Complaints and difficult messages

Difficult messages — complaints, refund demands, price objections, mistakes you made — are where AI helps most and where it must be supervised most closely. Its genuine advantage is emotional neutrality: it will not write the defensive sentence you would write at 9pm after a long day. Its weakness is that it does not know what you are willing to concede, so it will happily draft an apology that promises something you cannot deliver. The reliable structure is fixed and worth memorizing: acknowledge the specific problem in their words, take responsibility for what is genuinely yours, state exactly what you will do, and say by when. Instruct the tool to follow it and to avoid three failure patterns — over-apologizing, which reads as guilt; explaining your internal processes, which reads as excuses; and vague future assurances, which read as nothing. Give it your actual remedy boundaries in the prompt, because a remedy you did not authorize is worse than a slow reply. Two controls make this safe in practice. First, never let a difficult reply go out on the same pass it was drafted; read it once more, aloud if you can, before sending. Second, escalate rather than draft when the message involves injury, legal threats, discrimination claims, regulated advice, or a customer relationship large enough that a wrong sentence has real cost. AI helps you respond well to a bad day; it does not decide what your business owes anyone. The hardest part of a complaint reply is deciding what you owe, and that decision has to be made before you prompt. If you are still undecided, the draft will read as undecided. A short reply almost always outperforms a long one here. Length reads as defensiveness, and every additional sentence is another chance to say something you will regret in writing. A contractor who adopted a strict four-part structure and a 120-word ceiling found that his refund requests dropped, because customers stopped reading his explanations as excuses and started seeing a specific commitment.

Key takeaways

• The reliable structure is fixed and worth memorizing: acknowledge the specific problem in their words, take responsibility for what is genuinely yours, state exactly what you will do, and say by when.
• Two controls make this safe in practice.
• Without stated boundaries the tool may invent a remedy you cannot honor, which turns one problem into two.

Sign in to complete activities, save progress, and use the AI Coach.