What workflow automation is
Workflow automation is getting tasks to run by themselves: when something happens, or on a schedule, a series of steps runs across your apps without anyone clicking anything.
For example: when a customer fills in a form, add them to a spreadsheet, send them a welcome email, and post a message in the team’s chat. Each step uses a different app, and an automation tool connects them. n8n, the tool this course uses, lets you build these as a diagram of connected boxes, and it can include AI steps, such as having a model sort incoming emails or draft replies.
You could write a small script for each task instead. The trouble starts when you have fifty of them: where are the passwords for twenty services kept, what happened on last Tuesday’s run, what retries when a service is down, who gets told when something fails, and can anyone else understand or change it? An automation platform handles all of that around your steps.
What you’ll be able to do
By the end you will be able to:
- decide when a task deserves an automation platform and when a script is enough;
- build workflows in n8n that connect different apps;
- connect to any service’s API, and receive messages from services when something happens (webhooks);
- add AI steps safely, without letting a stranger’s text take control of your workflow;
- make workflows survive failures, and get alerted when they stop;
- run your own n8n server.
Where to start
Before you start: knowing what an API is helps. For note 06 you’ll need some Docker.
Read in order.
- What Workflow Automation Is: triggers and actions, the point where a platform beats scripts, the main tools, and what AI changed.
- n8n Core Concepts: boxes (nodes), runs, saved logins, and how data flows between boxes, which confuses almost everyone at first.
- Connecting APIs and Webhooks: logging in to services, fetching data in pages, staying within rate limits, and making sure a repeated message doesn’t do something twice.
- AI and Agent Workflows: the four ways to use AI in a workflow, from safest to riskiest, getting answers in a fixed format, and protecting against instructions hidden in the text an AI reads.
- Error Handling and Retries: sorting failures into kinds, retrying sensibly, and getting alerted when a workflow silently stops running.
- Self-Hosting n8n: running n8n on your own server, the one key you must never lose, and when hosting it yourself is the wrong choice.
Practice
- Build one real workflow with the setup from note 06: something you actually want automated.
- Break it on purpose: a wrong password, a broken incoming message, an expired login. Watch what happens.
- Send the same webhook twice and see whether your workflow does the job twice. Most first workflows do.
Key ideas to remember
- You’re paying for everything around each step, not the step: logins, history, retries, alerts. → 01
- n8n passes a list of items between boxes, and most boxes run once per item. Almost every beginner confusion comes back to this. → 02
- Expect every incoming message to arrive twice, and make sure that’s harmless. → 03
- A “success” reply doesn’t mean it worked. Check what the service actually sent back. → 03
- If you can draw the steps as a flowchart, do that instead of using an AI agent. Agents are for when you can’t. → 04
- An AI step will follow instructions from anyone who can get text in front of it. Give it access to as little as possible. → 04
- The most valuable alert is for the workflow that didn’t run. A stopped trigger produces no errors at all. → 05
- Back up n8n’s encryption key separately. Lose it and every saved login is gone for good. → 06
Related
- AI Engineer: AI agents, structured answers and safety in depth.
- Using AI: using AI chatbots well, without code.
- Secret Management: keeping passwords and keys safe.
- Site Reliability Engineering: how to think about alerts, on a bigger scale.
About this course
- Size: 6 notes, about 5,500 words, written in August 2026.
- Where it came from: written from documentation, not from running n8n. Every claim about running it is read rather than experienced, which matters here because the subject is mostly about running things.
- Not covered yet: Zapier, Make and Temporal in depth; n8n’s code boxes and building your own boxes; testing workflows; cost at large volumes; and the legal side of sending personal data through other companies’ AI models.
Software engineering
Infrastructure & security
Computer science
Hardware & engineering