Rebuilt lifecycle email with AI segmentation and lifted trial-to-paid by 38%

The challenge

Fieldnote's 14-day trial converted at 9%, and every trial user got the same five emails regardless of what they did in the product. Sales had no signal on which trials were worth a call.

What I did

1. Mapped the activation moments that predicted conversion with the data team using Amplitude cohorts
2. Designed six behavior-triggered journeys in Customer.io keyed to those moments
3. Built a prompt pipeline that drafted persona-specific email variants, with a review rubric so brand and legal could approve in one pass
4. Added a product-qualified lead score that routed the top 15% of trials to sales in Slack
5. Ran holdout tests on every journey for four weeks before rolling out

Who I worked with

Led a working group of one data analyst, one PM and the sales lead; owned the roadmap and weekly readout.

What I learned

The AI saved time, but the lift came from triggering on behavior. The model was most useful for variation, not strategy.