Agentic AI
An AI system that can break down goals into steps, use tools, and iterate until a goal is achieved.
Agentic AI systems go beyond single-turn question-answering. They: 1. Receive a goal from a user. 2. Decide which tools to call (web search, database queries, APIs, code execution). 3. Examine the results. 4. Decide what to do next (iterate, refine, call another tool). 5. Repeat until the goal is achieved.
Examples: An agent that plans a multi-day trip (books flights, hotels, restaurants). An agent that debugs code by running tests, reading error messages, and trying fixes. An agent that researches a topic by searching, reading, and synthesizing.
Tradeoffs: Agentic systems are more flexible but also more complex—they require careful prompting, robust error handling, and sometimes human oversight to avoid taking unintended actions. The line between "agentic" and "simple tool-calling" is fuzzy, but agentic implies goal-driven iteration.
Example
An AI agent that writes a report: research the topic (web search), gather sources, outline the structure, write drafts, and refine based on feedback.