Skip to main content

Build a WhatsApp AI Support Bot in 15 Minutes

AI & Agentsintermediate15 minutes13 min read

Create an AI-powered WhatsApp support bot that handles 70% of inquiries automatically. Setup in 15 minutes. No coding required.

Your support team answers the same 10 questions 50 times a day. An AI bot handles those in seconds — freeing your team for complex issues that actually need a human. Here's how to build one in 15 minutes.

Unlike generic chatbot builders, MoltFlow's AI learns from YOUR knowledge base and responds in YOUR brand voice — not canned ChatGPT answers. Businesses using AI support bots handle 10x the volume without hiring. Average setup: 15 minutes.

By the end of this guide, you'll have a working WhatsApp chatbot that handles 70% of customer service inquiries automatically, routing only complex issues to your team.

What You'll Need

Before starting, make sure you have:

  • A MoltFlow account — AI auto-replies require the Growth plan or higher (unlock AI-powered automation). The free plan works for testing setup only. Compare plans here.
  • OpenClaw API key — Get your API key from OpenClaw's developer portal. The free tier provides 1,000 messages/month, perfect for testing.
  • Connected WhatsApp session — You'll need an active WhatsApp number connected to MoltFlow. If you haven't done this yet, follow our 2-minute connection guide first.
  • Basic understanding of AI prompts — You don't need to be an expert, but understanding how to write clear instructions for AI models will help you customize your bot's behavior.

Important: For production use, ensure your WhatsApp number is a WhatsApp Business account and that you have permission from your customers to use automated responses in compliance with WhatsApp's Business policies.

Step 1: Prerequisites and OpenClaw Account Setup

OpenClaw is an AI agent platform that provides production-ready language models with built-in safety guardrails, function calling, and memory management. Unlike raw LLM APIs, OpenClaw is designed specifically for building conversational AI agents.

Get your OpenClaw API key:

  1. Visit openclaw.com and create an account
  2. Navigate to the API Keys section in your dashboard
  3. Click "Create New Key" and give it a descriptive name like "MoltFlow Support Bot"
  4. Copy the API key immediately — it won't be shown again
  5. Store it securely (we'll add it to MoltFlow in the next step)

Choose your model:

OpenClaw offers several models optimized for different use cases:

  • openclaw-chat-v2 — Best for general support conversations, balanced speed and quality
  • openclaw-chat-pro — Higher quality responses, slightly slower, recommended for complex products
  • openclaw-chat-fast — Ultra-fast responses, good for simple FAQ-style support

For this guide, we'll use openclaw-chat-v2 as it provides the best balance for most support scenarios.

Step 2: Connect Your WhatsApp Session to MoltFlow

If you haven't already connected your WhatsApp account to MoltFlow, follow our quick connection guide. This takes about 2 minutes.

Once connected, verify your session is in WORKING state:

  1. Go to the Sessions page in your MoltFlow dashboard
  2. Check that your session shows a green "WORKING" badge
  3. If it shows "STOPPED" or "FAILED", restart the session using the action menu

Your bot will only respond to messages when the session is in WORKING state, so this step is critical before proceeding.

Step 3: Configure OpenClaw as Your AI Provider

Now we'll connect OpenClaw to MoltFlow so your bot can use AI to generate intelligent responses.

  1. Navigate to AI Configuration in the MoltFlow dashboard left sidebar
  2. Click "Add AI Provider" or edit your existing configuration
  3. Select OpenClaw from the provider dropdown
  4. Paste your OpenClaw API key in the API Key field
  5. Choose openclaw-chat-v2 as your model
  6. Set the system prompt to define your bot's behavior

Example system prompt for a customer support bot:

text
You are a helpful customer support agent for [Your Company Name]. Your role is to:

1. Answer questions about our products and services accurately and professionally
2. Use the knowledge base documents provided to give specific, detailed answers
3. Maintain a friendly, patient tone even with frustrated customers
4. If you don't know the answer or the question requires human judgment, politely say: "I'd be happy to connect you with a member of our support team who can help with that. Would you like me to do that?"

Never make up information. Never promise features or pricing that aren't in the knowledge base. Always prioritize customer satisfaction while staying accurate.

Important customization tips:

  • Replace [Your Company Name] with your actual company name
  • Add specific policies or guidelines your support team follows
  • Include your brand voice (formal vs casual, emojis vs professional, etc.)
  • Specify any topics the bot should never handle (refunds, account deletion, legal issues, etc.)
  1. Click Save Configuration

Your AI provider is now connected and ready to generate responses.

