ClawDeployCustomer Support Prompts for Every Scenario
Prompts for ticket triage, response drafting, escalation summaries, and knowledge base generation — deploy them as a 24/7 AI support agent.
These prompts handle the most common customer support scenarios with empathy and accuracy. Each is designed to be deployed directly into a ClawDeploy agent or used manually for drafting responses.
Ticket Triage
Prompts for categorizing and prioritizing incoming tickets.
Categorize this customer message: "I was charged $49 but I'm on the $28 plan. Also the mobile app crashed twice today." Identify: category, priority, sentiment, and whether it needs multiple teams.
Categories: Billing (primary), Bug Report (secondary) Priority: HIGH (billing error = revenue impact) Sentiment: Frustrated but not angry Routing: Billing team for charge investigation, Engineering for crash report Note: This is a multi-issue ticket. Recommend splitting into two tracked items.
Score these 5 support tickets by priority (P1-P4). Consider: user plan, issue severity, churn risk, and number of affected users. Ticket 1: Enterprise user can't log in. Ticket 2: Free user wants dark mode. Ticket 3: Team plan user reports data loss. Ticket 4: Pro user asks about a feature. Ticket 5: Enterprise user reports slow performance.
Priority scoring: 1. Ticket 3 [P1]: Team plan data loss — Severity: Critical, churn risk: very high 2. Ticket 1 [P1]: Enterprise login blocked — High-value user, complete blocker 3. Ticket 5 [P2]: Enterprise slow performance — Important but not blocking 4. Ticket 4 [P3]: Feature question — Low urgency, can be handled async 5. Ticket 2 [P4]: Dark mode request — Free user, cosmetic, no urgency
Analyze the sentiment and urgency of this customer message: "This is the THIRD time my agent has disconnected this week. I'm paying for a service that doesn't work. If this isn't fixed by Friday I'm switching to [competitor]."
Sentiment: Angry (escalating frustration, third occurrence) Urgency: CRITICAL — Explicit churn threat with deadline Key signals: "THIRD time" (recurring issue), "paying for" (value expectation), "switching to" (named competitor) Recommended action: Escalate to engineering immediately + have customer success reach out personally within 1 hour.
Response Drafting
Prompts for drafting empathetic, accurate customer responses.
Draft a response for a customer who was charged twice. Tone: empathetic, take ownership, provide clear next steps. Don't blame the payment processor.
Hi [Name], I'm sorry about the double charge — that shouldn't happen and I understand how frustrating it is. I've flagged your account for an immediate refund of the duplicate charge. Here's what happens next: 1. Refund will be processed within 1-2 business days 2. You'll receive a confirmation email when it's complete 3. Your subscription remains active and unaffected If you don't see the refund by [date], reply here and I'll escalate it personally. Again, sorry for the inconvenience.
A customer reports a bug with detailed steps to reproduce. Draft a response that acknowledges the issue, thanks them for the detail, and sets expectations for resolution.
Hi [Name], Thank you for the detailed bug report — the steps to reproduce are really helpful and will speed up our investigation significantly. I've forwarded this to our engineering team with high priority. Here's what to expect: - Our team will investigate within 24 hours - If we need any additional info, we'll reach out - I'll update you here once we have a fix timeline In the meantime, if you need a workaround, [suggest workaround if applicable]. Thanks for helping us make the product better.
A customer requested a feature we've decided not to build. Decline gracefully without burning the relationship. Explain the reasoning.
Hi [Name], Thanks for suggesting [feature] — I can see why it would be useful for your workflow. After reviewing it with our product team, we've decided not to add this to our roadmap. Here's the honest reason: it would add complexity for the 90% of users who wouldn't use it, and we want to keep the product simple and focused. That said, here's a workaround that might get you close: [suggest alternative]. And if your needs change or you find a creative solution, we'd love to hear about it. Appreciate you taking the time to share your ideas with us.
A customer is angry about repeated service issues and threatening to cancel. Draft a response that validates their frustration, takes ownership, and offers a concrete resolution.
Hi [Name], You're right to be frustrated, and I'm sorry we've let you down. Repeated issues are unacceptable and I take full ownership of that. Here's what I'm doing right now: 1. I've escalated your specific case to our senior engineer — they're investigating the root cause today 2. I'm applying a credit of [amount] to your account for the disruption 3. I'll personally follow up within 24 hours with an update on the fix I don't want to lose you as a customer, and more importantly, I want to make sure this never happens again. Can I set up a quick call to walk through what happened and our plan to prevent it?
Knowledge Base
Prompts for creating and maintaining support documentation.
Based on these 50 common support tickets, generate the top 10 FAQ entries. Each should have a clear question and a concise, accurate answer.
1. **How do I change my subscription plan?** Go to Dashboard → Billing → Change Plan. Select your new plan and confirm. Changes take effect immediately. If upgrading, you'll be charged the prorated difference. 2. **Can I deploy my agent to multiple channels?** Yes. Go to your agent settings and add additional channels. You can run the same agent on Telegram, Discord, and WhatsApp simultaneously. 3. **My agent isn't responding — what do I do?** First, check your agent status in the Dashboard (green = running). If it shows as stopped, click "Restart." If the issue persists, check that your API key is still valid in Settings → API Keys...
Write a troubleshooting guide for "Agent not responding" — the most common support issue. Include: common causes, step-by-step diagnosis, and resolution for each cause.
# Agent Not Responding — Troubleshooting Guide ## Quick Check (30 seconds) 1. Dashboard → Agents → Check status light (green/yellow/red) 2. If red: Click "Restart Agent" → Wait 30 seconds → Test again ## Common Causes ### Cause 1: API key expired Symptoms: Agent was working, suddenly stopped Fix: Settings → API Keys → Check expiration → Generate new key → Paste into agent config ### Cause 2: Message channel disconnected Symptoms: Agent runs but messages don't arrive Fix: Settings → Channels → Check connection status → Reconnect if needed...
We just launched a new feature: "Scheduled Messages." Write a help center article explaining what it does, how to set it up, and common use cases.
# Scheduled Messages Send messages from your AI agent at specific times — perfect for daily digests, weekly reports, and automated check-ins. ## How It Works Scheduled Messages let your agent send pre-configured messages to your channels on a schedule you define. ## Setup (2 minutes) 1. Go to your Agent → Settings → Scheduled Messages 2. Click "Add Schedule" 3. Choose: frequency (daily, weekly, custom), time, and channel 4. Write your message template or prompt 5. Save and activate ## Common Use Cases - **Daily standup prompt:** "Good morning team! What are you working on today?" - **Weekly report:** "Here's your weekly metrics summary: [auto-generated]" - **Customer check-in:** "Hi! Just checking in — everything working well?"
Frequently Asked Questions
Can I deploy these prompts as an always-on support agent?
Yes! That's the ideal use case. Deploy a ClawDeploy agent with these prompts built into its instructions. It handles tier-1 support 24/7 and escalates complex issues to your team.
How do I maintain accuracy as my product changes?
Update your agent's knowledge base whenever you ship features or change pricing. The prompts adapt to the knowledge base content you provide.
What about sensitive customer data in prompts?
Never include real customer data (emails, account numbers) in prompts. Use placeholders and let the AI agent pull context from the conversation. ClawDeploy doesn't store customer data.
How do I measure if AI support is working?
Track: resolution rate (% of tickets handled without human), response time, customer satisfaction scores, and escalation rate. A good target: 50-70% AI-resolved in the first month.