Chatbot Demo of Your Own Business in 15 Minutes: Turn Your FAQs Into a Live Preview and Test It With Your Team (Free Demo Builder)

Chatbot demo videos on vendor websites always look perfect, because they were made for a business that isn’t yours. A far more useful demo is one that answers your own customers’ questions, in your own words, so you can see where it helps and where it gets stuck before you spend a rupee or an evening on setup. This guide shows you how to build that preview from a handful of FAQs, how to run a short test session with your staff or business partner, and how to read the results. A free tool on this page turns your questions and answers into a working preview you can type into.

👤 What the visitor seesYour own preview
🤖 Demo botHi! Welcome to Green Leaf Plants. Ask me anything, or type “talk to a person”.
cod available?
🤖 Demo bot · from “Is cash on delivery available?”Yes, cash on delivery is available on orders up to ₹3,000.
where is my order 48213
🤖 Demo botI’ll pass this to our team so someone can check it for you.

Illustrative shop. Your demo uses your own answers.

⚡ In short

  • A vendor’s demo shows what a bot can do for someone else. Your own preview shows what it will do for you.
  • Write 5 to 10 real questions with short answers, paste them into the free builder, and type into it as a customer would.
  • Let two or three people who know your customers try to break it for 10 minutes. Note answers, handovers and misses.
  • Fix the misses, then move the same answers into a real bot on a test page before it goes on your whole site.
Chatbot demo in LiveChatBot: AI Agent knowledge sources and a sample chat answered from them
How it looks in LiveChatBot: AI Agent knowledge sources and a sample chat answered from them.

The problem

Why the demo on a vendor’s website tells you so little

Every chatbot company has a demo. Usually it’s a short video or a sample bot on their home page, answering questions about the vendor’s own product or about a made-up shoe store. It looks smooth because the questions were chosen to match the answers. Nobody types “cod hai kya?” or “my parcel is 4 days late, what now” into a demo that was scripted in advance.

That doesn’t mean the vendor is hiding anything. A generic demo simply can’t answer the question you actually have: will this work for my customers, with my questions, in the way they write? To learn that, you need to see a bot answer your own material.

There’s a second reason to build your own preview. Most of the work in a chatbot isn’t the software; it’s the answers. If you can’t write clear, short answers to your top questions, no tool will fix that. A quick preview exposes this early. You’ll find out in fifteen minutes that your returns answer is too long, that you have three different delivery answers that contradict each other, or that half your customers ask about something you never thought to write down.

Finally, a chatbot demo built on your own answers is something you can show. Business partners, a manager or a sceptical senior salesperson often say no to a chatbot because they imagine the clumsy bots they’ve met elsewhere. Putting your own answers in front of them, in a chat they can type into, turns an argument into a test. If they can break it, you learn something. If they can’t, you’ve made your case.

If you’re still at the stage of comparing vendors, our list of 12 written questions to send before any sales call pairs well with this guide. Use the vendor’s answers to shortlist, and your own preview to decide.

The method

Fifteen minutes from FAQ list to tested preview

Minutes 0–5Write

List 5 to 10 questions customers ask most, each with a one- to three-sentence answer.

Minutes 5–6Paste

Put them into the free builder below as Question | Answer lines.

Minutes 6–10Play

Type 10 messages the way real customers write. Include one order-specific question.

Minutes 10–15Read

Check the scorecard: what was answered, handed over or missed, and why.

LaterShow

Run a short session with your team, then move the fixed answers into a real bot.

Step 1

Write the questions and answers for your chatbot demo

Open a notes app and write down the questions you answer most often. Don’t brainstorm; remember. Think about the last week of WhatsApp messages, calls and walk-in customers. For most small businesses the list looks something like this: delivery areas and times, charges, cash on delivery or payment options, opening hours, returns or cancellations, prices of popular items, and one or two questions specific to your trade, like “do you do home visits?” for a clinic or “is there a demo class?” for a coaching centre.

Five lines is enough to start. Ten is better. More than fifteen is too many for a first preview; you’ll spend your time writing instead of testing.

For each question, write the answer you’d give a customer in a hurry:

  • Key fact first. “Yes, cash on delivery is available on orders up to ₹3,000” beats “We offer various payment options including…”.
  • One to three sentences. If it needs a paragraph, it probably needs two separate questions.
  • Real numbers. Days, rupees, times, areas. Vague answers feel evasive in a chat.
  • Your own tone. If you’d say “we”, say “we”. If you use Hinglish with customers, it’s fine to use it here.

Write the question the way a customer would ask it, not as a heading. “Do you deliver to my city?” is better than “Delivery information”, because the preview matches words in the visitor’s message against words in your question.

