The two things called AI cold email
The term gets applied to two approaches with very different results:
| Approach | How it works | Result |
|---|---|---|
| Template variation | GPT rewrites one template 500 times with company name + job title swapped | 1-2% positive reply rate, declining |
| Per-prospect research | AI reads each prospect's site, finds a specific hook, writes a unique email | 4-8% positive reply rate, stable |
Key point
The distinction isn't subtle in the results. Template variation campaigns are seeing positive reply rates drop as inboxes adapt. Research-based personalization holds 4%+ positive reply rates because the content is genuinely relevant, not just syntactically varied. The subject line patterns guide covers what this looks like in practice.What the numbers look like from real campaigns
A $25 starter pack reaches roughly 350 contacts. At a 4% positive reply rate, expect about 14 replies. Realistically 4-8 useful conversations from a first campaign. Two converting at $500 MRR each: $1,000 MRR from a $25 spend. The full ROI breakdown shows how the math scales with deal size.
Pro tip
Track positive replies per campaign, not just clicks. If you're getting clicks but no replies, the subject line and offer work but the email body misses the pain. If you're getting replies from the wrong people, the ICP is off. See the reply rate benchmarks for what numbers to aim for.Why SaaS founders specifically benefit
Cold email works best when you have a specific person with a specific problem. SaaS almost always has both: a defined ICP and a concrete pain the product solves.
The old version required an SDR, a data provider, a sending tool, and 2-3 weeks of setup. Most solo founders can't afford that and don't have weeks. AI tools collapsed that stack into a $25 experiment that produces results in 48 hours.
Key point
What AI cold email doesn't replace is knowing who to email and why. The AI reads your site to understand what to say. If your homepage is vague about the problem you solve, the emails will be vague. Fixing your positioning is often higher-leverage than optimizing the outreach tool.AI-automated vs DIY cold email stack
| Element | DIY stack | distribute |
|---|---|---|
| Prospect data | Apollo: $49-149/month | Included, no per-message fees |
| Domain + inbox setup | 2 weeks setup, $20-50/month | Handled automatically |
| Email copy | Template + merge fields | AI reads each prospect's site |
| Sending tool | Instantly/Lemlist: $37-97/month | Included |
| Ongoing maintenance | 3-5 hours/week | Set budget, review replies |
| Time to first campaign | 2-3 weeks | 30 seconds |
| Cost to start | $200-500/month + time | $25 |
What consistently doesn't work
Broad ICP. "B2B SaaS companies" produces mediocre results regardless of personalization quality. "Operations managers at courier companies running their own driver fleet" is the right level. Specific enough that the email could only have been written for that person's situation.
A vague product page. The AI reads your URL to understand what you do. If your homepage leads with company values instead of the specific problem you solve, the emails will reflect that vagueness.
Asking for too much in the first email. "I'd love to give you a 430-second walkthrough" hits the same wall as any unsolicited pitch. Emails that get replies ask for something small: "Would a 130-second call this week make sense?"
Scaling before validating. 1,000 emails to the wrong ICP performs worse than 200 to the right one. Start narrow. Read the first replies carefully. Adjust before scaling.
Getting your first campaign right
The setup is 5 minutes, but those 5 minutes go better with two things ready:
A product URL that states the pain. Not your homepage if it's vague. A specific landing page, a use case page, or a blog post that leads with the problem you solve. The AI uses this to understand what to say.
An ICP description. Job title + company size + industry + one specific pain point. "Head of Customer Success at B2B SaaS companies with 20-200 employees struggling with onboarding drop-off" is the right level of specificity.
A $25 campaign reaches about 350 contacts. Expect 2-5 qualified replies in the first week. The first campaign validates targeting and messaging. The second, informed by the first, usually performs noticeably better.