CentriWebAn AI chatbot helps a small business website when it answers common questions from your own content, works after hours, and captures the enquiry so a person can follow up. It hurts when it guesses at prices, availability or advice, traps people who want a human, or replaces your contact form entirely. You are responsible for what it tells customers, so it needs tight limits, a clear handover to a person and proper testing before it goes live.
This guide gives you a fit test, what to train it on, how to hand over to a person, the legal risks and a testing checklist.
The short answer
A chatbot is worth it if people regularly ask the same questions outside business hours and you can answer those questions from information you already publish. It isn’t worth it if most enquiries need a quote, a site visit or a judgement call, or if nobody will check the conversations it has. Either way, keep your contact form and phone number. A chatbot is an extra door into your business, not a replacement for the front one.
Good fits and poor fits
| Usually a good fit | Usually a poor fit |
|---|---|
| Opening hours, locations, service areas, parking | Quoting prices for custom work |
| What a service includes and how it works | Medical, legal, financial or safety advice |
| How to book, reschedule or prepare | Complaints and disputes |
| Which service suits a common situation | Anything where a wrong answer costs the customer money |
| Collecting details for a callback after hours | Businesses with very few enquiries to begin with |
A quick way to decide: look at your last 50 enquiries by email, form and phone. If a large share are the same handful of questions with answers already on your website, a chatbot can take that load. If most are unique, a faster human reply will do more good. Our guide on how fast to reply to an enquiry covers that side.
What to train it on (and what to keep out)
Most modern chatbots answer from a set of documents you give them. What you include decides what it says.
Put in
- your service pages, FAQs and booking information
- your service area, hours and contact details
- policies customers ask about: cancellations, warranties, payment terms
- a short list of approved answers to your ten most common questions
Keep out
- internal price lists, margins and supplier terms
- old pages, superseded policies and draft documents
- anything with customer or staff personal information
- promises you can’t keep every time, like “same-day service”
Set clear instructions as well. It should only answer from your content, say “I’m not sure, let me get someone to help” when it doesn’t know, never quote a price unless the price is fixed and published, and never give advice outside your published information.
Handing over to a person
The quickest way to annoy a customer is a bot that won’t let them reach a human. Build the handover in from day one.
- Always offer a way out. “Talk to a person” should be one tap away, at any point.
- Hand over automatically when the customer asks for a person, gets frustrated, mentions a complaint, or asks something outside the bot’s content.
- During business hours, pass the chat to a staff member live, or offer a callback within a stated time.
- After hours, collect name, contact details and the question, and tell them when they’ll hear back. Then make sure they do.
- Pass the context along. Whoever follows up should see the whole conversation, so the customer doesn’t repeat themselves.
Capturing the enquiry, not just answering
A chatbot that answers a question and lets the visitor leave has done half the job. The value is in turning conversations into enquiries you can follow up.
- Ask for name and phone or email once the conversation shows real interest, not as a gate before any answer.
- Ask two or three qualifying questions where they fit naturally, like suburb, type of job and timing.
- Send every captured enquiry straight into your CRM or inbox, tagged as a chatbot lead.
- Trigger the same follow-up you’d use for a form enquiry, including a prompt reply from a person.
- Review a sample of conversations each week to spot questions the bot couldn’t answer. Those are gaps in your website content too.
An AI receptionist does a similar job on the phone, and the same rules about handover and follow-up apply.
The risks: wrong answers, privacy and consumer law
This section is general information, not legal advice.
You’re responsible for what it says
The best-known case is Moffatt v. Air Canada, 2024 BCCRT 149, decided by British Columbia’s Civil Resolution Tribunal in February 2024. Air Canada’s website chatbot told a customer he could apply for a bereavement fare discount after travelling, which wasn’t the airline’s policy. Air Canada argued the chatbot was responsible for its own actions. The tribunal rejected that, found the airline liable for negligent misrepresentation because it hadn’t taken reasonable care to make sure its chatbot was accurate, and ordered it to pay the fare difference plus interest and fees. It’s a Canadian small claims decision, not Australian law, but the principle is easy to follow: your chatbot speaks for your business.
Australian Consumer Law
In Australia, businesses must not engage in misleading or deceptive conduct. The ACCC’s guidance says it makes no difference whether a business intends to mislead or not. A chatbot that gets your prices, warranty terms or consumer guarantee rights wrong can create the same problem as a misleading web page.
Privacy
Chat conversations often include names, phone numbers and details of a customer’s situation. The OAIC’s guidance on commercially available AI products recommends making it clear when people are dealing with an AI tool such as a chatbot, updating your privacy policy to explain how you use AI, and checking how the vendor stores and uses chat data before you sign up.
Testing it before it goes live
Spend an afternoon trying to break it before your customers do. Have a few people, ideally not the person who set it up, work through this list:
- Ask your ten most common questions and check each answer against your website.
- Ask for a price on custom work. It should hand over, not guess.
- Ask something outside your business entirely. It should decline politely.
- Ask for a person. It should hand over straight away.
- Try to get it to promise a discount, a refund or a timeframe.
- Type a complaint and see whether it escalates.
- Leave details after hours and check they reach your CRM or inbox, and that someone follows up.
- If it books appointments, test double bookings, cancellations, public holidays and times outside your hours.
- Test it on a phone, not just a desktop.
Keep testing after launch. Read a sample of conversations every week for the first month, then monthly, and update its content whenever your services, prices or policies change.
Where to start
- Sort your last 50 enquiries and count how many are repeat questions answered on your site.
- Tidy your website FAQs and service pages, since they’ll be the chatbot’s source material.
- Write your rules: what it can answer, what it must hand over, and what it must never say.
- Set up the handover and connect captured enquiries to your CRM.
- Work through the testing checklist, then launch on a few pages first.
- Review conversations weekly and adjust.
A chatbot works best on a website that already answers people’s questions clearly, so it often goes hand in hand with website design and build. If you run a business in local services and want help setting up a chatbot with sensible limits, our AI for business work covers it.