Fifteen prompts for the hardest parts of support — angry tickets answered with grace, clean escalations engineering will actually read, and docs that deflect the next hundred tickets.
💡 Tired of copy-pasting? Import this whole pack into PromptDock and insert any prompt by typing // inside ChatGPT, Claude, Gemini & 7 more AI chats. Free, no account, stored on your device.
Draft a reply to this angry customer message. Acknowledge the specific frustration in their words (no generic "sorry for the inconvenience"), take ownership without over-apologizing, state exactly what happens next and when, and give them one thing they can do right now. Calm, human, under 150 words.
Their message (PII removed):
Refund request reply
A customer is requesting a refund. Our policy: {{policy}}. Their situation: {{situation}}. Draft a reply that is generous in tone whatever the outcome: if approved, make it fast and graceful; if declined, explain why honestly, offer the best alternative we can ({{alternative}}), and leave them feeling respected. Under 140 words.
Feature request response
Reply to a customer requesting {{feature}}. Be honest about status ({{status — planned / considering / not planned}}), never fake-promise, explain the reasoning in one sentence, offer the closest existing workaround, and genuinely thank them — requests like this shape the roadmap. Under 120 words.
Bug acknowledgment + workaround
Draft a reply confirming the customer has hit a real bug in {{feature}}. Own it plainly, give the workaround below with exact steps, set honest expectations about the fix timeline ({{timeline or 'no ETA yet'}}), and promise a follow-up when it ships — only if we'll actually send one.
Workaround:
Billing dispute reply
A customer disputes a charge: {{situation}}. Our records show: {{facts}}. Draft a reply that walks through the facts without being defensive, shows the exact charge breakdown, resolves it in the customer's favor where the case is ambiguous, and explains how to avoid the confusion next time. Under 150 words.
Troubleshooting
Symptoms → troubleshooting flow
Customer reports: {{symptoms}} in {{product/feature}}. Build a troubleshooting flow: most-likely causes ranked, the diagnostic question or step for each (cheapest checks first), and what result confirms or rules out each cause. Format as a decision tree I can follow live on a chat.
Repro-steps request customers answer
Write a short message asking this customer for reproduction details — but make it effortless: numbered yes/no and multiple-choice questions instead of "please describe," pre-filled guesses they can confirm ({{my guesses}}), and reassurance that this gets their issue fixed faster. Max 6 questions.
Escalation summary to engineering
Turn this ticket thread into an engineering escalation: one-line summary, impact (who/how many/severity), exact repro steps, expected vs actual behavior, environment details, what support already tried, and customer-visible deadline if any. No fluff — engineers should get it in 30 seconds.
Thread (PII removed):
Explain the technical thing simply
Rewrite this technical explanation for a non-technical customer: what happened in plain words, one everyday analogy, what it means for them practically, and what (if anything) they need to do. No jargon survives. Under 120 words.
Technical explanation:
Incident status update
Write a status update for an ongoing incident affecting {{feature/service}}: what's affected and what isn't, what we know so far (honest about unknowns), what we're doing, when the next update comes ({{time}}), and any workaround. Calm, no minimizing, no blame. Two versions: status page and direct reply to an affected customer.
Docs & Ops
Macro pack for a recurring issue
Customers keep contacting us about {{issue}}. Write a macro set: first reply (asks the one clarifying question that splits the causes), resolution reply for cause A ({{cause A}}), resolution reply for cause B ({{cause B}}), and a follow-up check-in. Use {{customer_name}} and {{order_id}} placeholders our helpdesk can fill.
Help-center article from a ticket
Turn this resolved ticket into a help-center article: a title matching what customers actually search, the problem in the customer's words, numbered fix steps with expected results, a troubleshooting section for when it doesn't work, and related-article suggestions. Scannable — bold the actions.
Resolved ticket (PII removed):
FAQ from ticket themes
Here are our most common ticket topics from the last month. Draft an FAQ: group into sections, write each question the way a customer would phrase it, answer in under 60 words each, and link placeholders to deeper articles. Order by ticket volume.
Topics + volumes:
QA rubric for support replies
Create a QA scorecard for reviewing support replies: criteria (accuracy, tone, completeness, next-steps clarity, policy compliance), 1-5 anchors for each with example phrasing at 2 vs 5, auto-fail conditions, and a weighting that reflects what actually drives customer satisfaction.
Voice-of-customer digest
Summarize these tickets into a weekly voice-of-customer digest for the product team: top 5 pain themes with counts and representative quotes, emerging issues (new this week), feature requests grouped by underlying job-to-be-done, and the one chart-worthy stat. One page max.
Tickets (PII removed):
How great support teams get better answers from ChatGPT
Paste the customer's exact words. Tone-matching starts with their message. The AI de-escalates better when it can see the frustration, not your summary of it.
Give it your policy, not just the situation. "Our refund policy is X, customer is asking Y" produces replies you can send. Without the policy you get generic apologies.
Strip personal data first. Replace names, emails and order numbers with placeholders before pasting tickets — support tools are for PII, AI chats aren't.
Macro the winners. A reply that turned an angry customer around is an asset. Save it with {{variables}} in PromptDock and every teammate can insert it with //.
Insert these with two keystrokes instead of copy-paste
Download the pack, import it once into PromptDock, and every prompt above is // away inside ChatGPT, Claude, Gemini and 7 more — with fill-in-the-blank {{variables}} that ask for the specifics as they insert.
Yes — as a drafting layer. Paste the anonymized ticket and your policy, get a draft, then review before sending. The human stays accountable for accuracy and promises made.
Are these prompts free?
Yes. Copy any in one click, or download the pack and import it into PromptDock to insert them by typing // in ChatGPT, Claude or Gemini.
Do these work with Zendesk/Intercom workflows?
The prompts are tool-agnostic — draft in any AI chat and paste into your helpdesk. The macro and help-article prompts are designed to feed whatever system you use.