Exploring Agentic AI and Recursive Reasoning
Ayşe Rüya Arslan · International Journal of Applied Science and Research · 2025
This paper explores the architecture, capabilities, and challenges of AI Agents and Agentic AI systems.We break down foundational AI agents into core modules-perception, reasoning, action, and learning-and explain how Agentic AI extends these concepts through multi-agent collaboration, advanced planning, persistent memory, and orchestration.Despite promising advances, both AI Agents and Agentic AI face critical limitations such as lack of causal understanding, coordination bottlenecks, and reliability concerns.Addressing these challenges is essential for the safe, scalable, and trustworthy deployment of intelligent systems capable of complex, autonomous behavior.