Verified against ChatGPT · 2026-08-09
Build a welcome series that earns the right to sell instead of pitching in email one
Plans a multi-email welcome sequence around what a new subscriber or customer actually needs to succeed first, sequencing the ask so it lands after value has been delivered, not before.
The prompt
Ready to copy — highlighted parts are example details you can swap.
You are planning a welcome email series for people who just signed up or purchased, sequenced across the first days of their experience with us. WHAT SOMEONE JUST SIGNED UP FOR Signed up for a free 14-day trial of a project management tool, hasn't invited teammates or created a project yet. NUMBER OF EMAILS AND SEND WINDOW 5 emails over the 14-day trial window. BIGGEST EARLY DROP-OFF POINT About 60% never create a second project after the first one — most churn happens around day 4. THE EVENTUAL ASK Upgrade to a paid seat before the trial ends on day 14. PLANNING PHASE First, map out what a new subscriber actually needs to know or do to get their first real win, in the order they'd need it — before writing any copy, list this as a numbered sequence of jobs-to-be-done, not marketing beats. Identify which email in the sequence is positioned right before or right after the drop-off point named above, since that's the email doing the most load-bearing work in the whole series, and flag it as such. DRAFTING PHASE For each email in the sequence, write a subject line and full body copy. The first email should not mention the eventual ask at all — its only job is orienting someone who just arrived and doesn't yet trust us. Only introduce the eventual ask once at least one earlier email has already delivered a concrete piece of value the reader could act on without paying for anything further; state explicitly which earlier email earned that right before the ask appears. Every email needs exactly one call to action, not a menu of three links competing for attention. Write short, skimmable body copy assuming most recipients read on a phone and decide whether to keep reading within the first two lines. WHAT NOT TO DO Do not open any email in the series with "Welcome to the family" or an equivalent stock phrase. Do not repeat the same call to action in back-to-back emails — if someone didn't click yesterday's link, the next email should either address why or offer a different next step, not just resend the same ask. OUTPUT FORMAT 1. The jobs-to-be-done map, in order, with the load-bearing email flagged. 2. Each email as: send timing, subject line, body copy, single CTA. 3. One line stating which email first earns the right to make the eventual ask, and what it earned that right with.
Customize
Optional — swap in your own details for the highlighted parts above.
Why this works
Separating the planning phase from the drafting phase forces GPT-5.1 to commit to a jobs-to-be-done map before it starts generating persuasive copy, which matters because the model's default behavior when asked directly for "welcome emails" is to reach for a generic five-email arc — welcome, feature highlight, social proof, urgency, discount — regardless of what the actual new user needs to do first; making it plan the sequence of real jobs first, in a separate step it has to commit to in writing, anchors the later copy to your actual product's activation path instead of a template it has seen thousands of times in training data. Flagging the drop-off point as load-bearing changes what the model optimizes for in that specific email — without that instruction it treats every email in the sequence with roughly equal weight, when in reality one email is doing the real work of preventing the churn you actually measured. The rule against mentioning the eventual ask in email one addresses GPT-5.1's tendency to front-load a call to action because it was told there's an eventual ask at all — language models weight information given anywhere in a prompt as relevant to the immediate output unless explicitly told to withhold it until a condition (delivering value first) has been met in the sequence itself. Banning repeated identical CTAs in consecutive emails counters a specific failure mode where the model treats "remind them again" as sufficient sequencing logic, rather than treating an unclicked link as a signal that the previous ask's framing didn't land and needs to change, not just repeat.
What you get back
Email 1 (Day 0): "Here's what your first project should look like" — orientation only, no ask. Email 3 (Day 4, flagged as load-bearing, sits right at the drop-off point): "The teams that stick around all did this one thing by day 4" — invites creating a second project, still no upgrade ask. Email 5 (Day 12): first mention of upgrading, framed against the two wins already delivered.
Verified against
ChatGPT GPT-5.1 · 2026-08-09
Changelog
- 2026-08-09 — Initial publish, verified against ChatGPT GPT-5.1.
Need this built into your business?
If a prompt isn't enough — Google Ads management, built and maintained for you — that's Scult's day job.
EXPLORE GOOGLE ADS MANAGEMENT
