Should a small business use an AI chatbot?

An AI chat assistant is a chat box on your website that answers visitors’ questions in plain language. Older chatbots followed fixed menus. Newer ones use an AI model to understand the question and write a reply, based on information you give it: your services, prices, areas, opening hours and answers to common questions.

The good ones don’t invent. They answer from your content, say when they don’t know, and pass the person to you. Think of it as a well-briefed helper for the questions you answer ten times a week, not a replacement for you.

What it does well:

• Answers repeat questions: “Do you cover Bognor?”, “How much is a boiler service?”, “Are you open Saturday?”
• Works out of hours, when many people browse and nobody is at the phone.
• Collects the details you need (name, contact, what they want) before you call back.
• Hands the conversation to a person, by WhatsApp, email or a callback request, when the visitor asks or is ready to book.

What it can’t or shouldn’t do:

• Make up prices or promises. If a price isn’t in its information, it should say so and hand over.
• Give professional advice. Legal, medical, financial or safety advice belongs with a qualified person.
• Handle complaints. An upset customer wants a human. Route complaints straight to you.
• Replace you. It starts conversations. You still win the job.

Wrong answers are your answers. In a Canadian case in February 2024, Moffatt v Air Canada, a tribunal held the airline responsible after its website chatbot gave a customer wrong information about bereavement fares. The tribunal said it makes no difference whether information comes from a static page or a chatbot. That’s Canadian, not UK law, but the lesson travels: customers will treat what your chatbot says as what you said. Keep its information short, accurate and up to date, and test it with awkward questions before it goes live.

Data protection. Chat messages often include names, phone numbers and details about someone’s situation, which is personal data under UK GDPR. In plain words, the ICO’s guidance on the right to be informed says you must tell people, at the time you collect their data, what you use it for, how long you keep it and who you share it with, in clear and plain language. With an AI chatbot, “who you share it with” includes the AI provider that processes the messages. The ICO’s AI guidance says you need to be transparent about how you process personal data in an AI system.

In practice that means a short notice in the chat window, a link to your privacy notice that names the provider and how long chats are kept, and asking people not to share health or payment details in the chat. I’m not a lawyer and this isn’t legal advice: if you handle sensitive information, such as health details in a clinic, read the ICO guidance or get advice for your situation.

There are usually two running costs, plus whatever the set-up costs.

• AI usage. AI providers charge by the amount of text processed, priced per million “tokens” (small chunks of words). OpenAI’s published rates for its smallest models are a fraction of a US dollar per million tokens, so for a small business site the monthly usage cost is often modest. It rises with the number of conversations and the model you choose.
• The chat tool. Some tools charge a monthly fee per seat, some charge per conversation. Intercom, for example, lists its Fin AI agent at $0.99 per outcome, on top of seat prices from $29 a month billed annually.

Ask any provider how you’ll be charged if the chatbot gets busy, and whether there’s a cap. A per-conversation price that looks small can add up.

People search for “which AI chatbot is best for small business”, but the honest answer is that the best one is the one that meets these tests for your business:

1. It answers only from your content. Write down your services, areas, prices or price ranges, hours and your 20 most common questions. That document is the chatbot’s brief.
2. It says it’s an automated assistant. Don’t pretend it’s a person.
3. It hands over cleanly. One tap to reach you, and a clear message about when you’ll reply.
4. It tells people how their data is used, in the chat window and your privacy notice.
5. It has a cost you can predict, with a limit on usage.
6. You test it and check it. Ask it tricky questions before launch, then read a sample of conversations each month and fix its brief where it went wrong.

My own site has a working chat assistant. It answers from a set of M1 service information I’ve written, gives a set answer if the AI reply fails or breaks my rules, and I don’t store chat messages on my servers. The privacy page says which AI provider processes the messages.

The Website Chat Assistant puts the same kind of assistant on your current website for a fixed £1,800, paid £900 to start and £900 at launch, typically 2 to 3 weeks. I set it up with your prices, services and FAQs, in your colours, and it hands people to you on WhatsApp when they want a person. AI usage (the cost of each answer) and any chat tool fee are billed at cost or paid directly by you, and listed before you agree.

If most of your enquiries come by phone and calls get missed, the Missed-Call Rescue is the better fit, at a fixed £800. Not sure which leak is biggest? Take the Scorecard.

• Right to be informed, Information Commissioner’s Office, accessed 30 September 2026
• How do we ensure transparency in AI?, Information Commissioner’s Office, accessed 30 September 2026
• Moffatt v. Air Canada: A Misrepresentation by an AI Chatbot, McCarthy Tétrault, accessed 30 September 2026
• API pricing, OpenAI, accessed 30 September 2026
• Pricing, Intercom, accessed 30 September 2026