Draft a Reply to a Frustrated Customer

Prompt

Draft a reply to this customer message. Requirements: acknowledge their specific problem in the first sentence (not a generic apology), explain what I can actually do about it based on the facts below, give one clear next step with a timeframe, and keep it under 150 words. Don't promise anything not listed in what I can offer.

Customer message:
{{customer_message}}

What I can actually offer:
{{options}}

Fill in these placeholders

• {{customer_message}}
• {{options}}

How to use it

Be honest in the options field — the prompt is designed to stop the AI from promising refunds or fixes you can't deliver. Read the draft once as if you were the customer before sending.

Example use case

Replying to a customer whose order arrived late, offering what your policy actually allows.

Why this prompt works

The instruction to acknowledge their specific problem in the first sentence prevents the generic apology opening that makes support replies feel automated. Listing what you can actually offer is the important constraint: without it, models promise refunds and fixes that do not exist, which turns one unhappy customer into a much bigger problem.

Common mistakes

• Being vague in the options field, which is exactly how AI ends up promising something you cannot deliver.
• Sending it unread. This message goes to an already-annoyed person, and one wrong detail confirms their suspicion.
• Over-apologizing. Repeated sorries read as evasion rather than accountability.

Variations worth trying

• Add a maximum word count. Frustrated customers do not read long replies.
• Ask for a version that leads with the fix and apologizes second, which often lands better.
• Ask what question the customer is likely to ask next and answer it pre-emptively.