Building Decision Agents with LLMs & Machine Learning Models

In this monologue podcast, the speaker discusses the importance of decision agents in agentic AI for solving complex problems, highlighting the limitations of large language models (LLMs) for this purpose due to their inconsistency, lack of transparency, and inability to process historical data effectively. The speaker advocates for using decision platforms or business rules management systems instead, emphasizing their consistency, transparency, agility, and low-code environment. The podcast further explores how to build stateless, side-effect-free decision agents using decision platforms, which include editors, repositories, validation tools, testing mechanisms, and deployment engines. The discussion extends to incorporating probabilistic elements using machine learning platforms and enhancing decision agents with LLMs for data ingestion and explanation of results, concluding with strategies for continuous learning and improvement of both analytic and decision agents within the agentic AI framework.