What is AI Agent?
Also called LLM agent
Definition
In language-model applications, an AI agent is a system in which a language model directs actions and tool use to pursue a goal. Unlike a fixed workflow, an agent can choose its next step based on intermediate results.
The operation of AI Agent is explained.
Anthropic distinguishes agents from workflows with predefined code paths. An agent can retrieve information, inspect the result, and decide whether another tool call is needed. Its autonomy depends on the implementation and permissions.
The relevance of AI Agent is explained.
A procurement assistant could inspect product documentation before comparing integration requirements.
Public documentation should explain capabilities and limitations accurately; machine access alone does not establish that a vendor is suitable for the buyer.
The evaluation of AI Agent is explained.
Agent evaluation covers tool accuracy, stopping conditions, permissions, traceable evidence, and failures involving missing or conflicting documentation. Human approval provides oversight for consequential actions.
Frequently asked questions about AI Agent
Is every chatbot an agent?
A chatbot that only generates replies need not control tools or a multistep process.
Are agents always autonomous?
Agent autonomy varies with the system’s design and permission boundaries.
Does an agent need web search?
An agent can use other tools and does not necessarily need web search.