Churn-risk outreach

The automation watches account activity in your product data each day, spots the patterns that precede churn, a steady decline, a key user who stopped showing up, and drafts a warm, specific check-in from you to the right contact. The drafts wait in Gmail for your approval.

You get a ready-to-send note the week the signal appears, written as if you had been paying close attention, because the automation was.

Scoped access

• Supabase, read usage data
• Gmail, draft check-ins only

How it fails silently

Churn detection fails silently because a dip in usage is ambiguous. A team goes quiet for a week because it is a holiday, or because they finished the project the product was for and will be back next quarter. The automation sees the drop, labels the account at risk, and drafts a we-have-missed-you note, and sent to a perfectly happy customer that message lands as out of touch. The run looks like attentive customer success. To the customer it reads as a company that mistakes a calendar for a problem, and it can plant a doubt that was not there before.

What the overseer catches

After the agent drafts the note, the overseer weighs the dip against the rest of the picture, a recent renewal, a support thread that went well, a quiet week that lines up with a holiday. When the draft treats a customer as slipping away while the fuller signal says they are fine, it flags it for you rather than leaving a worried check-in queued to a happy account.

Clear, sustained decline gets a drafted check-in queued for your review without fuss.

What reaches you is the ambiguous signal: the seasonal dip, the account with a known reason to be quiet, the pattern that could go either way. Those come to you before a personal note goes out, so your outreach always reads as informed rather than automated.