Step 4: Define Your Bot's Knowledge Base and Personality

The quality of your automated WhatsApp replies depends heavily on the knowledge base you provide. MoltFlow uses RAG (Retrieval-Augmented Generation) to let your AI customer service bot search through your documentation and pull relevant information into responses.

Knowledge base RAG requires the Business plan — your AI knows your product as well as your best support agent. Without RAG, the bot relies on the system prompt only. Learn more about building a knowledge base for AI.

Upload your knowledge base documents:

  1. Go to AI Configuration → Knowledge Base section
  2. Click Upload Documents
  3. Upload your support docs in these formats:
    • PDF files (product manuals, FAQ documents)
    • Markdown files (help articles, policy pages)
    • Text files (troubleshooting guides, scripts)
  4. MoltFlow will automatically chunk and index the content

What to include in your knowledge base:

  • Product documentation — Features, how-to guides, technical specs
  • FAQ responses — Common questions your team answers repeatedly
  • Troubleshooting guides — Step-by-step solutions for known issues
  • Policy documents — Return policies, warranty info, terms of service
  • Pricing information — Plans, add-ons, discounts (if you want the bot to discuss pricing)

Important: Review all documents before uploading. The bot will use this information to answer questions, so ensure everything is accurate and up-to-date. Remove any internal-only information that customers shouldn't see.

Configure your bot's personality using style profiles:

MoltFlow's style profiles let you train the bot to match your brand voice. This ensures your WhatsApp chatbot sounds like your team, not a generic AI:

  1. Go to Style Profiles in the dashboard
  2. Create a new profile called "Support Bot Voice"
  3. Paste 5-10 example messages written in your desired tone
  4. Enable the profile for your AI configuration

Learn how to train your AI's communication style in our style training guide.

For example, if your brand is casual and friendly, your samples might include:

  • "Hey there! Happy to help with that 😊"
  • "No worries — this happens sometimes. Here's what you can do..."
  • "Great question! Let me break it down for you..."

If your brand is more formal:

  • "Thank you for contacting us. I'm happy to assist."
  • "I understand your concern. Here's the information you need..."
  • "To resolve this issue, please follow these steps..."

The AI will learn from these examples and adapt its response style accordingly.

Step 5: Set Up Automatic Reply Rules

Now we'll configure when the bot should respond automatically and when to leave messages for human review.

Configure auto-reply triggers:

  1. Navigate to AI Configuration → Auto-Reply Rules
  2. Enable Automatic AI Responses
  3. Choose your trigger mode:

Option A: Reply to all inbound messages (recommended for dedicated support lines)

  • Best for: WhatsApp numbers used exclusively for customer support
  • Responds immediately to every message
  • Ensures customers never wait for basic questions

Option B: Reply only to specific keywords (recommended for mixed-use numbers)

  • Best for: Numbers that handle both automated and personal conversations
  • Only triggers bot when message contains keywords like "help", "support", "question"
  • More control, but customers need to know the keywords

Option C: Reply to messages in monitored groups (for community support)

  • Best for: WhatsApp groups where you provide product support
  • Bot monitors group messages and responds when it detects questions
  • Can be configured to only respond when mentioned

Learn how to set up group monitoring in our group monitoring guide.

Set business hours (optional but recommended):

Configure when the bot should be active:

  1. Enable Business Hours toggle
  2. Set your timezone
  3. Define operating hours (e.g., Monday-Friday, 9 AM - 6 PM)
  4. Set an out-of-hours message: "Thanks for your message! Our support team is available Monday-Friday, 9 AM - 6 PM EST. We'll respond during our next business day."

Outside business hours, the bot will send this message instead of trying to answer questions.

Configure escalation triggers:

Tell the bot when to hand off to a human:

  1. Enable Smart Escalation
  2. Set escalation keywords (e.g., "speak to human", "manager", "not helping", "frustrated")
  3. Configure escalation message: "I'll connect you with a team member who can better assist. Someone will respond within 2 hours during business hours."
  4. Set up notifications so your team knows when escalations happen

Important: Always provide a way for customers to reach a human. This builds trust and prevents frustration when the bot can't help.

Step 6: Test the Bot with Sample Conversations

Before going live, thoroughly test your bot to ensure it responds correctly.

