Agentic AI, Autonomous Agents, and Multi-Agent Systems: Concepts, Challenges, and Research Pathways

The recent surge in interest around agentic AI—systems that plan and act autonomously—and the long-standing study of multi-agent systems (MAS) converge into a rapidly evolving research space where algorithmic autonomy, coordination, safety, and societal impact intersect. This paper synthesizes foundational ideas from agent theory and MAS, surveys recent trends in agentic systems (including autonomous LLM-driven agents and advances in multi-agent reinforcement learning), and proposes a structured research agenda. We introduce an experimental methodology for evaluating agentic behaviors along axes of capability, safety, cooperation, and resource efficiency, and discuss open technical and ethical challenges. The paper concludes with prioritized research directions intended to accelerate robust, explainable, and societally aligned agentic systems. (Wiley)