Leave out anything that depends on one customer’s details: where a specific order is, why a payment failed, a complaint about a delivery. Those aren’t answers; they’re jobs for a person. The preview will show you how it hands them over.

If you want help choosing which questions belong on the list, our guide to building a bot from questions you already get has a grouping tool that ranks topics by how often they come up.

Free tool

Build a working chatbot demo from your own answers

Add your business name and your Question | Answer lines, then start the demo and type into it. Under every reply it explains why the bot said what it said, and a scorecard tracks answers, handovers and misses. Load the example to see a plant shop’s preview with ten test messages already played.

1. Your business name

2. Your questions and answers, one per line, written as Question | Answer. Five to ten lines is plenty.

3. Optional: messages to play automatically, one per line. Use the way real customers write, typos and Hinglish included.

Runs only in your browser; nothing is sent or saved. This is a simple word-matching preview, not the real LiveChatBot AI Agent.

Step 2

Play the customer, not the owner

When you type into your own preview, you'll be tempted to ask your questions exactly as you wrote them. Don't. You already know the answers; you're testing whether a customer who doesn't would get them.

Use this mix of ten messages, which covers most of what goes wrong in real chats:

  1. A greeting on its own. "hi" or "namaste". Many visitors start this way and wait. The bot should say what it can help with.
  2. A short, lazy version of a top question. "cod?" or "delivery pune". Real visitors often type two words.
  3. A Hinglish or misspelt version. "shipping charges kitna hai", "deliverey time". This shows whether your answer wording is too narrow.
  4. A question about a nearby topic you didn't write down. "do you sell bonsai" or "can I pay by UPI". The bot should admit it doesn't know, not guess.
  5. A specific order question. "where is my order 48213". The bot should hand over, not answer with your general delivery time.
  6. A request for a person. "talk to a human". It must hand over straight away.
  7. Contact details. Type an email or phone number after a handover, as a visitor would when asked.
  8. Two questions in one message. "do you deliver to Mysore and is COD available". See which half wins.
  9. A complaint. "the plant came damaged". Decide whether your bot should answer with your replacement policy or pass it to a person.
  10. Something rude or off-topic. It happens. A calm "I can help with…" is all you need.

The example in the tool already includes most of these. Watch two in particular: "do you have bonsai" and "can i pay by upi" both come back as misses, because the plant shop never wrote those answers. That's the most valuable thing a preview shows you, and it's why it's worth playing the customer honestly.

Step 3

Read the scorecard and the reasons

Every reply in the preview carries a short "Why" line. Read them; they're more useful than the replies themselves.

Matched line 3 (shared: cod). The bot found your cash-on-delivery question because the visitor's message shared a word with it. If a match was right but only shared one weak word, your question is fragile. Add more of the words customers use to the question line, for example "Is cash on delivery or COD available?".

Handover phrase. The bot spotted words like "my order", "talk to" or a long number and passed the chat to your team. That's what you want for anything only a person can check. If a handover fired on a general question, your trigger words are too broad.

No saved question shares enough words. A miss. Either write an answer for it or accept that a person should handle it. Misses are listed under the scorecard in the order you typed them.

Contact details found. After a handover or miss, the visitor left an email or phone number. In a real LiveChatBot setup, this is what the away form does when nobody is online: it collects an email or phone number so the conversation isn't lost, and offline messages can become tickets your team follows up.

The scorecard adds these up. The percentage counts answers plus correct handovers out of all real questions, leaving out greetings and contact details. Around 80% is a sensible bar before you show the preview to anyone else. Below that, spend five more minutes on the misses.

One honest caveat: this preview matches words. A real AI-based bot understands meaning, so it will cope with more wordings than the preview does. That's a good thing. If your answers work in a simple word-matching preview, they'll usually work even better with an AI reading them. The reverse isn't true: vague or contradictory answers stay vague and contradictory, however clever the bot.

Example

What the plant shop changed after one round

“can i pay by upi” came back as a miss
🚩 Instead ofLeave it and hope few people ask.
✅ Do thisAdd a line: “Which payment methods do you accept? | UPI, cards, net banking and cash on delivery up to ₹3,000.” (Illustrative.)
“do you have bonsai” came back as a miss
🚩 Instead ofWrite a long answer listing every plant in stock.
✅ Do thisHand it to a person: stock changes daily, so the bot says it will check with the team.
“are you open on sunday” matched the timings line
🚩 Instead ofAssume it's fine because a match happened.
✅ Do thisRead the answer aloud: it already says “Closed on Sunday”, so the visitor gets a direct reply.

Step 4

Run a ten-minute session with your team

A preview you've only tested yourself is half-tested. You wrote the answers, so you unconsciously ask questions that fit them. The fastest fix is to hand the screen to someone else.

