An AI chatbot for a small business: what actually changes

01The difference with older chatbots

The old ones followed a decision tree written in advance. The moment a visitor stepped outside the script, they hit "sorry, I didn't understand".

A modern assistant reads your documents — your prices, your terms, your product sheets — and answers in your words. It handles a badly phrased question, and it knows how to say it doesn't know.

02What it actually saves

Repetitive questions disappear: opening hours, lead times, service area, terms. That's often the bulk of incoming messages.

Enquiries arrive qualified. Instead of "hi, how much?", you get the need, the context and a rough budget.

Nights and weekends stop being black holes. A visitor who gets an answer at 11 p.m. doesn't shop around the next morning.

03What you should never hand it

Firm commitments. A final price, a delivery date, a contractual promise: that stays your decision, not the machine's.

Sensitive matters. A complaint, a dispute, an angry customer: the assistant must hand over immediately, and say so clearly.

And it must never pass for a human. That's a matter of trust — and in the European Union, a matter of law.

04What makes it work

Sources that are current. An assistant wired to outdated prices repeats mistakes at scale. Its sources need an owner.

An honest scope. An assistant that answers twenty questions perfectly and hands over the rest beats one that answers everything, vaguely.

A visible way out. At any moment the visitor must be able to reach a person — that's what makes the tool acceptable.

05And data protection?

Conversations contain personal data: they must be hosted under GDPR-compliant terms, kept for a defined period, and the visitor must be told they're talking to a machine.

These points are settled at design time. Bolted on afterwards, they cost far more.