Automation · STEP-BY-STEP GUIDE

Build your first n8n workflow, one step at a time

Learn the basics with a simple workflow that turns sample data into a clear update.

8 min readPublished 17 Sept 2026Updated 17 Sept 2026
An overhead view of a notebook, pen and laptop on an organised desk

Start with a small, reversible workflow. This example uses a manual trigger and sample data to produce a project update. There is no email sending, no customer data and no AI bill to manage.

1. Define the input and the result

The input is a fictional project name, status and next action. The output is one readable summary. Write that transformation down before building anything.

Use n8n Cloud’s trial or an existing installation. A self-hosted edition is available, but running it safely requires maintenance and is outside this first exercise.

2. Add a manual trigger

Create a workflow and choose Manual Trigger. This lets you decide when to run the example while you learn. Keep the workflow inactive.

Do not connect live accounts for this exercise. Starting manually makes it easier to isolate a mistake.

3. Create the sample fields

Add an Edit Fields (Set) node after the trigger. Set three string fields: project, status and nextAction. Use the fictional values below. Inspect the output after running the node.

project: Website content
status: Draft ready
nextAction: Review the introduction

4. Build the summary

Add another Edit Fields node. Create a field named summary, switch its value to Expression and enter the expression below. Run the workflow and inspect the output.

{{ $json.project + ": " + $json.status + ". Next: " + $json.nextAction + "." }}

5. Test a missing value

Clear nextAction in the first node, then run again. The incomplete sentence shows why a successful execution is not the same as useful output. Restore the field.

Before extending this workflow, decide how missing information should be handled: stop for review, set a clear default or take another branch.

6. Choose a sensible next step

Once the sample output is correct, consider reading approved data from a real system. Review access permissions and expected execution volume first.

AI can help with a task such as summarising unstructured text, but this example needs only ordinary automation. Keep an approval step before any external action.

Product interfaces and plan limits can change. These guides provide a starting point; check the official documentation as you go.