Pick two or three people who talk to customers: the person on the shop counter, whoever answers WhatsApp, a salesperson, your business partner. Give each of them ten minutes and one instruction: "Ask it what customers really ask, and try to make it give a wrong answer."

While they type, sit beside them and keep quiet. Write down three things on paper:

  • Every miss, word for word. These become your next answers.
  • Every wrong answer, where the bot matched the wrong line. These are more serious than misses, because a confident wrong reply misleads a customer.
  • Every time they said "a customer would never ask it like that" or "they'd ask it like this instead". That's free training data.

After the session, ask each person one question: "Would you be happy for a customer to get these answers when we're closed?" If the answer is no, ask what would change their mind. It's usually one or two specific answers, not the idea of a bot.

This session does a second job too. The people who'll work alongside the bot have now seen exactly what it says and when it hands chats to them. They're far less likely to resist it later, and far more likely to suggest new answers once it's live.

For a more formal pass/fail test before launch, our golden-question regression test gives you 20 questions to re-check after every change.

Compare

Vendor demo, free trial and your own preview

Vendor demoFree trial or free planYour own preview (this page)
Whose questionsThe vendor'sYours, once set upYours
Time to see resultsMinutesAn hour or more of setup15 minutes
Shows real handoverSometimesYes, to your real inboxSimulated, with reasons
Understands meaningYes, if the product has AIYes, if the product has AINo, word matching only
Best used forSeeing what's possibleTesting on real visitorsChecking your answers and getting buy-in

Step 5

From preview to a real bot on a test page

Once your preview scores well and your team is comfortable, the answers you wrote are the main asset. Moving them into a real bot is mostly copying and pasting.

In LiveChatBot, general answers go into the AI Agent's knowledge; see how to train your AI chatbot. Save the same answers as saved replies too, so your team can send them in one click and the free keyword bot can answer from them if AI replies are unavailable or used up. Our guide to saved replies shows how to set those up, including the {name} placeholder.

Don't switch it on across your whole site on day one. Use "where it shows" to limit the widget to one page first, or hide it on pages you're not ready for; showing live chat on specific pages walks through it. Visit that page yourself from your phone and repeat your ten test messages, this time against the real bot. Then ask your team to do the same.

Set your opening hours before real visitors arrive. Outside those hours the away form collects an email or phone number, which is the real version of the "contact details" step you tested in the preview.

When the test page behaves, widen the rollout. Our staged soft-launch plan covers which pages to add next and what to watch in the first week. After launch, the misses you noted in your preview session are just the start: every week, read the chats the bot couldn't handle and add answers for the ones that repeat.

Checklist

Before you show your preview to anyone

Preview progress

Getting started

Build your preview tonight, show it tomorrow

Write down your top five customer questions with short answers, paste them into the builder above and type ten messages as a customer would. Fix the misses, then hand the screen to a colleague tomorrow morning.

When you're ready for the real thing, LiveChatBot's free plan includes an AI Agent with a small free monthly allowance, a free keyword bot that answers from saved replies, a team inbox, the Android app, Flows, opening hours with an away form and tickets, for two websites with 100 visitors per website each month and no per-seat charges. See pricing and start at livechatbot.in.

Frequently asked questions

How long does a chatbot demo of my own business take to build?

About 15 minutes with the free builder on this page: five minutes to write 5 to 10 questions and answers, a minute to paste them, and the rest to type test messages and read the scorecard.

Is the preview the same as the real LiveChatBot AI Agent?

No. The preview matches words between the visitor's message and your questions. The real AI Agent answers from the knowledge you give it and copes with more wordings. The preview is for checking your answers, not the AI.

Who should test the preview?

Two or three people who talk to customers every day, such as counter staff, whoever answers WhatsApp, or a salesperson. Ask them to type what customers really ask and try to get a wrong answer.

What score is good enough?

Aim for about 80% of real questions answered or handed over correctly, with no wrong answers. Write answers for the misses, or decide that a person should handle them.

Does the demo builder save or send my answers?

No. It runs entirely in your browser and nothing is sent or saved. Refreshing the page clears it, so keep your Question | Answer lines in a notes file.

Keep reading

Related guides

📘

Chatbot Conversation Storyboards: Write Your Top Question's Chat Five Ways Before You Build Anything (Free Storyboard Builder)

Read the guide →

🧭

Chatbot Design Blueprint: Sketch Your Bot's Menu Tree on Paper and Catch Problems Before You Build

Read the guide →

🛠️

Chatbot Companies: Send These 12 Questions in Writing Before Any Demo, Then Grade the Reply (Free Vendor Reply Checker)

Read the guide →

Turn your preview into a real bot

Paste the answers from your preview into the AI Agent and saved replies, test on one page, then go live. Start free on livechatbot.in.