PODCAST: explore the advancement of automation systems beyond traditional Robotic Process Automation (RPA), leveraging Large Language Models (LLMs) and deep learning for greater adaptability and cognitive capabilities. The first paper introduces UINav, a system for mobile UI automation agents designed to execute human tasks on devices like Android phones, achieving high accuracy through an error-driven training process and novel techniques like utterance masking. The second source presents SmartFlow, an AI-based RPA system that combines LLMs with computer vision and natural language processing to generate navigation workflows for automating web application tasks, demonstrating robustness across diverse GUI layouts. Finally, the third paper proposes Text2Workflow, a method that uses a multi-layered LLM prompting strategy to generate executable business process workflows in JSON format from natural language requests, aiming for generalized automation across various tools and APIs and significantly outperforming single-prompt baselines.