AI Agents vs. Agentic AI: A Conceptual Taxonomy and Applications
Comprehensive taxonomy reviewing application domains spanning customer support, healthcare, research automation, and robotics coordination.
This multi-institutional paper establishes a critical distinction that the AI industry has struggled to articulate clearly: the difference between AI agents (individual autonomous entities) and agentic AI (a design philosophy emphasizing autonomy, adaptability, and goal-directed behavior). The taxonomy provides a rigorous framework for classifying systems along dimensions of autonomy, reactivity, and social ability.
The review spans application domains including customer support (where agents handle multi-turn conversations with context retention), healthcare (diagnostic assistants that coordinate with specialist systems), research automation (literature review and hypothesis generation), and robotics coordination (multi-robot task allocation). Each domain reveals different requirements for agent architecture and different failure modes.
The authors argue that the term "agentic AI" should be reserved for systems exhibiting genuine goal decomposition, planning under uncertainty, and the ability to learn from environmental feedback — criteria that exclude many systems currently marketed as "AI agents." This conceptual clarity is essential for setting realistic expectations and designing appropriate evaluation benchmarks.
