I've spent years getting certified in RPA, but now I see AI Agents that can navigate UIs dynamically without needing pre-defined selectors or brittle workflows. Are tools like UiPath and Blue Prism dead, or is there a way these legacy systems integrate with the new "Agentic" approach for enterprise-scale automation?
3 answers
RPA isn't dead; it's evolving into "Agentic RPA." Traditional RPA is great for high-volume, repetitive tasks where the UI never changes. It's deterministic and fast. However, AI Agents excel where the UI is dynamic or the logic requires reasoning. In 2025, the industry is moving toward a hybrid model. You use RPA for the "heavy lifting" of data entry into legacy systems and an AI Agent to handle the exceptions or the decision-making parts of the workflow. UiPath has already integrated "Autopilot" features that essentially turn their robots into agents that can handle unstructured data.
Dorothy, do you think the skill set for an RPA developer needs to shift more toward Python and LLM prompt engineering then, rather than just dragging-and-dropping blocks in a GUI?
RPA is for the "rails" and AI Agents are for the "off-roading." Most enterprises still have a lot of rails that need maintaining!
That's a great way to put it, Nancy. Many of our banking clients still rely on legacy mainframe systems where RPA is the only way to interface. Agents are just the new "smart driver" for those old vehicles.
Absolutely, William. The "low-code" era is merging with "pro-code." Knowing how to write custom API integrations and how to "ground" an LLM using RAG is becoming as important as knowing how to handle a selector in a web browser. I’d suggest looking into LangChain and Semantic Kernel as they are the new "orchestration" layers for modern automation.