[Paper Review] Generative to Agentic AI: Survey, Conceptualization, and Challenges

Johannes Schneider|ArXiv.org|Apr 26, 2025

Artificial Intelligence in Healthcare and Education4 citations

TL;DR

This paper surveys Generative AI (GenAI) and Agentic AI, defining concepts, comparing capabilities, and outlining evolution, reasoning, interaction, and challenges toward AGI.

ABSTRACT

Agentic Artificial Intelligence (AI) builds upon Generative AI (GenAI). It constitutes the next major step in the evolution of AI with much stronger reasoning and interaction capabilities that enable more autonomous behavior to tackle complex tasks. Since the initial release of ChatGPT (3.5), Generative AI has seen widespread adoption, giving users firsthand experience. However, the distinction between Agentic AI and GenAI remains less well understood. To address this gap, our survey is structured in two parts. In the first part, we compare GenAI and Agentic AI using existing literature, discussing their key characteristics, how Agentic AI remedies limitations of GenAI, and the major steps in GenAI's evolution toward Agentic AI. This section is intended for a broad audience, including academics in both social sciences and engineering, as well as industry professionals. It provides the necessary insights to comprehend novel applications that are possible with Agentic AI but not with GenAI. In the second part, we deep dive into novel aspects of Agentic AI, including recent developments and practical concerns such as defining agents. Finally, we discuss several challenges that could serve as a future research agenda, while cautioning against risks that can emerge when exceeding human intelligence.

Motivation & Objective

• Define GenAI and Agentic AI and clarify their differences and similarities.
• Compare capabilities and limitations of GenAI and Agentic AI based on literature.
• Describe the evolution from GenAI to Agentic AI and identify key milestones.
• Analyze reasoning, memory, tools, and interaction components of Agentic AI.
• Discuss evaluation strategies and research challenges toward AGI and responsible deployment.

Proposed method

• Conduct a structured literature survey contrasting GenAI and Agentic AI.
• Provide formal definitions and a high-level taxonomy (Table 1) of characteristics.
• Trace the evolution from GPT-2/3/ChatGPT 3.5 to agentic architectures and RL-based interactions.
• Synthesize reasoning patterns (CoT, ToT, FoT, GoT) and memory/tool interaction mechanisms.
• Propose evaluation criteria for Agentic AI and outline a future research agenda.

Experimental results

Research questions

• RQ1 How do GenAI and Agentic AI differ in reasoning, interaction, and autonomy?
• RQ2 What capabilities enable Agentic AI to solve multi-step tasks and interact with environments and tools?
• RQ3 How should Agentic AI be defined, specified, and evaluated (single and multi-agent systems)?
• RQ4 What challenges and risks emerge when progressing toward AGI with Agentic AI?
• RQ5 What are the practical and ethical considerations for deploying Agentic AI?

Key findings

• Agentic AI adds iterative planning, reflection, and environment/tool interaction beyond GenAI.
• Reasoning improvements in Agentic AI can reduce errors and hallucinations via verification and recursion.
• Memory, retrieval augmentation, and dynamic resource allocation enable more scalable and controllable AI workflows.
• In practice, GenAI may outperform Agentic AI in some domains when domain-specific data is abundant or costs are considered.
• Agentic AI represents a major step toward AGI but faces data, training, safety, and interpretability challenges.
• Different reasoning decompositions (CoT, FoT, GoT) and reflection strategies enable deeper problem solving.