# How to Set Up No Match Reply With AI in ZazzyAgent A customer will eventually type something your predefined chatbot doesn't recognize. Maybe your bot has a pricing flow, but the customer asks: > Do you offer discounts for annual plans? No keyword matches. Without a fallback, the customer may receive nothing useful. ZazzyAgent provides **No Match Reply** for this situation. ## What is No Match Reply? No Match Reply is the response used when a customer's message doesn't match an existing predefined bot flow or trigger. Instead of leaving the customer without an answer, you can configure a response. That response can also use AI. The current AI training workflow allows No Match Reply to use an AI Reply connected to the relevant trained knowledge. ## Why does No Match matter? Without it: ```text Customer ↓ Unknown message ↓ No match ↓ Nothing useful ``` With it: ```text Customer ↓ Unknown message ↓ No Match ↓ AI Reply ↓ Useful answer ``` ## No Match vs normal flow A normal flow handles something you've deliberately configured. For example: > `pricing` → Pricing flow. No Match handles something that doesn't match. For example: > Do you have a discount for startups? ## Why use AI for No Match? You don't know every way a customer will phrase a question. A customer may ask: > How much? Another: > What's the price? Another: > Can you tell me the cost? A keyword-based system might require several triggers. An AI fallback can understand the question more naturally. ## Before you configure it You need: * AI knowledge/training * Useful business information * A configured AI Reply * A connected bot/channel * A test account ## Step 1: Open the No Match configuration Open the relevant bot settings in ZazzyAgent. Find the **No Match** action/configuration. This is the system-level fallback for unmatched customer messages. ## Step 2: Open its Flow The No Match action can be configured through Flow Builder. Open the flow associated with No Match. ## Step 3: Add AI Reply Add or open the **AI Reply** component. Connect it to the No Match path. The current reference workflow specifically configures No Match by opening its flow and assigning the AI training campaign to the AI Reply component. ## Step 4: Select the appropriate AI knowledge Choose the AI training/knowledge configuration that should answer the customer's questions. For example: > Customer Support Knowledge rather than a sales-only knowledge source. ## Step 5: Enable No Match Return to the relevant bot configuration and enable No Match Reply. Save the settings. ## What happens now? Suppose your existing bot recognizes: > pricing > booking > support but a customer asks: > Do you offer a free trial? If that isn't a predefined flow trigger, the No Match Reply can hand the question to AI. ## No Match doesn't mean "AI knows everything" This is important. AI should answer using the Knowledge you provide. If your Knowledge says nothing about free trials, the agent shouldn't invent one. ## Tell AI what to do when information is missing A useful instruction is: > If the answer isn't available in Knowledge, don't guess. Tell the customer you don't have the information and offer human assistance. ## No Match and AI Agent are different approaches ### No Match + AI Reply AI acts when predefined automation doesn't match. ### AI Agent for All Queries AI handles the conversation broadly as the primary assistant. Neither is automatically right for every business. ## When No Match is useful No Match is particularly useful for: * FAQ bots * Support bots * Structured menus * Lead-generation bots * Booking systems * Businesses with many predefined flows ## Example: Support bot Configured: **Track Order** **Returns** **Refunds** **Contact Support** Customer asks: > How long does delivery normally take? No predefined path matches. AI fallback answers from Knowledge. ## Example: Lead-generation bot Configured: **Request Quote** **Book Demo** **Pricing** Customer asks: > Can you work with companies in Dubai? No exact trigger. AI answers from the business Knowledge. ## No Match and human handoff No Match can also be used as a safety net. For example: ```text Unknown Question ↓ AI tries to answer ↓ Information available? ↙ ↘ Yes No ↓ ↓ Answer Human ``` This prevents the AI from making unsupported claims. ## Configure a clear fallback Your fallback shouldn't simply be: > I don't understand. It should still help. For example: > I can help with questions about our services, pricing and bookings. What would you like to know? Or, when AI is connected: > Let me check that for you. then let the AI answer. ## No Match and AI context The current AI configuration supports contextual memory, which can be useful when a customer's unmatched question depends on previous messages. For example: Customer: > Do you have the premium plan? AI: > Yes. Customer: > Is that available for teams? The second question is difficult to understand without the previous context. ## Test No Match deliberately Don't test only your known keywords. Send unusual questions. Examples: > What is your refund policy? > Do you have annual billing? > Are you open Sunday? > Can I speak to a human? > I have a problem with something else. ## Check the difference between "No Match" and "No Answer" These aren't necessarily the same. **No Match** means the predefined trigger/path wasn't matched. The AI may still be able to answer. **No Answer** means the AI itself may not have enough information. Your fallback should account for both situations. ## Common problems ### No Match never fires Check whether No Match is enabled. ### No Match fires but AI doesn't respond Check the AI Reply configuration and selected knowledge. ### AI responds with incorrect information Check your AI training/Knowledge. ### AI replies to everything You may have configured AI for all queries instead of fallback-only. ## When should you use AI for No Match? A strong default for many structured bots is: > **Use predefined flows for important actions, and AI fallback for natural-language questions.** This gives you both control and flexibility. ## Recommended architecture ```text Customer Message ↓ Matches predefined flow? ↙ ↘ Yes No ↓ ↓ Structured AI No Match Flow ↓ ↓ Knowledge available? ↙ ↘ Yes No ↓ ↓ Answer Human ``` This is one of the simplest ways to add AI to an existing chatbot without rebuilding the entire automation.