Skip to content
findyouragent.ai

How to Build Your First AI Automation With n8n (No Code)

Find Your Agent Team · · 7 min read
How to Build Your First AI Automation With n8n (No Code)

The easiest first n8n automation is a three-node workflow: a trigger that catches something new (like an email), an AI step that summarizes or sorts it, and an action that saves the result (like a new row in Google Sheets). You can build it in under an hour on the free cloud trial, with no code, and this guide walks through each step.

Pricing and interface details below are from n8n's pricing page and documentation as of November 2026. We have not hands-on tested every step, so treat node and field names as a guide and check them against what you see on your own canvas.

What you will end up with and what you need

You will end up with a live workflow that watches a Gmail inbox for new messages, asks an AI model to write a one-line summary and label each email as urgent, routine, or ignorable, and appends the result to a Google Sheet. It is a small but genuinely useful triage log, and the same pattern works for form submissions, support tickets, or reviews.

  • Tools: n8n (cloud trial or self-hosted), a Google account, and access to an AI model.
  • Time: 45 to 60 minutes the first time, mostly spent on connecting Google credentials.
  • Cost: $0 to start. See the cost section below.

Cost and setup options

n8n charges by workflow executions, not by number of steps or users. Per n8n's pricing page (as of November 2026), the cloud Starter plan is listed at 20 euros per month billed annually for 2,500 executions, and Pro at 50 euros per month billed annually for 10,000 executions. The page states that all plans include unlimited users, workflows, and integrations, and that the AI Agent node and AI steps are included on every plan.

For learning, you do not need to pay. n8n lists a cloud free trial with no credit card required, including 1,000 executions and a 180-second timeout. n8n also offers a Community Edition you can self-host from GitHub, which is free to run but means you handle hosting, updates, and security yourself. Beginners should start with the cloud trial and decide later.

One honest caution: a self-hosted setup is where many older tutorials lose people, because it involves Docker, a server, and a public URL for webhooks. If your goal is to finish your first workflow today, skip it.

Step by step: build the workflow

Step 1: Create the workflow and add a trigger

  1. Sign up for the n8n cloud trial. In the left menu, open Personal and select Create workflow. Name it "Inbox triage".
  2. On the canvas, add a Gmail Trigger node with the event Message Received.
  3. Connect a Google credential when prompted. n8n has a dedicated Google credentials guide in its docs; follow it, since this is the step that takes the longest.
  4. Set Poll Times to a sensible interval, and set Read Status to unread only (the default). To keep testing cheap, use the Search filter with Gmail syntax, for example "label:test" or a specific sender, so only a few messages qualify.

Per the docs, the trigger returns up to 10 emails per poll by default (maximum 50). Each email processed counts toward your execution allowance, which is why filtering matters.

Step 2: Test the trigger

Send yourself a test email, then select Execute step on the trigger. You should see the message appear in the output panel. Look at the fields: you will drag them into the next node, so note where the subject, sender, and body text live. With Simplify on (the default), the output is condensed; turn it off if the body text you need is missing.

Step 3: Add the AI step

  1. Add a Basic LLM Chain node after the trigger.
  2. Under Prompt, choose Define below so you can write the prompt and drag in fields from the input panel.
  3. Attach a chat model to the node. For example, add the OpenAI Chat Model sub-node. n8n's docs say it needs an OpenAI API key, and that on n8n Cloud you can instead choose gateway credits, so no OpenAI account is required. You can swap in another provider's chat model sub-node if you prefer.
  4. Write the prompt, using expressions (the double curly braces n8n replaces with live data) for the email fields.

An example prompt:

You triage emails for me. Read the email below. Reply with exactly two lines. Line 1: a summary in 15 words or fewer. Line 2: one label from this list: URGENT, ROUTINE, IGNORE. Subject: {{ $json.Subject }} From: {{ $json.From }} Body: [drag the email text field here]

The exact field names depend on your trigger output, so drag them from the input panel instead of typing them. Run Execute step and check that the answer is two lines. If you want guaranteed structure, turn on Require Specific Output Format and attach a Structured Output Parser, which the docs list as one of the supported parsers.

