Cold email works when it’s genuinely personalized and delivers clear value. AI makes genuine personalization at scale possible. Here’s how to do it without looking like you used AI.
The Two Cold Email Sins
Sin 1: Generic templates that feel like templates. “I came across your profile and was impressed by your work” is not personalization.
Sin 2: Obvious AI output. Phrases like “I hope this message finds you well” and “I wanted to reach out” are immediate delete signals.
The goal: AI does the personalization research faster, but the output sounds genuinely human.
Research-First Cold Email
Prospect Research Prompt
Before writing, research the prospect:
Research this prospect for a cold email:
Name: [Name]
Title: [Title]
Company: [Company]
LinkedIn: [paste profile text or describe]
Recent news: [any recent company news or announcements]
Our product: [description — 1 sentence]
Problem we solve: [specific problem]
Find:
1. The most specific personal hook (something real about them or their work)
2. A business pain they likely have given their role and company
3. How our product could specifically help their situation
4. The best angle for outreach — what would make them actually want to reply?
Be specific. "They're growing" is not specific. "They hired 3 engineers last month per LinkedIn" is specific.Writing the Email
Write a cold email to [Name] at [Company].
Research:
- They just [specific trigger — closed a funding round, launched a product, posted about a challenge]
- Their role involves [relevant responsibilities]
- Their company [relevant company context]
Our product: [description]
Angle: [specific relevance to them]
Requirements:
- Subject line: 5-8 words, no questions, no "quick" or "help"
- Opening line: references the specific thing about them (not "I came across your profile")
- Body: 2-3 sentences connecting their situation to what we do
- Proof: one specific result (customer name and outcome, not "companies like yours")
- CTA: one specific, easy ask
- Total: under 90 words
Do NOT include: "Hope this finds you well", "I wanted to reach out",
"I'm reaching out because", "I came across your profile"Email Templates by Trigger
Trigger: Job Change
Write a cold email for someone who just started a new job:
Person: [Name] just became [new title] at [company]
They came from: [previous role]
Our product: [description]
Why relevant: New leaders often evaluate existing vendors and tools in the first 90 days
Angle: Welcome to the role + here's something that might help your new challengesTrigger: Company Funding
Write a cold email for a company that just raised funding:
Company: [name] raised [amount] from [investors]
Contact: [name, title]
Our product: [description]
Connection: [how our product helps companies that just raised — hiring, scaling, etc.]
Angle: Congratulate + connect their growth plans to our product's valueTrigger: Job Posting
They're hiring for [role]. This signals they care about [problem].
Company: [name]
Our product: [description — relevant to what they're building]
Angle: "You're hiring [role] which means you're focused on [X] — we help with exactly that"Subject Lines That Work
Write 10 subject line variations for a cold email about [offer/product].
Target: [job title] at [company type]
Main value proposition: [what you do for them]
Principles:
- Curiosity without clickbait
- Specific to their situation
- Under 8 words
- No questions
- No "help", "quick", "just"
Provide: 10 options across different angles (personal, problem-led, proof-led, direct)
Then recommend the top 3 for A/B testing.High-performing subject line patterns:
| Pattern | Example |
|---|---|
| [Company] + [observation] | “Noticed your team doubled last quarter” |
| Name + question | ”[Name], have you solved [problem]?” |
| Direct benefit | ”[3x result] for [company type]“ |
| Mutual connection | ”[Person] suggested I reach out” |
Multi-Step Sequences
Write a 5-email cold email sequence for:
Prospect type: [title at company type]
Our product: [description]
Sequence goal: book a 20-minute call
Email 1 (Day 1): Initial outreach — [specific angle based on research]
Email 2 (Day 4): Different angle — add value or new proof point
Email 3 (Day 8): Social proof — relevant case study
Email 4 (Day 15): Try a different approach — shorter, more direct
Email 5 (Day 22): Breakup email — final attempt, leave on good terms
Requirements:
- Each email references the previous only briefly
- Each adds new value, doesn't just "bump" the thread
- Progressive calls to action (reply? vs. 20-min call? vs. OK if not right fit?)
- Sequence should build a relationship, not pressure themWriting More Like a Human
After AI generates the email, run it through this filter:
Make this cold email sound more human and less like it was written by AI:
[paste AI-generated email]
Changes to make:
1. Remove all corporate buzzwords ("leverage", "synergies", "solutions")
2. Shorten sentences that are too perfectly structured
3. Add one small imperfection or casual element
4. Remove any hedging language ("I believe", "I think", "potentially")
5. Make the CTA less formal
6. Cut any sentence that starts with "I"
Keep the same message and length. Just make it sound like a person wrote it quickly.A/B Testing Framework
I'm A/B testing cold email variables. Set up a testing plan:
Variable to test: [subject lines / opening lines / CTAs / email length / offer]
Current performance: [reply rate if known]
Sample size needed: [100-200 sends minimum per variant for statistical significance]
Create:
1. Two variants to test (A and B)
2. What's different between them (one variable only)
3. How to measure winner (reply rate / positive reply rate / meetings booked)
4. How long to run the test
5. Decision criteria
One variable at a time — don't change multiple things simultaneously.Avoiding Spam Filters
AI-generated cold email at scale can trigger spam filters:
Technical:
- Use a separate sending domain (not your main company domain)
- Warm up new email accounts slowly (start at 20/day, increase over weeks)
- Keep text-to-link ratio high (more text, fewer links)
- Don’t use spam trigger words: “guaranteed”, “free”, “limited time”, “act now”
Content:
- Personalize: spam filters now detect template patterns
- Vary your templates: don’t send identical emails at scale
- Include unsubscribe option for bulk sends
Infrastructure:
- Use a dedicated IP (better for deliverability at scale)
- Set up SPF, DKIM, DMARC records
- Monitor your sender reputation with Google Postmaster Tools