What is n8n? Workflow automation explained

The short version

• n8n automates flows between apps: email, spreadsheets, calendar, website, social media, AI models. One trigger, several steps, one result.
• The cloud plan costs 20 dollars or euros a month billed annually, with 2,500 runs; there is a free trial first, no credit card needed. On your own server, the Community Edition is free, but you handle hosting and updates yourself.
• You don't need AI for every flow. Passing data along, publishing posts, filing receipts: none of that needs a model. AI joins in when something has to be read, summarized or written.
• Start with one flow that repeats every week. Below are three you can rebuild.

n8n Cloud: free trial with no credit card, then from 20 dollars or euros a month billed annually.

Try n8n Cloud*

Affiliate link: I earn a commission, nothing changes for you.

What n8n is, in three terms

Every workflow in n8n is made of the same three things. Once you know them, every template and tutorial makes sense.

• Trigger: the event that starts the flow. A new email, a submitted form on your website, a new row in a spreadsheet, or simply a schedule, say every Monday at eight.
• Steps: what happens next, one after another. n8n calls them nodes. One step fetches data, one transforms it, one sends it on. AI steps belong here too: they summarize a text or draft a reply.
• Connections: the apps n8n can talk to. Common services have ready-made building blocks, everything else goes through a generic interface. According to the pricing page, all connections are included in every plan.

A few facts before it gets concrete:

Three example flows

These are flows that save freelancers real time. Each has a trigger, two or three steps and a result. Two of them you can rebuild as they are; the third runs like this in a client project.

1. A website enquiry becomes a ready-made card

The contact form on your website sends every enquiry to n8n. Instead of a raw email you get a sorted card with everything you need to reply.

1. TriggerForm submitted

   Your website reports the new enquiry to n8n via a webhook.

2. StepInto the spreadsheet

   Name, request and date land in your client list, for example in Google Sheets.

3. StepAI summarizes

   An AI step reads the enquiry, states the request in one sentence and drafts a reply.

4. ResultA card in your inbox

   An email with summary, draft and a link to the spreadsheet row. Replying is still your job.

What you need: a form that can send a webhook, a spreadsheet, access to an AI model. Where it bites: the AI draft is a draft. Read it before you reply.

2. Recording in, summary and three post ideas out

You record a client call, a talk or a voice note. n8n turns it into minutes and three post suggestions you only have to check.

1. TriggerNew file in a folder

   You drop the recording into a cloud folder, n8n notices the new file.

2. StepTranscribe

   A transcription step turns audio into text, locally on your machine if you prefer.

3. StepAI condenses

   A summary with decisions and open points, plus three post ideas in your style.

4. ResultDocument plus drafts

   Minutes and drafts land in your notes app or arrive by email.

What you need: a cloud folder, a transcription service or a local model, a style briefing for the AI. Where it bites: check names and numbers in the transcript before anything goes out.

3. Handing posts over to Metricool

This flow runs like this in a client project: finished posts come out of a fixed process, n8n hands them to Metricool, and Metricool publishes them on the channels at the scheduled time. Nobody copies text into planning tools any more.

1. TriggerPost is ready

   Text, image and date are in place and the post is approved.

2. Stepn8n bundles

   n8n takes text, image and date and checks that nothing is missing.

3. StepHand-over to Metricool

   n8n creates the post as a scheduled post in the Metricool calendar.

4. ResultPublished on time

   Metricool publishes automatically. The calendar shows what goes out when.

What you need: access to Metricool, either through its API (from the Advanced plan, per Metricool) or through Metricool's MCP server (on every plan, per Metricool), and a place where finished posts live. Where it bites: image formats per channel. Set them once and it runs. More in the Metricool guide.

Notice what the three have in common? None of them replaces your decision. They take the sorting, retyping and copying off your hands and put the result in front of you.

n8n Cloud or self-hosted?

Both are the same software. The difference is who runs the server. The numbers come from the n8n pricing page and the license page, as of October 24, 2026.

A run, per n8n, is one complete execution of your workflow, no matter how many steps it has. Flow 1 above uses one run per enquiry. 2,500 a month goes a long way for a freelancer; the next plan, Pro, with 10,000 runs costs 50 dollars or euros a month.

My take: if you are using n8n for the first time, start in the cloud. Everything that can go wrong with a server is gone, and after an hour you know whether the way of working suits you. If you already run a server, save the fee.

Do you need AI for automation at all?

Often not. If a flow runs the same way every time, say moving data from one tool to the next or publishing posts at fixed times, n8n without an AI step is enough. AI joins in only when content has to be understood, summarized or written, as in flows 1 and 2. Flow 3 does without. Which model you pick for the AI steps is secondary; how to decide is in the guide ChatGPT, Claude or Gemini.

Where to start

1. 1

   One task that repeats

   Pick something you do by hand every week and that annoys you. Not the biggest task, the most frequent.

2. 2

   Name the trigger and the result

   What starts the flow, and what should end up where? If you can say it in one sentence, the task is ready.

3. 3

   Build with two steps, then extend

   First the trigger and one step, test, then the next step. That way you find mistakes while they are small.

That is exactly how I work with clients: first the task, then the tool. What that looks like as guided work is on AI for freelancers.

Sources

Retrieved on October 24, 2026.

Related

• My AI Stack: The AI Tools I Actually Use
  n8n and Metricool working together, with examples from client projects.
• ChatGPT, Claude or Gemini? The Task Decides, Not the Model
  Which route fits your task: automation, script, AI automation or chat assistant.
• Function Calling Explained
  How an AI model calls tools, the foundation for AI steps in workflows.