AI Concept Atlas
FIELD NOTES / 01
Guided learning · 6 lessons

Making an agent dependable

Separate testing, execution evidence, access boundaries, and human decisions.

Lesson 1 of 6

Agent

An agent is a system that uses a model to choose actions toward a goal, then responds to what happens. How much freedom it has depends on the application.

[Anthropic]

In practice

A support assistant looks up an order, checks its delivery status, then decides whether more information is needed.

[Anthropic]

A little deeper

Here we use agent for a system with model-directed steps. Some authors also include more prescribed workflows under the same name. [Anthropic]

Helpful to know: Language Model

Explore Agent and its connections ↗