AI Learning Hub › Learn AI › Self-Paced Courses › Agents Without the Hype

Course ยท Beginner to intermediate

Agents Without the Hype

A hands-on guide to how AI pursues goals, uses tools, and stays under human control

Start with What Makes Something an Agent? Back to the tutorials →

First, put aside the word “autonomous.”

Agent describes who chooses the next step rather than how independent a system is, and an agent can be kept on a very short leash.

AI products increasingly use the word agent, but the label can hide several different system designs. Two tools described in the same language may differ completely in who is actually deciding what happens.

This course starts with the mechanics rather than the marketing. You will learn to look at a system and describe how it works: what goal it was given, what it can reach, what it does with an intermediate result, and where a human can still intervene. Across the ten modules you will explore what separates a single response from a workflow, what makes a system more agent-like, how goals, context, tools, and intermediate results shape what happens next, where humans can retain control, and how to supervise increasingly consequential actions.

“Agent” is a marketing word at least as often as a technical one. A product that calls itself an agent may run a fixed script underneath. A product that never uses the word may let a model choose its own steps. The label on the box does not reliably tell you the architecture inside it, which is why this course teaches you to look at behavior instead.

There is no single official boundary around the word. Different products and researchers use “agent” somewhat differently. In this course, we use a practical distinction: a workflow follows a path chosen in advance; an agent lets the model make meaningful choices about how to pursue a goal.

Before asking what an agent can do, ask who decides what it does next.

What this course promises

You do not need to build an agent, or write any code, to understand one. The course builds a single habit: describing an AI system in terms of control rather than capability. By the end you should be able to look at an unfamiliar product and answer:

The emphasis is judgment and supervision rather than building. The modules are read-and-decide: there is nothing to install and no account to create. A capstone at the end asks you to review a proposed agent before it is given the job.

Format

Ten modules and a capstone, each on its own page. Each module introduces only the vocabulary the next step needs.

Time

About 20 minutes a module. Modules 1 to 5 are beginner level; 6 to 10 are intermediate.

Written for

Students, faculty, and staff at Stanford Law School. No programming background is assumed.

What you need

Nothing: no account, no software, no AI subscription, no terminal. Module 10 adds one small task in a Stanford-managed AI environment.

Course path Beginner Delegator Supervisor

The course

  1. Module 01 · Beginner

    What Makes Something an Agent?

    Workflows follow a path chosen in advance; agents let the model choose the next step. Tell them apart by asking who decides, rather than by reading the product label.

    Start module

  2. Module 02 · Beginner

    Give the Agent a Job

    A goal, a success condition, boundaries and a stop condition: what the agent is actually being asked to do, and where its discretion ends.

    Start module

  3. Module 03 · Beginner

    Give It Context

    What information the system can see when it decides, why context windows have limits, and why handing over everything you have is rarely the right move.

    Start module

  4. Module 04 · Beginner

    Give It Tools

    Tools are how a system reaches past its own text: reading a file, running a search, sending a message. Which ones belong in the toolbox depends on the job.

    Start module

  5. Module 05 · Beginner

    Let It Take More Than One Step

    One result changes what the next step should be. How iteration works, how a run differs from a single response, and why the route cannot be written down in advance.

    Start module

  6. Module 06 · Intermediate

    When the Plan Changes

    Plans meet evidence. Adaptation, fallbacks and dead ends, and why flexibility about the route leaves the boundaries where they were.

    Start module

  7. Module 07 · Intermediate

    Read Is Not Act

    Reading returns information; acting changes something outside the conversation. Reversibility, side effects, and the questions worth asking before a tool call takes effect.

    Start module

  8. Module 08 · Intermediate

    Put a Human in the Loop

    Checkpoints, approval and approval scope: where human judgment belongs, what the human needs shown in order to decide, and what a single approval actually authorizes.

    Start module

  9. Module 09 · Intermediate

    When an Agent Goes Off Course

    Failure modes, prompt injection and recovery. How to tell a route change from a departure from the job, and how to locate the point where a run diverged.

    Start module

  10. Module 10 · Intermediate

    Supervise an Agent

    Traces, verification and escalation. What to check before, during and after a run, and why a polished completion message is not evidence that anything changed.

    Start module

Continue Learning from the Source