Customer Support Chatbot: Handle 80% of Tickets Automatically
A well-configured support chatbot handles order status questions, returns, product FAQs, and common issues without any human involvement — 24/7. This guide builds a complete customer support chatbot that resolves 70–80% of typical support tickets automatically.
Before You Start
You’ll need:
- A working chatbot (Set Up First Chatbot)
- Content-Aware mode enabled (Content-Aware Mode guide)
- AIWU Pro with API key configured
- Your support content: return policy, shipping info, FAQs, common issues
- Time needed: 30–45 minutes for full setup
- Plan required: Pro (embeddings + chatbot training)
The 5 Components of a Support Chatbot
A support chatbot that actually resolves tickets needs all five:
- Knowledge base — your policies, FAQs, and product info as training data
- Clear instructions — system prompt that defines behavior and limits
- Escalation path — what happens when the chatbot can’t help
- Correct placement — where on your site it appears
- Regular updates — keeping knowledge current as policies change
Step 1: Prepare Your Knowledge Base
Collect the content your support chatbot needs to know. For most online stores, this includes:
| Content type | Example topics |
|---|---|
| Returns & refunds | Return window, conditions, how to initiate, refund timeline |
| Shipping | Carriers, delivery times, international shipping, tracking |
| Orders | How to check status, change/cancel an order, missing orders |
| Products | Sizing guides, compatibility, care instructions, FAQs per product type |
| Account | Password reset, email change, deleting account |
| Payment | Accepted methods, payment issues, invoices |
Add each topic to your embeddings dataset. Follow the Embeddings in 10 Minutes guide if you haven’t created a dataset yet, then connect it to your chatbot using the Training guide.
Step 2: Write the Support System Prompt
Go to your chatbot’s Context tab. Replace the default instructions with a support-focused prompt:
You are a customer support assistant for [Store Name]. Your job is to help customers with questions about their orders, returns, shipping, and products.RULES:
1. Always answer from the knowledge base first. If the answer is there, use it exactly.
2. For order status questions: always ask for the order number first before trying to help.
3. If you don't have enough information to fully answer, say: "For this specific question, I'll need to connect you with our support team. You can reach us at [support email] or via the contact form at [URL]."
4. Never make up policies, delivery dates, or product specifications.
5. Always be polite and empathetic — customers contacting support are often frustrated.
6. Keep answers concise and clear. Use bullet points for multi-step instructions.
7. If a customer is very upset or mentions a legal issue, immediately escalate: "I'd like to have a team member assist you personally. Please email [support email] with your order number and we'll prioritize your case."Your knowledge base contains: return policy, shipping information, product FAQs, and account management help.
Step 3: Set Up the Escalation Path
Every support chatbot needs a clear “I can’t help you” exit. Configure in your chatbot settings:
- Escalation message: A consistent message the AI uses when it can’t resolve the issue
- Contact options to include: Email address, contact form URL, live chat link (if available), phone number
- Business hours (optional): “Our team is available Monday–Friday 9am–6pm EST”
Add these to your system prompt so the AI always escalates consistently rather than making up a resolution.
Step 4: Configure Display Rules
A support chatbot should appear where support is needed. Go to the Display tab:
- Show on: Order confirmation pages, Account pages, Product pages, Contact page
- Hide on: Homepage (use a sales-focused bot there instead), Blog posts
- Or show everywhere — simpler, and many customers expect chatbot access site-wide
Step 5: Test the Full Support Flow
Before going live, test these five scenarios from an incognito window:
- “What is your return policy?” → should give your actual return policy from training data
- “Where is my order?” → should ask for order number
- “I want a refund” → should explain the refund process from your training data
- “Do you ship to Australia?” → should answer correctly if in training data, escalate if not
- “This is ridiculous, I’ve been waiting 3 weeks” → should respond with empathy and escalation
Common Issues
Problem: “Chatbot makes up return policies that don’t exist.”
Fix: Add to your system prompt: “Never state a policy that is not in your knowledge base. If unsure, escalate to the support team.” Also check your return policy document is in the training data and re-generate embeddings.
Problem: “Customers complain the chatbot doesn’t actually help, just redirects to email.”
Fix: Your knowledge base has gaps. Review your analytics for the top unanswered questions and add those topics to your training data. The chatbot can only answer what it’s been trained on.
What’s Next
- 📊 Track resolution rate and top questions: Chatbot Analytics
- 🧠 Keep knowledge current: Update Your Embeddings when policies change
- 🏥 See it in action — dental clinic: Use Case: Dental Clinic with AI Chatbot
- 💻 See it in action — SaaS company: Use Case: SaaS Support Automation
- 🔔 Get notified of escalations via Slack or Telegram: Workflow Notifications guide
Last verified: AIWU v.4.9.2 · Updated: 2026-02-25