Robotic Process Automation (RPA)

RPA uses software “robots” that operate existing applications through their user interface, the way a person would: clicking, typing, copying between screens. It automates the task, leaving the underlying systems and the process design unchanged.

What it is good at

  • Linking systems that have no API (legacy desktop apps, web portals).
  • Repetitive, rule-based, high-volume data entry and reconciliation.
  • Fast, low-disruption pilots: no change to the target application.

Two modes: attended bots work beside a person on their desktop; unattended bots run on servers on their own.

Limits

  • Brittle: bots follow screen layouts, so UI changes break them.
  • Automates the process as is, including waste; a bad process just runs faster (business-process-reengineering and lean say: fix first, then automate).
  • Maintenance cost: scripts grow into undocumented technical debt without governance.
  • Rule-based only: judgement, unstructured documents and exceptions need other tools (see hyperautomation-and-intelligent-automation, ai-and-bpm).
  • Not a replacement for bpm-suites-and-workflow-engines: bots do steps, an engine manages the flow, state and people.

How to use it well

Use process-mining to pick stable, high-volume paths; prefer an API or integration when one exists; keep bots in the same repository and change process as other software.

Vendors

uipath is the best-known platform; other vendors exist. UiPath itself defines RPA as using software robots to automate repetitive, rule-based tasks such as data entry and system integration, says the value is greatest where systems lack APIs, and says a platform should support attended and unattended robots and UI-, API- and AI-based integration. Its agentic-era positioning is covered in its own note.

BPM hub, ai-automation

Open items

  • Vendor landscape not systematically surveyed.

Sources