Actuarial Notes

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

By Ranjan Sapkota, Konstantinos I. Roumeliotis and Manoj Karkee · arXiv · 2025 · CFE201-117-27

Study guide: Exam CFE-201

A review that distinguishes AI Agents, modular systems driven by large language models for task-specific automation, from Agentic AI, systems marked by multi-agent collaboration, dynamic task decomposition, persistent memory and coordinated autonomy. It offers a conceptual taxonomy, a mapping of applications, and an analysis of the challenges of each paradigm, and proposes solutions and a roadmap for robust, scalable and explainable AI-driven systems. It is the arXiv preprint (2505.10468) of an article later published in Information Fusion.

On the syllabus

  • Exam CFE-201 — Sections IV and V (pp. 16-30); outcomes 4a, 4b.

I Introduction

  • A Methodology Overview
    • 1 Search Strategy

II Foundational Understanding of AI Agents

  • 1 Core Characteristics of AI Agents
  • 2 Foundational Models: The Role of LLMs and LIMs
  • 3 Generative AI as a Precursor
  • A Language Models as the Engine for AI Agent Progression
    • 1 LLMs as Core Reasoning Components
    • 2 Tool-Augmented AI Agents: Enhancing Functionality
    • 3 Illustrative Examples and Emerging Capabilities

III The Emergence of Agentic AI from AI Agent Foundations

  • 1 Conceptual Leap: From Isolated Agents to Coordinated Systems
  • 2 Key Differences between AI Agents and Agentic AI
  • A Architectural Evolution: From AI Agents to Agentic AI Systems
    • 1 Core Architectural Components of AI Agents
    • 2 Architectural Enhancements in Agentic AI

IV Application of AI Agents and Agentic AI

  • 1 Application of AI Agents
    • 1 Customer Support Automation and Internal Enterprise Search
    • 2 Email Filtering and Prioritization
    • 3 Personalized Content Recommendation and Basic Data Reporting
    • 4 Autonomous Scheduling Assistants
  • 2 Applications of Agentic AI
    • 1 Multi-Agent Research Assistants
    • 2 Intelligent Robotics Coordination
    • 3 Collaborative Medical Decision Support
    • 4 Multi-Agent Game AI and Adaptive Workflow Automation

V Challenges and Limitations in AI Agents and Agentic AI

  • 1 Challenges and Limitations of AI Agents
    • 1 Lack of Causal Understanding
    • 2 Inherited Limitations from LLMs
    • 3 Incomplete Agentic Properties
    • 4 Limited Long-Horizon Planning and Recovery
    • 5 Reliability and Safety Concerns
  • 2 Challenges and Limitations of Agentic AI
    • 1 Amplified Causality Challenges
    • 2 Communication and Coordination Bottlenecks
    • 3 Emergent Behavior and Predictability
    • 4 Scalability and Debugging Complexity
    • 5 Trust, Explainability, and Verification
    • 6 Security and Adversarial Risks

VI Potential Solutions and Future Roadmap

  • A Potential Solutions
  • B Future Roadmap

VII Conclusion

References

Sources