Author
Ranjan Sapkota, Konstantinos I. Roumeliotis, Manoj Karkee
Year
2025
<Insert Cover Image Here & On The Page Summary>
AI Agents vs. Agentic AI: A Conceptual Taxonomy, Applications and Challenges
Ranjan Sapkota, Konstantinos I. Roumeliotis, Manoj Karkee. 2025. (View Paper → )
This review critically distinguishes between AI Agents and Agentic AI, offering a structured, conceptual taxonomy, application mapping, and analysis of opportunities and challenges to clarify their divergent design philosophies and capabilities.
Here’s my summary of the difference between the two…
Aspect | Agentic AI | AI Agent |
Scope | Handles complex, multi-step workflows via dynamic goal decomposition and role assignment | Tackles narrow, well-defined tasks |
Architecture | A coordinated system of multiple specialized agents with an orchestrator | A single, tool-augmented executor |
Autonomy | Exhibits broader, end-to-end autonomy across tasks and sub-tasks using meta-agents/orchestration | Has high autonomy within their task |
Interaction Model | Adds inter-agent communication (centralised or decentralised protocols) to align decisions | Mainly interacts user↔agent↔tool |
Planning & Reasoning | Coordinates multi-agent planning with reflective loops and cross-checks | Relies on sequential tool use (e.g., ReAct/CoT) for a single thread |
Memory | Uses persistent/shared memory(episodic & semantic) so collaborating agents keep and reuse context | Often uses little or short-lived state |
Failure Modes & Governance | Adds risks like error cascades, coordination breakdowns, emergent behaviour, and heavier needs for orchestration, audit, and safety | Struggles with hallucination, brittleness, long-horizon planning |
Typical Applications | Research automation, multi-robot coordination, medical decision support | Customer support, scheduling, enterprise search, summarisation |