Did AI make your apology warmer—or harder to believe?
Imagine asking an AI assistant to soften this: ‘We cancelled your appointment and didn't tell you.’ If the draft comes back as ‘We're sorry for any inconvenience you may have experienced,’ I'd put the original sentence back. The person already knows whether they were inconvenienced. They need to know whether you understand what happened. The Parliamentary and Health Service Ombudsman's existing NHS complaints guide explicitly advises against ‘I'm sorry you feel like that’ and recommends ‘I'm sorry X happened.’ It also asks for responsibility, an explanation and a remedy. That's complaint-handling guidance, not research on AI writing. I'd borrow it when checking an AI-assisted customer reply: keep the known mistake visible, say what you know so far, and include only a next step somebody has actually agreed to take. Don't ask the assistant to invent a reassuring explanation. In this hypothetical appointment case, ‘We don't yet know why the notification failed; I'll update you on Friday’ could be honest if the sender owns that commitment. What would you refuse to let an AI rewrite out of an apology?
Comments
Before buying an AI reply tool, I'd ask the support lead what staff can actually offer when the company makes a mistake. If a receptionist can write an apology in seconds but needs three approvals to waive the resulting fee, better wording won't get the customer much further. Give staff a clear remedy they can authorize, with a route for exceptions. Then let the assistant help explain the decision. Otherwise the team may send more polished apologies while customers still wait for somebody with permission to fix the problem.