# AI Agents vs. Agentic AI: A Conceptual Taxonomy, Applications and Challenges

**Authors:** Ranjan Sapkota, Konstantinos I. Roumeliotis, Manoj Karkee  
**arXiv:** [2505.10468](https://arxiv.org/abs/2505.10468)  
**Date:** 2025-05-15  
**Summary page:** https://www.summarizepaper.com/en/paper/2505.10468v5/

## Key points

- AI Agents: driven by LLMs and LIMs for task-specific automation
- Agentic AI: characterized by multi-agent collaboration, dynamic task decomposition, persistent memory, and coordinated autonomy
- Generative AI as a precursor to AI agents
- Agentic AI represents a paradigm shift with unique features
- AI Agents: customer support and data summarization
- Agentic AI: research automation and medical decision support
- Challenges include hallucination, brittleness, emergent behavior, and coordination failure
- Proposed solutions include ReAct loops, retrieval-augmented generation (RAG), automation coordination layers, and causal modeling
- Provide a roadmap for developing robust, scalable, explainable AI-driven systems
- Support from NSF, USDA, NIFA, and other funding sources mentioned

## Summary

In this comprehensive review, Sapkota et al. critically analyze the distinction between AI Agents and Agentic AI, providing a structured conceptual taxonomy, application mapping, and in-depth analysis of their divergent design philosophies and capabilities. The authors begin by outlining their search strategy and foundational definitions. They define AI Agents as modular systems driven by LLMs and LIMs for task-specific automation. They also position Generative AI as a precursor to AI agents, which advance through tool integration, prompt engineering, and reasoning enhancements. In contrast to AI Agents, Agentic AI systems represent a paradigm shift characterized by multi-agent collaboration, dynamic task decomposition, persistent memory, and coordinated autonomy. Through a chronological evaluation of architectural evolution, operational mechanisms, interaction styles, and autonomy levels across both paradigms  the authors present a comparative analysis highlighting the unique features of each approach. The review further delves into application domains enabled by AI Agents such as customer support and data summarization compared to Agentic AI deployments in research automation and medical decision support. Unique challenges in each paradigm are explored including hallucination  brittleness  emergent behavior  and coordination failure. The authors propose targeted solutions such as ReAct loops  retrieval-augmented generation (RAG)  automation coordination layers  and causal modeling to address these challenges. Sapkota et al. 's work aims to provide a roadmap for developing robust  scalable <kd>,</kbd> explainable AI-driven systems. The authors acknowledge support from the National Science Foundation (NSF), United States Department of Agriculture (USDA), National Institute of Food and Agriculture (NIFA), as well as other funding sources for their research on artificial intelligence applications in agriculture. No data was used for the research described in the article. This review is published in Information Fusion 126 (2026) 103599.