Pick a depth. Each prompt opens in your AI pre-loaded with the lesson. Click a row to preview the prompt.
The gap between a hobby project and something a business would actually put on its website is two short lines of writing. A cold chat box that opens silent and blank makes visitors bounce; a warm, on-brand greeting that names the business and invites a question makes them engage. Even more important is the fallback — what the bot says when it doesn't know. A bot that confidently invents a wrong answer to "do you offer refunds?" can genuinely damage a business, while a bot that says "I'm not sure about that one — let me get you to a person" protects the client and builds trust. This is your first taste of conversation design, and it's the moment your bot stops sounding like a raw language model and starts sounding like it works for the business.
In Chatbase these are two settings: the "Initial message" (the greeting that appears when the chat opens) and the fallback behaviour, shaped by the bot's instructions/system prompt. Here is exactly what I wrote for Flow Room. Greeting: "Hi! Welcome to Flow Room. I can help with class timings, passes, beginner classes, our location, and freezing your membership. What would you like to know?" Notice it does three jobs in one sentence — it names the business, sets expectations by listing what it can help with (so nobody asks it to book a flight), and ends with an open question that invites a reply.
For the fallback, I added this line to the bot's instructions: "If you don't know the answer or the question is outside class info, passes, location, or membership, do NOT guess. Say: 'That's a great question — I'm not certain on that one. Please WhatsApp us at 98xxxxxxxx or drop your email and we'll get right back to you.'" Then I tested it by asking something off-topic: "do you sell yoga mats?" — a fact I never trained. Before the fallback line, the bot cheerfully invented "Yes, we have mats for 800 rupees" (completely made up, and dangerous). After adding the fallback instruction and retraining, the same question got the honest handoff. That single edit is the difference between a bot a business can trust and one that will eventually embarrass them. Rule of thumb: your greeting sets the menu, and your fallback admits the edges of that menu gracefully.