How to Write Win-Back Emails with AI (Prompts + a Real Workflow)

Every list is quietly leaking. Subscribers don’t unsubscribe — they just stop opening, and each quiet month makes the next send a little worse for everyone (engagement is a deliverability signal, and dead weight drags it). The win-back is the highest-leverage campaign most programs never run. Here’s the whole play, with the briefs to run it.

Step 1: Define “lapsed” from your data, not a blog post

The classic threshold is “no opens in 90 days,” but your program has its own rhythm. A daily newsletter’s lapsed point is not a monthly digest’s. Start with the question:

Brief: “How many subscribers haven’t opened anything in 60 days? How does that compare to 90 and 120 days?”

You’re looking for the elbow — the window where the group is big enough to matter but not so cold it’s hopeless. (This is exactly the kind of question conversational analytics answers from your live data; with other stacks, export and count.)

Step 2: Build the segment — and slice out buyers

Not all lapsed subscribers are equal. Someone who bought and went quiet is worth a different email than someone who never engaged at all.

Brief: “Build a segment: purchased in the last 12 months, no opens in 60 days. And a second one: never purchased, no opens in 90 days.”

With Airis this produces production segments directly; elsewhere, it’s your criteria checklist for the segment builder.

Step 3: Write the email — short, and honest about the situation

Win-back copy has one job: acknowledge the silence without guilt-tripping, and make re-engaging one click. The brief that works:

Brief: “Write a win-back email for lapsed buyers, in my usual voice, under 90 words. Acknowledge they haven’t opened in a while without being weird about it. One CTA. Three subject-line options with genuinely different angles — not three phrasings of one idea.”

Notes that separate good win-backs from spam-folder fodder:

• Subject lines should differ in angle: curiosity (“Still want these?”), utility (“What you’ve missed”), and directness (“Should we stop emailing you?”) test differently. Three rewordings of one angle test nothing.
• The unsubscribe offer is a power move. “If this isn’t useful anymore, unsubscribe here” costs a few dead contacts and signals confidence — and every voluntary exit is one less quiet drag on your engagement rates.
• In your voice. A win-back that sounds like a different company confirms the reader’s suspicion that nothing valuable is here. (Voice-learned drafting is the fix.)

Step 4: Make it a sequence, not a send

One email recovers the easy cases. The full play is two or three:

1. The value reminder — best-of content or what’s changed since they left.
2. The direct ask (a week later, to non-openers) — “should we stop?”
3. The goodbye (optional) — last chance, then a sunset to a suppression-safe state.

Brief: “Turn this into a 3-step flow: email 1 now, email 2 after 7 days to non-openers, then flag anyone still quiet for sunset. Draft all three.”

Ship it as a drafted automation you review and publish — win-backs are the canonical “build once, runs forever” flow.

Step 5: Read the result like an analyst

The win-back’s success metric isn’t opens on the win-back — it’s what the recovered cohort does next month. Ask for it:

Brief: “Of the win-back recipients, how many opened something in the 30 days after? Compare their engagement to the list average.”

Then keep the loop going: the thresholds you learned in Step 1 become a standing monitor, and the next win-back builds itself. That loop — question → segment → send → automation — is the whole conversational email marketing model in miniature.