Send test messages from another phone:

  1. Save your MoltFlow-connected WhatsApp number on a different phone
  2. Send a simple question: "What are your business hours?"
  3. Verify the bot responds within a few seconds
  4. Check that the response is accurate and in the right tone

Test edge cases:

  • Unknown questions: "Do you sell purple elephants?" — The bot should admit it doesn't know or redirect to human support
  • Multi-language: Send messages in different languages if you serve international customers — check if responses are appropriate
  • Follow-up context: Send a question, then a follow-up — the bot should remember the conversation context
  • Escalation triggers: Send "I want to speak to a manager" — verify the escalation workflow activates
  • Outside business hours: Test a message when your bot is set to off-hours — confirm the out-of-hours message appears

Review the message log:

After each test:

  1. Go to the Messages page in MoltFlow
  2. Find the conversation in the list
  3. Click to view the full thread
  4. Verify the bot's responses are logged correctly
  5. Check that metadata shows it was AI-generated

Important: Test with real questions your customers ask. Go through your support ticket history and pull out common inquiries. If the bot can't answer them correctly, improve your knowledge base or system prompt.

Step 7: Monitor and Refine Bot Responses

Launch your bot and monitor its performance over the first week. This is where you'll identify gaps and improve accuracy.

Use the Messages page to review conversations:

  1. Navigate to Messages in the dashboard
  2. Filter by AI-generated to see all bot responses
  3. Review conversations daily for the first week
  4. Look for patterns in questions the bot struggles with

Identify knowledge gaps:

When you see the bot give weak answers or escalate unnecessarily, it usually means:

  • Missing documentation — Add the missing info to your knowledge base
  • Unclear system prompt — Refine instructions for how to handle that type of question
  • Need for human judgment — Add escalation rules for those topics

Refine your system prompt based on real interactions:

If you notice the bot is:

  • Too verbose — Add to system prompt: "Keep responses under 100 words when possible."
  • Too formal — Add style profile examples with a friendlier tone
  • Missing details — Add: "Always provide specific steps, not general advice."
  • Making assumptions — Add: "If you're not sure what the customer means, ask a clarifying question."

Track key metrics:

Monitor these in your MoltFlow dashboard:

  • Auto-reply rate — How many messages the bot handles vs. escalates
  • Average response time — Should be under 5 seconds for AI responses
  • Escalation rate — If over 30%, your knowledge base may need expansion
  • Customer satisfaction — Send follow-up surveys asking "Was this helpful?"

Aim for an escalation rate of 10-20% for mature bots. Higher rates mean you need better documentation or narrower auto-reply triggers.

Tips for Production

Rate limiting considerations:

OpenClaw free tier provides 1,000 messages/month. For production:

  • Upgrade to OpenClaw Pro (10,000 messages/month) or Enterprise (unlimited)
  • Monitor your usage in the OpenClaw dashboard
  • Set up billing alerts to avoid service interruptions

Human escalation paths:

Always have a clear escalation process:

  • Use MoltFlow's webhook notifications to alert your team of escalations in real-time
  • Consider routing escalated conversations to a specific WhatsApp group where your support team monitors
  • Set up a schedule so someone is always on-call to handle escalations during business hours

Learn how to configure webhooks in our webhook setup guide.

Monitoring response quality:

  • Review a random sample of 20 AI conversations per week
  • Create a checklist: accuracy, tone, helpfulness, escalation appropriateness
  • Keep a "bad response" log and use it to refine your system prompt
  • When you fix an issue, re-test with the original question to confirm improvement

Handling off-hours:

Consider these strategies for 24/7 coverage:

  • Use the bot for basic questions outside business hours, but clearly state response times for complex issues
  • Set expectations: "Our team will follow up on this during business hours (9 AM - 6 PM EST)"
  • For urgent issues, provide an escalation option even off-hours if you have on-call staff

Compliance and privacy:

Important legal considerations:

  • Add a disclaimer to your first bot message: "This is an automated assistant. For complex issues, you'll be connected to a team member."
  • Follow WhatsApp Business policies on automated messages
  • Ensure your knowledge base doesn't include customer data or PII
  • Review your jurisdiction's requirements for chatbot disclosure

What's Next

Now that you have a working WhatsApp chatbot handling customer inquiries automatically, explore these advanced capabilities:

Ready to scale your customer service? Start your Business plan trial to unlock AI-powered automation with knowledge base RAG. View pricing.

Questions? Join our community Discord or check out the MoltFlow documentation for more advanced AI agent patterns.

Ready to automate your WhatsApp?

Start for free — set up in under 2 minutes.

Get Started Freeor explore more guides →