Does a small business need a chatbot?

Does a small business need a chatbot? Often not. A page that answers the question clearly, and a phone number that gets picked up, beats most chatbots we have seen.

That is not a case against them. There is a narrow set of jobs where a chatbot genuinely earns its place, and a wider set where it makes things worse. Knowing which side you are on is worth more than choosing a tool.

Two very different things get called a chatbot

A scripted bot follows a tree. Press 1 for hours, press 2 for pricing. It does exactly what it was built to do and nothing else. It is predictable and cheap to run, and it frustrates people the moment their question is not on the menu.

An AI chatbot reads what the customer typed and writes an answer, usually drawn from your own documents. It handles questions nobody planned for. It can also state something false in a confident voice, which a scripted bot cannot.

Most businesses need to decide which of these they are buying before they decide whether to buy at all.

When a chatbot is worth having

The same ten questions arrive all day. Hours, delivery areas, price lists, order status, how to book. If your team answers these by hand and the answers do not change, a bot saves real time. This is the same logic as what to automate first: a repeated step that follows a rule.

Customers write outside working hours. A reply at 11pm that answers the question or books a slot is better than silence until morning.

The answer lives in a system. "Where is my order" is a lookup. A bot connected to your order data answers it faster than a person and never gets tired of it.

You have the documents. An AI bot is only as good as what it reads. If your prices, policies and product details are written down and current, it has something to stand on.

When it is the wrong purchase

You get a few enquiries a day. At that volume a person answers faster and better. The bot costs money to build and attention to keep correct.

Your website does not answer the basics. A chatbot placed over a confusing site hides the problem rather than fixing it. Fix the page. Most enquiries are lost after they arrive, not before.

Questions are high-stakes. Pricing quotes, medical advice, legal terms, refunds. A wrong answer here costs more than the wait for a human. Let the bot collect the details and hand over.

Nobody will read the failures. Every bot needs a person who reads the conversations it got wrong, every week. Without that, it quietly gets worse.

A chatbot that cannot say "I do not know, here is a person" is not finished, however good the rest of it is.

What it should never do

Never pretend to be a person. Never invent a price, a date or a policy. Never trap someone in a loop with no way out. Never take a payment or change an order without a clear confirmation from the customer.

A well-built one has a visible way to reach a human and passes over the whole conversation when it does. That handoff is the part worth testing hardest.

What it costs

A scripted bot on an existing platform can be set up for a small sum. An AI bot connected to your documents and systems is a small custom project, and it carries a running cost: model usage, hosting, and the weekly review. Budget for the review as seriously as the build.

If the channel is WhatsApp, read what the rules allow first, because they decide what the bot may do before the technology does.

A sensible way to start

Take a month of customer messages. Count how many were one of the same ten questions. If most were, a bot is worth a small trial, built by a workflow automation project or an off-the-shelf tool. If they were all different, you do not have a chatbot problem. You have a service business, and a person is the right answer.