What Is a Chatbot? And When Should You Actually Use One?

Chatbots have a reputation problem. Years of frustrating customer service bots, "I'm sorry, I didn't understand that. Did you mean: return policy?", have made people sceptical of the entire category.

The technology has changed dramatically. But the scepticism is still worth taking seriously, because a badly implemented chatbot creates more frustration than no chatbot at all.

Two types of chatbot

The old kind: rule-based. Someone writes decision trees, if the customer says X, respond with Y. Predictable, controllable, brittle. Falls apart quickly when a customer asks anything outside the script.

The new kind: AI-powered. Built on large language models, these can understand natural language, handle varied questions, and generate responses that actually address what was asked. They are not perfect, but they handle ambiguity far better than their predecessors.

When a chatbot makes sense

A well-implemented AI chatbot adds genuine value in specific situations. When the same questions are asked repeatedly and the answers are consistent, hours, pricing, returns, product specs. When customers need answers outside business hours. When the volume of enquiries is too high for a small team to handle without delays.

When it does not

When the questions require genuine judgment, access to customer-specific data, or emotional sensitivity. When the stakes of a wrong answer are high. When your customers expect human contact and a chatbot would feel dismissive.

The honest rule: a chatbot should handle what is genuinely handleable, and make it frictionless to reach a human for everything else. Hiding the "speak to a person" option is the fastest way to destroy trust.