What is an Agent?

The article explores the concept of "agentic" AI systems, emphasizing their distinction from traditional generative AI models. An agentic system employs a reasoning model, typically a large language model (LLM), to determine and execute actions by interacting with various tools, creating a dynamic decision-making loop. This approach contrasts with non-agentic systems, which rely on static automation or single LLM outputs. The text highlights the challenges and potential of agentic systems in exceeding the capabilities of copilot-like systems, which perform specific tasks with human oversight. It discusses the evolving landscape of AI, focusing on the balance between autonomous and collaborative agents, and the importance of integrating human feedback into agentic workflows. Despite the promise of these systems, the article cautions against over-reliance on current AI capabilities, advocating for a realistic assessment of technology readiness and the importance of maintaining human oversight in complex decision-making processes.