Agentic AI

AI systems that can autonomously plan, execute multi-step tasks, use tools, and adapt their approach based on intermediate results, moving beyond single-turn question-answering to goal-directed behavior.

Key Characteristics

Goal-Directed: Given an objective, the system determines the steps needed rather than requiring explicit instruction for each step.

Tool Use: Can invoke external tools, APIs, or services to accomplish tasks that pure language generation cannot.

Iterative Reasoning: Evaluates intermediate results and adjusts approach, rather than committing to a single generation path.

Persistence: Maintains state across multiple interactions to complete extended tasks.

The Shift from Generative to Agentic

Generative AI produces outputs (text, images, code). Agentic AI produces outcomes (completed tasks, achieved goals). This shift changes the risk profile: errors can compound across steps, and the system may take actions with real-world consequences.

Governance Implications

Agentic systems require different oversight than generative systems. The question shifts from “is this output appropriate?” to “should this action be authorized?” Human-in-the-loop becomes human-on-the-loop, monitoring rather than approving each step.

Related: 05-atom—agentic-ai-definition, 05-molecule—generative-agent-agentic-comparison, 01-atom—human-in-the-loop