How to use these
Copy a prompt into Claude (or your AI assistant of choice), fill in the bracketed parts, run it, and iterate. These produce a strong first draft — a person reviews and approves before anything goes to a customer. Keep what works in your own context library.
Response drafting
What it’s for: a fast, on-brand first draft of a customer reply.
Draft a reply to this customer message:
[Paste message]
Context: customer relationship is [strong / strained / new]; the issue is [resolved / in progress / can't be fixed].
Tone: warm, clear, accountable. Acknowledge the issue, say what we're doing, set a concrete next step. No corporate hedging. Keep it short. Give me two versions — one a touch warmer, one a touch more formal.Escalation triage
What it’s for: deciding fast what needs to jump the queue.
Here are [N] open tickets:
[Paste tickets]
Triage each: priority (P1-P4) with a one-line reason, anything that's a churn or legal/PR risk, and which need a human now vs. can wait. Output as a sorted table, highest risk first.Knowledge-base article from tickets
What it’s for: turning repeat questions into a reusable help article.
We keep getting this question: [topic]. Here are a few example tickets and how we resolved them:
[Paste]
Write a clear knowledge-base article: the question, a short answer up top, then step-by-step detail, plus "if that doesn't work" troubleshooting. Plain language, screenshots noted where they'd help.Ticket theme analysis
What it’s for: finding what’s actually driving contact volume.
Here is a sample of [N] recent tickets:
[Paste tickets]
Identify the top 5 themes by volume, the likely root cause of each, and which are product issues vs. education/UX issues. Recommend the one fix that would reduce contact volume most. Quote real customer phrasing where it's vivid.Churn-risk flag
What it’s for: spotting at-risk accounts in the support signal.
Based on this account's recent support history and usage:
[Paste tickets / sentiment / usage]
Assess churn risk (low/medium/high) with reasoning, the specific signals behind it, and 3 concrete retention actions the account team should take this week. Be honest if the signal is weak.QA review of agent responses
What it’s for: consistent quality coaching without nitpicking.
Review these agent responses against our standards (accurate, clear, warm, sets a next step):
[Paste responses]
For each: a 1-5 score, what was strong, and one specific improvement. Then name the single coaching theme across all of them. Constructive, not nitpicky.Macro / template generation
What it’s for: building reusable response templates for common cases.
Create response templates for our 8 most common support scenarios: [list scenarios].
Each template: a warm opener, a placeholder for the specifics, a clear resolution/next step, and a friendly close. Written so an agent can personalize in 30 seconds. On-brand, human, never robotic.Want this running in your help desk?
This is a starter set. Inside an AI Office engagement, your team turns the highest-value of these into running workflows — draft-and-review reply assistants, triage, and KB generation inside your help desk — with a person approving every customer-facing reply. When one is worth shipping as a tool, a Value Sprint delivers it in weeks. Start a conversation at frogslayer.com.