Build an AI Sales Assistant: Automate Lead Outreach

The Email Grind Will Kill Your Soul (Let’s Fix It)

Let me tell you about a founder I know. Let’s call him Dave. Dave is building something cool. He’s got a product. He’s got a list of potential customers. And he has a very, very special relationship with his keyboard.

Every morning, Dave opens his spreadsheet. He looks at a company name. He opens a new tab, Googles the company. He reads their “About Us” page. He finds the CEO’s name. He tries to guess their email. He writes a “personalized” email that starts with, “Hi [Name], I saw your company is doing cool things in [Industry]…” He does this. Then he does it again. And again. For two hours straight.

Dave isn’t selling. He’s doing digital manual labor. He’s an unpaid intern for his own sales process. His soul is slowly leaking out through his fingertips.

What if Dave had a tireless, super-smart intern who could do all that research and draft those emails in 30 seconds? An intern who never gets tired, never complains, and can do it for 100 leads at once? That’s what we’re building today. Welcome to the factory.

Why This Matters: From Chore to Choreography

This isn’t just about saving time. It’s about changing your role from a doer to an architect.

When you automate your sales outreach, you unlock a few things:

Scale: You can’t personally email 500 qualified leads a week. An AI assistant can. That’s not 10x improvement; it’s 100x.
Sanity: You stop dreading the “outreach” block in your calendar. It becomes a background process. You focus on high-value conversations with people who reply.
Intelligence: A good AI assistant doesn’t just spam. It can tailor the message based on real data, making your outreach feel more personal, not less.

This replaces the manual, repetitive work of a junior sales development rep (SDR). It frees you up to be the account executive, the closer. It turns a chore into a well-oiled machine.

What We’re Actually Building (The Blueprint)

We’re going to build an automation that takes a list of new leads and spits out a perfectly drafted, personalized email for each one. This is a three-step pipeline:

The Trigger: New lead data appears. This could be from a form on your website, a CRM, or even a simple Google Sheet.
The Brain (Research & Write): An AI looks up the company, understands the context, and crafts a unique email message. This is where the magic happens.
The Output (Draft in Inbox): The final email is saved as a draft in your Gmail or Outlook, ready for your final review and a single click to send.

What this does NOT do: This is not a spam bot. We are not blasting emails without oversight. The final step is a draft, so you always have control and can add that human touch before hitting send. This is about augmentation, not mindless automation.

Prerequisites: What You Need to Bring to the Lab

Don’t panic. You don’t need a computer science degree for this. You just need:

An n8n account: The free tier is more than enough to get started. It’s our workshop floor.
An OpenAI API key: This is for the AI brain. You’ll need to set up a payment method, but the cost for this workflow is pennies. We’re talking fractions of a cent per lead.
A Gmail or Outlook account: For sending the drafts.
A list of leads: For our example, we’ll use a simple Google Sheet with names and company domains.

If you can sign up for a web service, you can do this. Let’s fire up the machinery.

Step-by-Step Tutorial: Building Your Sales Intern

Open your n8n dashboard. Let’s create a new workflow. This is our assembly line.

Step 1: The Trigger – Getting Leads Into the System

First, we need to tell n8n when to start working. We’ll use a schedule to simulate checking for new leads, but in a real system, this would be a webhook or a database trigger.

Click the + (Add Node) button in the top right.
Search for Schedule Trigger and select it.
Set it to run every minute for testing, or once a day for a real workflow.

Now, we need data. For this example, let’s pretend we have a Google Sheet with two columns: Full Name and Company Domain. We’ll add a hard-coded set of data for simplicity, but you can connect to a Google Sheet node later.

Step 2: The Brain – AI Research & Email Writing

This is the heart of the operation. We’ll use a single AI agent to do the research and writing. We’ll use the powerful gpt-4o-mini model.

Add a new node after the trigger. Search for AI Agent.
For the Model, choose OpenAI Chat Model. Connect your OpenAI credentials.
We need to give the agent a clear system prompt. This is its job description. Think of it as training your new intern. Paste this into the System Message field:

You are an expert Sales Development Representative. Your job is to write a concise, personalized, and compelling first-touch email.

You will receive a lead's name and their company domain. You must:
1. Acknowledge the lead's role and company.
2. Identify what the company does (make a smart assumption based on the domain or common knowledge).
3. Write a short, 3-4 sentence email that introduces my service/product (an AI Automation Agency) and explains how it can help their business save time and scale.
4. Do not use fluff. Be direct and professional.
5. Your output should ONLY be the email body text.

Now, we need to feed the agent our lead data. In the Input for Agent field, we’ll use an expression to combine our lead’s name and company.

Step 3: The Output – Creating a Draft Email

The AI has written our email. Now we need to put it into our drafts folder.

