Seven chatbot mistakes small businesses make

A chatbot can take a lot of repeat questions off a small business: opening hours, prices, whether you have a spot on Saturday. It can also quietly annoy customers for months if it is set up carelessly, and the owner often never finds out, because unhappy people just leave the chat.

Most of the problems are not technical. They come from the same seven mistakes, and all of them are fixable before launch or in an afternoon after it. This guide goes through each one, then ends with a short checklist you can use to test any chatbot, whether you built it, bought it or are still comparing options.

It is written for Australian small businesses: cafés, salons, studios, clinics, trades and shops.

1. Launching without a clear job

"Answer customer questions" is not a job. A chatbot with no defined purpose tries to do everything and does none of it well.

Pick one or two jobs and write them down in a sentence each:

• Answer the ten questions we get most often.
• Collect a name, number and what the person needs when we are closed.
• Point people to the right booking page.

If you cannot say what a good conversation looks like, you cannot tell whether the bot is working. Start narrow. You can add jobs once the first one is solid. Our guide to what an AI chatbot costs a small business in Australia also helps you decide how much to ask of one.

2. Letting it guess

The most damaging chatbot answer is a confident wrong one: the wrong closing time, a price you stopped charging, a service you do not offer.

This happens when a chatbot is allowed to fill gaps from general knowledge instead of from your information. The fix is simple to state and easy to skip:

• Give it approved information. Hours, prices, locations, policies, how booking works.
• Tell it what it does not know. Anything outside that information should get "I'm not sure, let me pass you to the team", not an invention.
• Keep the source of truth in one place. If your hours change for a public holiday, the bot should not be the last place to find out. Our guide to keeping your name, address and phone consistent covers the same habit for every listing.

Ask it about something you do not do. If it says yes, it is guessing.

3. No way to reach a person

Customers forgive a bot that cannot answer. They do not forgive one that cannot hand over.

Every chatbot needs an exit that works:

• A visible way to ask for a human, in plain words, not hidden in a menu.
• A handoff that passes along the conversation, so the customer does not repeat themselves.
• A real next step when nobody is in: a callback, a form, or a clear "we reply by 9am".

If your team cannot take a handoff quickly, say so honestly inside the chat. "We reply within one business day" is better than a promise nobody keeps.

4. Pretending to be a person

Giving a chatbot a human name and a friendly tone is fine. Making customers believe they are talking to a staff member is not, and it backfires the moment someone asks a question the bot cannot handle.

The Office of the Australian Information Commissioner's guidance on privacy and the use of commercially available AI products recommends that public-facing tools such as chatbots are clearly identified as AI, with privacy policies updated to match. Being upfront also helps with trust: people are usually happy to use a quick automated answer when they know what it is and can reach a person when they need one.

5. Collecting personal details without thinking

A chat window invites people to type things: full names, addresses, health details, even card numbers. A small business that has not planned for this can end up holding information it never meant to collect.

• Ask only for what the job needs. For a booking enquiry that is usually a name, a contact method and the time wanted.
• Never take payment or ID details in chat. Send people to a secure payment or booking page.
• Say what happens to the information. Link your privacy policy next to the chat, and make sure it covers the chatbot.
• Know where chats are stored and who can read them.

The same OAIC guidance recommends that organisations avoid entering personal information into publicly available generative AI tools, and asks whether using personal information with an AI product is necessary in the first place. Whether the Privacy Act applies to your business depends on its size and activities, so check with your adviser. How Realtiq handles data on this website is in our privacy policy.

6. Letting it make promises you cannot keep

A chatbot that says "yes, we can fit you in tonight" or "that will definitely be refunded" is making statements on your behalf. If they are wrong, the customer will hold you to them.

The ACCC explains in its guidance on false or misleading claims that businesses must not mislead customers about what they offer. Chat messages are part of how you present your business, like a sign or a social post. Check with your adviser how this applies to what your assistant says.

In practice:

• Keep quotes, refunds, warranties and availability for a person to confirm.
• Have the bot say "the team will confirm" instead of "yes".
• Review any wording about prices against your real price list.

7. Never reading the conversations

A chatbot is not finished when it goes live. The first weeks are when you learn what customers actually ask, which is often different from what you expected.

Set a simple routine:

• Week one: read every conversation, daily if you can.
• Weeks two to four: read a handful each week, and all of the ones that ended in a handoff or a complaint.
• Monthly after that: note the top five questions and any wrong answers, then update the information the bot uses.

Repeated questions are also a hint for your website. If ten people ask where to park, the answer belongs on your Google Business Profile and website, not only in a chat.

A quick checklist to test any chatbot

Run these before launch, and again after any big change:

If it fails more than one, fix those before it faces customers.

Realtiq's chat assistant answers on your website, WhatsApp, Instagram, Facebook, email and SMS from information you approve, and passes anything it cannot answer to you. The details are on the AI chatbot page.

Common questions

What is the most common chatbot mistake?

Letting the chatbot answer from guesses instead of from approved information. A wrong price or wrong opening time costs more trust than no answer at all. Give it your real details and a clear instruction to hand over when it is unsure.

How do I know if my chatbot is working?

Read the conversations. Check that the answers are right, that people who asked for a person got one, and that the questions it handled are the ones you wanted it to handle. A bot that is busy is not necessarily a bot that is helping.

Should a chatbot say it is a bot?

Yes. The OAIC recommends that public-facing chatbots are clearly identified as AI. It also avoids the awkward moment when a customer realises they were never talking to a person.

Can a chatbot take bookings or payments?

It can send people to your booking page or take a simple enquiry. Payment details and other sensitive information should go through a secure payment or booking page, not the chat window. If you want to know what happens on Instagram first, see replying to Instagram DMs.

Do I need a chatbot at all?

Not always. If you get a handful of enquiries a week, a clear website and a good Google Business Profile may be enough. A chatbot starts to pay off when the same questions arrive often, or when enquiries come in while you are closed.