I use AI every day. It’s fast, it’s good at structure, and it rarely gets tired. But on one project it gave me advice that sounded neat and would have quietly cost a lot.
The setup
We were redesigning how a CRM tracked consent: who can get marketing email, who can’t, and why. It’s unglamorous work, but it decides who you’re allowed to talk to.
I asked an AI assistant to help me map the statuses. One of its suggestions: mark churned customers as unsubscribed. Cleaner list, fewer edge cases, less risk.
It sounded tidy. It was wrong.
Why churned isn’t unsubscribed
They’re two different things:
- Churned is a business status. They stopped paying.
- Unsubscribed is a consent status. They asked you to stop emailing them.
Someone who leaves can come back. Former customers are often your best win-back audience. They know the product and they know your team, and their reason for leaving may have been budget, timing or a feature you’ve since shipped.
Marking them unsubscribed would have cut that path off. And once someone is flagged as unsubscribed, you generally can’t send them marketing email again without them opting back in. That decision is hard to undo.
What we did instead
- Kept the two fields separate: lifecycle stage (prospect, customer, churned) and subscription status (subscribed, unsubscribed).
- Built a churned-customer segment that we could email on purpose, with relevant content, while respecting anyone who had actually opted out.
- Wrote down why, so the next person, or the next AI suggestion, doesn’t merge them again.
Most tools handle this well if you set it up on day one. Our guides to the best email marketing platform and a CRM for a small team cover what to look for.
Who owns the decision
AI optimizes for what looks clean. It doesn’t know your business model, your win-back revenue or what can’t be undone. Use it for speed and structure. But somebody still has to own the decision, and that somebody has to understand what each field actually means.