Build Your First AI Email Assistant in n8n

The Inbox That Ate My Saturday

I once knew a founder who spent every Sunday night in “email jail.” You know the routine: 200 unread messages, a mix of urgent client fires, boring newsletters, and random sales pitches. He’d sit there, a digital janitor with a mop, trying to figure out what mattered. It was soul-crushing. He wasn’t building his business; he was managing a notification firehose.

This is manual labor in a world of automation. You’re being paid to be a human filter, a low-value intern for your own attention. This stops today. We’re going to build a digital employee that lives inside your inbox, reads your mail for you, and only bothers you with the good stuff.

Why This Matters: Reclaim Your Focus

Every email is an interruption. A context switch. Your brain is like a high-performance engine; it doesn’t like stopping and starting every five minutes. This automation is about building a bouncer for your attention.

What it replaces:

• The part-time intern who sorts mail.
• Your own ‘check-email-while-coffee-brews’ procrastination.
• The anxiety of ‘what if I miss something important?’

We’re not just saving time; we’re building a system that scales with you. Whether you get 10 emails a day or 1,000, this system works the same. It’s the first step toward a fully autonomous executive assistant.

What This Tool / Workflow Actually Is

We’re using a tool called n8n. Think of n8n as a visual factory floor. You’re not writing code from scratch; you’re connecting boxes on a screen with wires. Each box is a task: ‘read an email,’ ‘analyze the text,’ ‘draft a reply,’ ‘send it.’ It’s a logic pipeline.

What this does: It triggers on a new email, uses an AI to classify it (e.g., Urgent, Newsletter, Inquiry), and drafts an appropriate response. It will NOT hit ‘send’ without your approval, unless you tell it to. We’re building a helpful assistant, not a rogue robot.

Prerequisites

Zero coding skills required. If you can drag a file into a folder, you can build this. Here’s what you need:

• An n8n account (Cloud or self-hosted).
• An email account that supports IMAP (Gmail, Outlook, etc.). For Gmail, you’ll need an App Password.
• An OpenAI API key (or another LLM provider).

Don’t worry, we’ll go slow. This is like assembling a Lego set with very clear instructions.

Step-by-Step Tutorial: Building the Email Sorter

Let’s head to the factory floor. Open your n8n workspace.

Step 1: The Trigger (The Mailroom Bell)

First, we need to know when we have mail. We don’t want to constantly check; we want the mailroom to ring a bell.

1. Click the big + Add first workflow button.
2. Search for IMAP Email and select it as your Trigger node.
3. Double-click the node to configure it.
4. Enter your email settings (Host, Port, User, Password). For Gmail, use an App Password.
5. Set Trigger: On New Mail (polling interval) to something like every 1 minute.
6. Click Test step to make sure it can connect. It should fetch a recent email.

Step 2: The AI Classifier (Your Smart Intern)

Now that we have the email, we need someone to read it and tell us what it is. That’s our AI.

1. Click the + button to add a new node.
2. Search for Basic LLM Chain. This is our connector to an AI model.
3. Connect it to the IMAP node. Drag a wire from the IMAP’s right-side dot to the LLM Chain’s left-side dot.
4. Double-click the LLM Chain node. Under ‘Model,’ select OpenAI Chat Model and connect your API key.
5. Now we need to tell it what to do. Click the Prompt (Text) field. We’re going to write a simple instruction. This is our prompt engineering.

Step 3: The Prompt (The Job Description)

In the Prompt field, type something like this. You can copy-paste this directly:

You are an expert executive assistant. Your job is to classify incoming emails and draft a brief, helpful reply. 

Analyze the following email:
Subject: {{ $('IMAP Email').item.json.subject }}
From: {{ $('IMAP Email').item.json.fromText }}
Body: {{ $('IMAP Email').item.json.text }}

Return a JSON object with two keys:
1. "category": Choose ONE of: "Urgent", "News", "Promotion", "Inquiry".
2. "draft_reply": A 1-2 sentence professional response appropriate for the category. If the category is "Promotion", the reply should decline politely. If "Urgent", acknowledge receipt and say you'll respond shortly.

// n8n gives us the AI's text response in the first item
const aiResponseText = items[0].json.text;

let parsedResult;

try {
  // Try to find the JSON object in the text response
  const jsonMatch = aiResponseText.match(/{[\s\S]*}/);
  if (jsonMatch) {
    parsedResult = JSON.parse(jsonMatch[0]);
  } else {
    // If it's just plain text, handle it
    parsedResult = { category: 'Unclear', draft_reply: aiResponseText };
  }
} catch (error) {
  parsedResult = { category: 'Error', draft_reply: 'Could not parse AI response.' };
}

// Return the new item for the next step in the workflow
return [{
  json: {
    original_email: items[0].json,
    category: parsedResult.category,
    draft_reply: parsedResult.draft_reply
  }
}];