What is an Agent? A Conceptual Primer and History of Agents and Agentic AI

We are pleased to announce the release of Chapter 1: What is an Agent? from our forthcoming book Agentic AI in Law and Finance.

The Challenge of Definition

The term “agent” has become ubiquitous in discussions of artificial intelligence, yet its meaning remains frustratingly unclear. Marketing materials tout “agentic” capabilities while researchers debate whether particular systems qualify as agents at all. This confusion isn’t merely academic—it has real consequences for how organizations evaluate, deploy, and govern AI systems.

This chapter addresses that need for definitional clarity by drawing upon nearly a century of scholarship across eight disciplines—including philosophy, law, economics, cognitive science, and computer science.

A Three-Level Hierarchy

Synthesizing work from Anscombe to recent advances in Agentic AI, we propose a three-level hierarchy of agency:

• Level 1 (Agent) is defined by the minimal properties of Goal, Perception, and Action (GPA)
• Level 2 (Agentic System) adds the operational properties of Iteration, Adaptation, and Termination (IAT), implemented through traditional algorithms
• Level 3 (Agentic AI) fulfills these six properties using AI or large language models (LLMs) for planning and orchestration

Practical Evaluation

Beyond theoretical definitions, we provide a practical evaluation rubric to help practitioners distinguish genuine agentic systems from sophisticated tools or single-shot chatbots. The analysis explores critical dimensions such as:

• The autonomy spectrum
• Entity frames
• The boundaries between delegated proxies and self-directed entities

Safeguards for Professional Deployment

The chapter outlines essential safeguards for professional deployment in high-stakes domains like law and finance:

• Attribution and provenance
• Escalation protocols
• Confidentiality controls
• Human oversight requirements

Read the Full Chapter

The complete 60-page chapter is available on SSRN. We welcome feedback from practitioners, researchers, and anyone grappling with these definitional questions.

This chapter forms the foundation for understanding the architectural and governance considerations that follow in Chapters 2 and 3 of the book.