AI Agents, Simply · Level 1 · Beginner
Playing loads YouTube (privacy-enhanced mode).Watch on YouTube
Agent or workflow: how to choose
Fixed steps or a model that decides? A simple way to choose between a workflow and an agent, with examples.
Build the kind of agent these lessons explain, with a closed list of actions and an approval step. Needs some Python.
Course
Build an AI agent for your business that picks actions from a closed list you define, waits for your approval on anything you haven't cleared to run alone, and refuses five kinds of action in code: moving money, DNS, deleting, revoking keys and changing itself.
Agentic AI Explained with a Real Workflow: Model, Agent, Agentic System
All lessons in AI Agents Explained Simply
The lesson's script, line by line. The published video's own lines are not on record here, so the video may differ in places.
A workflow runs the same steps, in the same order, every time. Anthropic describes workflows as following predefined code paths.
That makes it easy to test, and easy to predict. If it worked today, it'll work the same way tomorrow. Boring, in the best way. And when something breaks, you can see exactly which step failed.
And a workflow can still use AI inside a step, like asking a model to write a summary. The steps around it stay fixed.
An agent is different. The model decides what to do next, based on what it found so far.
It uses tools in a loop: it looks, decides, acts, and checks the result, again and again until the job is done.
That's flexible, and great when nobody can write the steps in advance. But it's harder to predict, and harder to test, because two runs of the same job can take different paths.
Quick quiz. In a workflow, who decides the steps? Comment your answer.
It's B, you, in advance. That's what makes it predictable.
Here are the trade-offs. Speed first. A workflow usually finishes faster, because there's no thinking between steps.
Cost. An agent calls the model again and again in its loop, and every call costs something. A workflow might call it once, or never.
And predictability. A workflow does exactly what you wrote. That's why Anthropic suggests starting with the simplest solution, and adding complexity only when it clearly helps.
If this is helping, subscribe. Next, we draw an agent's tools, memory and planning, piece by piece.
You don't have to choose just one. A common pattern is a workflow around a small agent, so the agent only works inside clear edges.
The workflow does the fixed steps, like catching a form and saving it. The agent only does the fuzzy part, like researching an answer.
And a person approves the result before anything leaves. You get the flexibility where you need it, and control everywhere else. In n8n, that can be a normal workflow with an AI agent node in the middle.
Let's practise. Saving every form entry to a sheet: same steps every time. Workflow.
A weekly report from a spreadsheet: fixed steps on a schedule. Workflow.
A reminder the day before a booking: read the calendar, send one message. Workflow again.
Researching a new supplier across many web pages: you can't write those steps in advance. That's agent work.
And sorting a messy inbox needs judgment, so an agent can help, with you approving any reply before it goes out. Notice the pattern: when the steps are known, a workflow wins. When they're not, an agent is worth it.
Pause the video and pick one job from your own week.
Is it a workflow or an agent, and why? Resume when you've decided.
Last quiz. The steps are the same every time. Agent or workflow? Comment.
B, a workflow. Simple, fast and predictable.
Recap. A workflow follows steps you wrote in advance.
An agent lets the model pick its steps, in a loop.
Start with the simplest thing that works.
And mix them, with a person approving what matters. If you're unsure, build the workflow first. You can always add an agent later, once you know exactly where it helps. Next time: agentic AI, explained with a real workflow.
