How we launch 28 cold email campaigns in a single day

Here's how we launch 28 cold email campaigns in a single day.

That's how you actually create a self-improving AI Email engine.

I was pretty overwhelmed, with how to split-test campaigns, so I've came up with this framework:

There are 2 dimensions you can segment from:

WHO (the targeting): → industry → company size → geography → title → techstack

WHAT (the copy): → subject line → icebreaker → the pitch → the CTA

To make it actually measurable you test ONE variable at a time and measure to impact of that one.

Here is a rundown of a recent launch that I've done:

- 68,000 US leads in one table - split into 7 industries - each industry get's split by size, 10-50 and 50-200 = 14 clean audiences - 2 variants each, one variable different = 28 campaigns

To make it repeatable, I've created several claude code skills.

Claude code slices the TAM, tags every lead by industry and size, writes a personalized email per lead, and logs all 28 variants to one table in OXYGEN, so I can see which audience actually books calls.

The results of the campaigns gotta be logged in OXYGEN to inform the agent which angles and segments have performed best, so that it can double down on the winning formula.

DM me if you have any questions:)