Skip to content
LetsVibeAI

Autonomous GTM & Sales Automation Engine/Lab 3

Personalized Email Outreach Engine

Build an AI-powered outreach system that researches and personalizes every email.

Learning Objectives

  • Build an AI-powered outreach system that researches and personalizes every email
  • Integrate multiple AI services (Perplexity + DeepSeek) into a single automation
  • Understand the architecture of a research-driven cold email pipeline
  • Implement deliverability best practices for cold outreach

The Architecture

Google Sheet (leads)
    → Perplexity: Research prospect company
    → Perplexity: Research YOUR own products
    → DeepSeek: Draft personalized email
    → Resend: Send email
    → Google Sheet: Update status + timestamp

Cost to run: ~$5/month using free tiers and minimum credit purchases.

Step-by-Step

Step 1: Set Up Your Lead Sheet

Create a Google Sheet named "Leads" with these exact column headers:

first_name last_name email company_name company_domain prospect_linkedin title source status email_1_date email_1_message_id notes

Leave email_1_date and email_1_message_id blank for new leads — these act as sentinel columns.

Step 2: Configure Make.com Modules

Module 1 — Google Sheets: Search Rows

  • Query: email_1_date is empty
  • Limit: 5-10 leads per run (to manage API costs)
  • Add an Iterator to process rows one by one

Module 2 — Perplexity: Company Research

Prompt template for Perplexity API:

Research the company: 
- Company: {{company_name}}
- Domain: {{company_domain}}  
- Prospect LinkedIn (if provided): {{prospect_linkedin}}

Deliver a concise, factual brief:
1) What they do
2) ICP and segments
3) Recent news (last 90 days)
4) Competitive context
5) Potential pain points related to [your industry]

Output: Save to a company_research_summary variable.

Module 3 — Perplexity: Your Product Knowledge Snapshot

Run a second research call so the writer model knows what you sell.

Inputs: your primary domain and product pages (or a brief about what your company does).

Prompt:

Analyze this product/service:
[your product description]

Deliver:
1) Key value propositions
2) Target customer profile
3) Features most relevant to [target industry]
4) Case study or success story snippets

Output: Save to a product_snapshot variable.

Alternative: If you don't want to use Perplexity twice, replace one call with:

  • Gemini with web search enabled
  • OpenAI with web search (browsing mode)
  • Simple HTTP module to scrape the company homepage

Module 4 — DeepSeek: Draft Personalized Email

Combine the research outputs:

You are a sales email writer. Using the following research, write a short cold email body.

Company Research:
{{company_research_summary}}

Our Product:
{{product_snapshot}}

Prospect Name: {{first_name}} {{last_name}}
Prospect Title: {{title}}

Rules:
- No subject line — body only
- Max 150 words
- Reference something specific from the company research
- Clear, low-pressure call to action
- Professional but conversational tone

Module 5 — Resend: Send Email

  • To: {{email}}
  • From: your verified domain
  • Subject: {{first_name}}, quick question about [topic from research]
  • Body: DeepSeek output (plain text)

Module 6 — Google Sheets: Update Row

  • Map to the original row
  • Set email_1_date to {{now}}
  • Set email_1_message_id to the Resend message ID

Step 3: Schedule and Monitor

  • Run as a scheduled scenario (once daily)
  • Start at 5-10 emails/day, gradually increase
  • Monitor open rates and adjust subject line patterns

Alternative Providers

If Perplexity or DeepSeek aren't available:

Step Alternatives
Web Research Gemini + web search, OpenAI browsing, Google Custom Search API
Email Writing GPT-4, Claude, Gemini, any chat completion API
Sending SendGrid, AWS SES, Postmark

Deliverability Considerations

Cold email is a high-risk channel. Protect your domain:

Practice Why
Use a subdomain Isolates cold email reputation from primary domain
Warm up gradually Start with 5/day, increase by 5 every 3 days
Monitor bounce rates Keep under 2% — anything higher triggers spam filters
Personalize genuinely Generic templates get flagged immediately
Include unsubscribe Required by law (CAN-SPAM, GDPR)

Key Takeaways

  • A personalized outreach engine can be built for ~$5/month using free tiers
  • The three-step research-draft-send pipeline produces higher-quality outreach than batch-and-blast
  • Two research calls (company + product) gives the writer everything it needs
  • Plain text emails with genuine personalization outperform HTML templates
  • Domain warming and deliverability are the hardest part — not the automation

Extension Ideas

  • Add Perplexity research on the specific prospect (not just company)
  • Multi-step sequences: Day 1 email, Day 4 follow-up, Day 10 break-up
  • A/B test subject lines across different lead segments
  • Track replies via Gmail/Outlook webhook and auto-tag as "Replied"
  • Add LinkedIn outreach as a parallel channel