, not a tool. Pick a recurring admin flow with measurable volume — invoice approval, order entry, master-data maintenance — and write it down in fifteen steps or fewer. Only then decide whether RPA, a workflow engine, or process re-engineering is the right lever.
The BMWK Mittelstand-Digital programme supports SMEs with neutral, no-fee digitalisation advisory. The current funding period ends 31 December 2026 — a successor programme is planned from 2027. State-level funding and the BMWK "Digital Jetzt" investment grant programme cover concrete automation projects.
