Don't A/B test on your customers.

2 min read Original article ↗

Agentic A/B testing for marketing email

Pit two candidate emails against a fleet of 100 LLM persona agents in a realistic competitor inbox. Get open rate, click rate, and simulated revenue — in minutes, before you risk a send.

The problem

Email marketing is a critical channel for e-commerce, and A/B testing subject lines and copy is standard practice. But if Variant A is 10× better than Variant B and you split your list 50/50, you just torched half your potential revenue sending the loser.

Don't leave money on the table. Get A/B test results in minutes for your marketing emails — on a fleet of agents grounded in your actual customers — before you ever press send.

How it works

You pick two candidate emails for a brand. We assemble a 100-email inbox — your candidate plus 99 real emails from competitor and adjacent brands. Then we run 100 autonomous LLM persona agents, each one grounded in your target customer psychographics.

01

Open

Up to 20 emails per agent

02

Click

Up to 5 emails per agent

03

Purchase

Up to 2 products per agent

The output: open rate, click rate, and simulated revenue for A vs B.

The constraint

weekly spend

purchase budget — agents only buy what they actually want

How will your brand emails fare against your competitors?