RPA — Robotic Process Automation
Robotic Process Automation (RPA) describes software robots that mimic human interaction with applications — clicking, typing, reading screens — to automate repetitive tasks. RPA peaked in attention around 2018-2020 as a quick-win automation play; since 2022 it has converged with AI and broader hyperautomation platforms. For ERP-centric organisations, RPA remains a useful tool for filling automation gaps where APIs are not available or where legacy systems interact with modern ERP.
Common ERP-adjacent use cases
- Supplier-invoice processing — reading invoices, validating against POs, posting to AP. Today often replaced by AI-document-AI
- Customer-order entry from email or PDF orders into the ERP sales-order screen
- Master-data updates — bulk customer or supplier record changes across multiple systems
- Bank reconciliation downloading bank statements and matching to ERP open items
- Legacy-to-modern bridges — pulling data from an old AS/400 or mainframe screen into the modern ERP
- Report distribution — scheduled ERP report generation and email distribution to stakeholders
- Period-close tasks — scripted financial close checklist items
Leading RPA platforms
UiPath — market leader, broad enterprise presence in DACH. Strong AI features through Document Understanding and recently Agentic Automation. Cloud subscription from 1,500 EUR per bot per month. Microsoft Power Automate Desktop — bundled with Microsoft 365 E5, gaining rapid market share for entry-level RPA in DACH. Automation Anywhere (AA) — established enterprise platform, particularly in financial services. Blue Prism (now SS&C Blue Prism) — original enterprise RPA pioneer, particularly in banking and insurance. WorkFusion, NICE Robotic Automation, Pegasystems Robotic Automation — further enterprise options. Open source: Robocorp, OpenRPA, Robot Framework. For mid-market DACH operations starting RPA, Microsoft Power Automate is the most common entry point through Microsoft 365 bundling.
Convergence with AI and IDP
The RPA market is rapidly converging with adjacent automation categories. Intelligent Document Processing (IDP): Rossum, Hyperscience, ABBYY FlexiCapture, Microsoft AI Builder handle invoice and order extraction with LLM-augmented accuracy — replacing classical RPA for these high-volume use cases. Agentic AI: UiPath Autopilot, Microsoft Copilot Studio, OpenAI Operators move beyond scripted RPA toward AI agents that adapt to changing interfaces. Process Mining + RPA: process mining identifies automation candidates; RPA implements them. Celonis EMS plus UiPath integration is the most-evangelised pattern. For mid-market in DACH, the practical implication: do not over-invest in RPA scripts that AI document understanding will eat within 2-3 years.
Governance pitfalls
RPA suffers from the same governance challenges as low-code: easy to start, hard to operate. Unmaintained RPA bots accumulate across departments, breaking silently when underlying applications change. ERP upgrades regularly invalidate dozens of RPA scripts. Best practice: establish an RPA Centre of Excellence (CoE) within IT, document every bot with its triggers, business owner, criticality and expected lifetime. Schedule quarterly reviews to retire bots whose use case has been automated through native ERP features or AI document processing. Treat RPA as tactical automation with explicit sunset planning, not as a permanent platform.
Related Topics
Frequently Asked Questions
Is RPA still relevant in 2026?
Yes, but narrower than during the 2018-2020 peak. Use cases involving structured screen interaction with legacy systems remain valid; high-volume document processing has largely shifted to AI-based intelligent document processing. New RPA investment should focus on automating workflows that cannot be solved through native ERP features, APIs or AI in 12-24 months.
How does RPA compare to API-based integration?
API integration is more robust, faster and easier to maintain than RPA. Prefer API integration whenever the source system exposes APIs. RPA is the fallback for systems without APIs (legacy mainframes, vendor portals, older Windows applications) or for processes that span human approval steps within larger workflows.
What is a realistic ROI for RPA?
Successful RPA programmes report 3-7x ROI over 3 years through productivity gains and error reduction. Failures typically come from: automating too few processes (under 10 bots) so the platform cost dominates, or building unmaintainable bot estates that grow brittle. Pilot ROI is rarely a guide to programme ROI — the difference between proof-of-concept and stable operations is significant.