Chatbot.

There are two basic types. Rule-based chatbots work with predefined buttons and decision trees: the user picks 'opening hours' and gets the opening hours. They are predictable but rigid. AI chatbots use a large language model and understand questions in the user's own words. To answer correctly, they are given access to verified company content — usually through RAG — and clear rules about what they may and may not say.

A chatbot is useful where many similar questions come up and quick answers make a difference: opening hours, prices, procedures, availability, the status of an order. It can answer outside office hours, pre-qualify enquiries and pass them to the right person. Once it is meant to carry out actions, such as booking an appointment, the chatbot becomes an AI agent with access to tools.

The limits are just as clear. A chatbot without a good knowledge base makes up answers. A chatbot that offers no hand-over to a person frustrates. And a chatbot for a company with ten enquiries a week is often an expensive toy, where a good FAQ page and a phone number would help more. Since August 2026 the EU AI Act also requires that users in the EU can tell when they are talking to an AI.

A chatbot becomes measurable through a handful of metrics: the share of conversations resolved without human help, the number of enquiries handed over, satisfaction after the conversation and — especially important — a sample of answers that a person reviews regularly.