How to Build an AI Email Outreach System with N8N and OpenAI

By Monk Media One Tech | AI Automation Agency, Ahmedabad, India

Generic cold emails get a 2–3% reply rate. Personalized ones — where the opener references something specific about the recipient — get 8–12%. The problem: writing personalized openers at scale takes more time than most teams have.

This N8N workflow automates an AI Email Outreach System end-to-end: prospect research, AI-written first lines, scheduled sending, reply detection, and follow-up sequences.

Workflow Architecture

Lead List (Google Sheets / Apollo.io API)
         │
         ▼
Enrich: Company news, LinkedIn data
         │
         ▼
OpenAI: Write personalized first line
         │
         ▼
Assemble full email from template
         │
         ▼
Gmail: Send (rate-limited, scheduled)
         │
         ▼
Gmail Watch: Monitor for replies
         │
    ┌────┴────┐
  Reply    No Reply
    │         │
 Update    Follow-up
  CRM      sequence
            (Day 3, 7, 12)

Step 1 — Pull Leads from Apollo.io

Apollo.io has a database of 270+ million contacts. Use their API to pull targeted leads:

// N8N HTTP Request node
{
  "url": "https://api.apollo.io/api/v1/mixed_people/search",
  "method": "POST",
  "headers": { "Content-Type": "application/json" },
  "body": {
    "api_key": "YOUR_APOLLO_KEY",
    "q_organization_domains": ["target-industry.com"],
    "person_titles": ["CEO", "Founder", "COO", "Head of Operations"],
    "per_page": 25,
    "organization_industry_tag_ids": ["5567dd9973696439b10b0000"]
  }
}

Filter out leads missing email addresses or with low confidence scores.

Step 2 — Enrich with Recent Company News

For each lead, find something recent and specific to reference:

// N8N HTTP Request — NewsAPI
URL: https://newsapi.org/v2/everything
Params: {
  q: "{{ $json.organization.name }}",
  sortBy: "publishedAt",
  pageSize: 3,
  from: "[30 days ago]",
  apiKey: "YOUR_NEWSAPI_KEY"
}

Extract the most relevant headline. If no news found, fall back to their LinkedIn headline or recent company milestone from Apollo data.

Step 3 — AI-Personalized First Line

This is the ROI driver. Send each lead's data to OpenAI:

System prompt:
You write cold email opening lines for a B2B AI automation agency based in Ahmedabad, India.

Write ONE opening sentence (maximum 20 words) that:
- References something SPECIFIC about this person or company
- Sounds natural and human, not like a template
- Shows you've done real research
- Does NOT use: "I noticed", "I loved your", "I came across", "Congratulations on"

Lead info:
Name: {{ $json.first_name }} {{ $json.last_name }}
Title: {{ $json.title }}
Company: {{ $json.organization.name }}
Industry: {{ $json.organization.industry }}
Recent news: {{ $json.recent_news }}
LinkedIn headline: {{ $json.linkedin_headline }}
Company size: {{ $json.organization.estimated_num_employees }} employees

Return ONLY the sentence. No explanation.

Sample outputs:

• "Scaling to 200 employees in two years must have the operations team running at full capacity."
• "Entering the Toronto commercial real estate market right as interest rates stabilize — good timing."
• "Your AI-first approach to HR sounds like it's attracting exactly the kind of talent acquisition clients who would benefit from what we do."

Step 4 — Email Template Assembly

// N8N Code node
const first_name = $json.first_name;
const first_line = $json.ai_first_line;
const company = $json.organization.name;
const industry = $json.organization.industry;

const email_body = `${first_line}

I'm reaching out because we work with ${industry} companies to automate their most time-consuming operational workflows — typically saving 10–20 hours per week per team.

We recently helped [reference client in same industry] reduce their [specific pain point] by 70% using a combination of N8N automation and AI agents.

Would a 20-minute call make sense to explore if there's a fit?

Best,
Dhruv Desai
Monk Media One Tech | monkmediaone.tech
+91 88668 19349`;

return [{ json: { 
  to_email: $json.email,
  to_name: `${first_name} ${$json.last_name}`,
  subject: `${company} + AI automation`,
  body: email_body
} }];

Step 5 — Scheduled Sending

Don't blast all emails at once. Schedule for optimal delivery times:

// N8N Code node — scheduling logic
const recipient_timezone = $json.organization.country_code === "IN" ? "Asia/Kolkata" : "America/New_York";

// Best times: Tuesday–Thursday, 9am–11am local time
const scheduled_time = getNextBusinessDayMorning(recipient_timezone);

return [{ json: { ...originalData, send_at: scheduled_time } }];

Send via Gmail node with a 2-minute delay between each email to avoid triggering spam filters.

Step 6 — Reply Detection and Follow-Up

Set up a Gmail Watch trigger in N8N that monitors your inbox:

• If reply detected → mark lead as "replied" in Sheets, stop follow-up sequence, notify sales rep
• If no reply after 3 days → send follow-up 1
• If no reply after 7 days → send follow-up 2 (share case study, no ask)
• If no reply after 12 days → send final follow-up, remove from sequence

Our Results After 60 Days

Without the personalized first line (tested for 2 weeks), reply rate was 3.8%. The AI personalization was the single biggest variable.

Compliance Notes

• Include unsubscribe option in every email
• B2B cold email to business addresses is legal in India under IT Act
• For EU recipients: GDPR requires a legitimate interest basis — document this
• For Canadian recipients: CASL requires prior consent — consider opt-in sequences instead

We're Monk Media One Tech — AI automation agency, Ahmedabad. We build N8N workflows, sales automation, and AI systems.

Book a free discovery call: monkmediaone.tech/contact

📞 +91 88668 19349 | hello@monkmediaone.tech