AX-300 · Lesson 1 of 12 · Module 1: What Makes an Agent · 3:06

What makes an agent

A chat answers one message and waits; a workflow follows fixed steps a person designed; an AI agent is given a goal and works toward it in a loop, planning, acting with a tool, checking the result and deciding the next step, until the job is done or a person must decide. Agents cost more and can take a different path on each run, so they are harder to test: use one only when the next step depends on what it finds, and keep fixed jobs as workflows. Worked example: the marketing agent reads the best-sellers, checks stock, finds only 18 bags of the top tea and features another, then stops for approval before anything is sent.

Frame from lesson 1 of Sophyx Certified AI Agent Builder – Professional, "What makes an agent"

What you will learn

  • Tell an agent from a chat and a workflow
  • Name the four steps of the agent loop
  • Decide when a workflow is the better choice

The one thing to remember

Agents choose their steps; workflows follow them.

Common mistake: Using an agent where a workflow would do

Agents cost more and can take a different path each time, so a job with fixed steps gets slower, dearer and harder to test for no gain.

The fix: Use an agent only when the next step depends on what it finds.

Key terms

AI agent
AI that works toward a goal and chooses its next step.
Agent loop
Plan, act, check the result, then decide the next step.
Tool
An action the agent can ask the system to do for it.

Transcript

The brands in the worked examples are fictional, and the numbers are illustrative.

It's Monday, and Ines, who runs marketing at Larchmere Tea Company, has eleven jobs on her list.

Her chat assistant answers one question at a time, and she carries each answer on herself.

Welcome to AI Agent Builder. Lesson one is about what makes an agent.

You'll learn what sets an agent apart, the four steps of its loop, and when a workflow is better.

First, a quick recall from the last course. Where should a person approve what an AI workflow does?

Wherever a mistake would leave the building or can't be undone.

A chat answers one message and waits, while a workflow follows fixed steps that a person designed.

An AI agent gets a goal instead, and chooses each next step, using tools and checking what happens.

So Ines gives the agent builder a goal, to plan this week's campaigns for her approval.

Every agent runs the same loop. It plans, acts with a tool, checks the result and decides the next step.

It goes round again until the job is done, or until a person must decide.

It writes a plan, then calls a tool that reads the best-sellers of the last thirty days.

Earl Grey Cream sells best, so before featuring it, the agent checks the stock.

Only eighteen bags are left until November, too few for eighteen hundred and sixty readers.

So it features Assam Breakfast instead, with four hundred and twenty bags in stock.

After nine steps, the drafts are ready, and it stops to ask Ines before anything is sent.

Here's a quick check. Why did the agent change the featured tea?

It checked the stock and found only eighteen bags, so its next step changed.

Agents cost more, and they can take a different path each time, so they are harder to test.

The welcome email for new Box subscribers never changes, so Ines keeps it as a workflow.

If the steps never change, build a workflow instead.

A common mistake is using an agent where a workflow would do.

Instead, use an agent only when the next step depends on what it finds.

This week, list three jobs you'd like help with, and mark the steps that depend on what it finds.

Pick one for an agent, and say why a workflow won't do.

Before we wrap up, can you name the four steps of the agent loop?

Plan, act with a tool, check the result, and decide the next step.

So remember, agents choose their steps, and workflows follow them.

Now try the practice quiz, then join Ines in lesson two.

Part of AI Agent Builder (AX-300)

For people who run AI workflows: build an agent with the right tools, tests and guardrails, and approvals where it counts.

12 lessons, practice quizzes, a final exam and the Sophyx Certified AI Agent Builder – Professional certificate. Regular price USD 49.99. Your first certification is free. Free with any paid Sophyx plan.

Start the course free
Book a Demo