ARTIFICIAL INTELLIGENCE AGENTS: FROM THEORY TO REAL-WORLD APPLICATIONS

AbstractWe review the theory, architectures, and real-world deployments of artificial intelligence (AI) agents autonomous software systems that perceive environments, plan and act to achieve goals, and interact with humans and other agents. We synthesize foundational models (planning, reinforcement learning, symbolic systems), modern agentic systems built on large language models (LLMs), and evaluations used to measure capability and safety. We survey application areas (healthcare, robotics, customer support, finance), report recent empirical trends, and analyze societal, security, and issues. We close with open challenges and a research agenda to move agents from promising demonstrations to robust, trustworthy production system