I’m a traditional RPA developer using UiPath and Blue Prism. With the rise of agentic AI, I’m wondering if I need to dive deep into neural networks and PyTorch to stay relevant. Can I build these "intelligent agents" using existing low-code tools, or is the industry moving toward a "Code-First" approach where Python is the only way to build autonomy?
3 answers
You don't necessarily need to be a Deep Learning researcher, but you do need to understand how to orchestrate Large Language Models (LLMs). The industry is moving toward "Agentic RPA" where the bot doesn't just follow a fixed path but can reason through UI changes using vision models. In 2023, I saw a huge shift toward LangChain and Semantic Kernel. You don't need to build the model from scratch (that's what OpenAI/Google do), but you must know how to connect them to your legacy RPA workflows. Think of yourself as a "System Integrator" for AI.
Melissa, do you think the traditional RPA vendors will catch up, or will startups like CrewAI and AutoGen eventually make UiPath obsolete for complex automation?
I've already started learning Python. The flexibility you get with custom agents far outweighs what you can do in a drag-and-drop interface.
Smart move, Danielle. Python and APIs are the glue of the modern AI ecosystem. It's definitely worth the effort.
Steven, I think they will coexist. UiPath is great for "deterministic" tasks like moving files. CrewAI is better for "probabilistic" tasks like research and summarization. The winning strategy is to use RPA for the plumbing and AI agents for the brains. Most big companies won't ditch their expensive RPA licenses overnight, they'll just augment them.