Add a new node after the AI Agent. Search for Gmail.
Connect your Gmail account.
Choose the Operation: Create a draft.
In the To field, we need the lead’s email. Since our example only has the name and domain, we’ll generate a likely email. For this example, we’ll use an expression. A common pattern is firstname.lastname@company.com. Let’s assume that for our example. In a real-world scenario, you might use an enrichment tool like Hunter.io or a custom AI step to guess the email format. For now, we’ll hardcode a placeholder for demonstration: lead@company.com or use an expression like {{ $json.firstName.toLowerCase() }}.{{ $json.lastName.toLowerCase() }}@{{ $json.companyDomain }} if your data is structured that way.
For the Subject, we can ask the AI to also generate a subject line, or we can write a simple one using an expression. Let’s use an expression for the subject: Quick question about {{ $json.companyDomain }}. (Note: In a real setup, you’d get the subject from the AI agent’s full output).
For the Message Body, this is the key. We need to map the output from our AI Agent to this field. In n8n, you can use an expression. Click on the expression editor (the little fx button) and select the AI Agent’s output. It will look something like {{ $json.output }}. This tells n8n: “Take the text the AI generated and put it in the body of this draft email.”

Complete Automation Example: The ‘Ackerman Security’ Lead

Let’s watch our creation in action. Imagine our Schedule Trigger finds a new row in our lead sheet:

{
  "fullName": "Patricia Ackerman",
  "companyDomain": "ackermansecurity.com"
}

Here’s the n8n execution log in our heads:

Trigger: Fires. “New lead detected: Patricia Ackerman from ackermansecurity.com”.
AI Agent: Receives the data. It knows its job is to write an email. It infers that Ackerman Security is likely a security or protective services company. It crafts a message: “Hi Patricia, I saw you lead the team at Ackerman Security. My agency specializes in automating critical business workflows, which could help your team spend less time on admin and more time on protecting clients. Would you be open to a 10-minute chat next week?” It sends this text to the next step.
Gmail Node: Receives the AI’s text. It creates a draft. The ‘To’ field is prepped (e.g., patricia.ackerman@ackermansecurity.com), the subject is “Quick question about ackermansecurity.com”, and the body is the AI’s perfectly crafted message. It’s now sitting in Patricia’s drafts folder, waiting for a final review before sending.

Boom. In seconds, you went from raw data to a high-quality, ready-to-send outreach draft. You’ve just outsourced the boring part of sales to a robot.

Real Business Use Cases (Beyond Just ‘Sales’)

This pattern is a universal automation engine. Here are five ways you can use this exact workflow for different business needs:

The Real Estate Agent: New lead comes in from a Zillow inquiry. The AI drafts an email referencing the specific property they were looking at and offers a neighborhood guide. Problem Solved: No more generic “Thanks for your interest!” emails.
The Freelance Recruiter: A new candidate uploads a resume. The AI compares their skills to a job description and drafts an outreach email highlighting specific matches. Problem Solved: Personalized candidate outreach at scale.
The E-commerce Store: A user abandons their cart. The AI drafts a recovery email mentioning the specific items left behind and offering a small, time-sensitive discount. Problem Solved: Recovering lost sales automatically.
The Consultant: A potential client fills out a ‘Contact Us’ form describing their problem. The AI analyzes the problem description and drafts a response that acknowledges the pain point and suggests the next step (e.g., a diagnostic call). Problem Solved: Instant, intelligent response to complex inquiries.
The Networking Pro: After a conference, you have a list of new contacts. The AI can draft a “great to meet you” email for each person, referencing the event and the topic you discussed. Problem Solved: Following up with everyone without spending your whole weekend on it.

Common Mistakes & Gotchas

Everyone trips over the same hurdles at first. Here’s how to avoid them:

The ‘Spam’ Trap: Your AI prompt is everything. If you tell it to “sell hard,” it will write generic, spammy trash. Your prompt must be focused on personalization and value, not pushing for a sale.
Credential Timeout: n8n connections (like Gmail) need to be re-authenticated periodically. If your workflow stops working, check your credentials first. Don’t panic.
Bad Data In, Bad Data Out: If your lead list has broken company names or domains, your AI will get confused. Always validate your input data. A simple check in your workflow can save you a headache.
Forgetting the Human: I’ll say it again: always send drafts first. Let the AI do the grunt work, but you are the final quality check. Your reputation is on the line.

How This Fits Into a Bigger Automation System

This sales assistant is a single, powerful module. But its true power is when you plug it into a larger system.

CRM Integration: Instead of a Google Sheet, your trigger could be a new contact created in HubSpot or Salesforce. The final email draft could be logged as an activity on that contact’s record.
Multi-Agent Workflows: Imagine a two-agent system. Agent 1 researches the company and writes the email. Agent 2 acts as a ‘Critique Agent’—it reviews the first agent’s draft for tone, clarity, and length before it ever gets to your inbox.
Voice Agents: What if an email reply comes in asking for a call? A voice agent could handle the initial scheduling call, transcribe it, and update your CRM, all without you picking up the phone.
RAG Systems: You could connect this AI Agent to your company’s knowledge base (using RAG). Now, the AI can reference your specific case studies or pricing plans in the email, making it even more relevant.

What to Learn Next

You’ve just built the front-end of a self-driving sales machine. But what happens when someone replies?

In the next lesson, we’re going to build the other side of this coin: The ‘Inbox Zero’ Responder. We’ll create an AI that monitors your inbox, reads incoming replies, understands intent, and drafts context-aware responses for you.

You’re building a closed-loop system. The outreach bot and the response bot, working together. That’s how you truly scale.

Until then, go set up your n8n account. Class is in session.