Step 4: Save the result to Google Sheets

  1. Create a sheet with the columns Date, From, Subject, Summary, and Label.
  2. Add a Google Sheets node and choose the Append Row operation. Connect your Google credential and pick the spreadsheet.
  3. Map each column to an expression: sender and subject from the trigger, and the summary from the AI node's output. If you used two lines, an Edit Fields (Set) node between them can split the output into separate fields first.

Step 5: Run it end to end and publish

Select Execute on the whole workflow with a test email in place and confirm a new row appears. Then select Publish, which n8n's docs describe as activating the workflow so it runs on its own. Send two or three more emails to confirm it keeps working without you clicking anything. You can review past runs in the executions list if something fails.

Prefer to start from a prompt?

n8n's own beginner tutorial shows an Assistant that builds a workflow from a text description. Per n8n's pricing page, the Assistant is in preview and included on Starter, Pro, and Business with monthly credits that do not roll over. It is a fast way to get a draft canvas, but you still need to understand each node to fix it, so building the first one by hand is worth it.

Common mistakes

  • Letting the trigger run on your whole inbox. Filter by label or sender while testing, or you will burn executions and send private email content to an AI provider you did not intend to.
  • Typing field names from a tutorial. Names vary with settings like Simplify. Drag fields from the input panel.
  • Forgetting to publish. A workflow you only tested manually will not run on new email.
  • Trusting the AI blindly. A model can mislabel an urgent email as routine. Use this as a log to scan, not as a filter that hides messages.
  • Skipping the privacy check. Email bodies go to the model provider you connect. Do not point this at mailboxes containing sensitive customer or health data without checking your provider's terms.
  • Building an "agent" when you need a workflow. The AI Agent node lets a model decide which tools to call, which is powerful but less predictable. A fixed chain like this one is cheaper to debug. Our explainer on AI agents vs chatbots covers the difference.

The cheapest route

For $0: use the cloud trial while you learn, or self-host the Community Edition if you are comfortable with servers. Self-hosted n8n can use your own AI provider key. Be aware that API usage is billed by the model provider separately, though a short summary per email costs very little at small volumes. Check your provider's current price list before running it on a busy inbox.

If you outgrow it, the paid cloud tiers are priced by executions, so a workflow that triggers many times a day can reach 2,500 executions per month quickly. Estimate your volume before choosing a plan.

Where to go next

Once this works, swap the trigger for a form submission, a new row, or a webhook, and swap the sheet for Slack, Notion, or your CRM. n8n is one of several ways to automate with AI; if you want to compare other options, you can browse workflow automation agents on Find Your Agent.

Frequently asked questions

Is n8n really free?

n8n offers a cloud free trial with no credit card required and a self-hostable Community Edition on GitHub. Paid cloud plans start at 20 euros per month billed annually, per n8n's pricing page as of November 2026. Self-hosting is free to license but you cover hosting and maintenance.

Do I need to know how to code to use n8n?

No. You build workflows by connecting nodes on a canvas and filling in settings. A little comfort with expressions (placeholders for live data) helps, and code nodes exist if you want them, but this tutorial needs none.

What is the difference between a workflow and an AI agent in n8n?

A workflow follows steps you define in order. An AI agent node lets a model decide which tools to use and in what sequence. Workflows are more predictable and easier to debug, so they are a good first project.

How many executions will my first automation use?

Each time the workflow runs counts as one execution, per n8n's pricing model. An inbox trigger that processes ten emails in one poll runs once per poll, but a busy mailbox still adds up. Filter your trigger while testing.

Do I need my own OpenAI API key?

Not necessarily. n8n's docs say that on n8n Cloud you can use gateway credits in the OpenAI Chat Model credential field instead of your own key. Self-hosted setups use your own provider key.

Find the right AI agent for your work

Browse and compare hundreds of AI agents by category, pricing, and features.

Browse AI Agents