What an AI chatbot is and how it differs from the old ones
Chatbots from a few years ago worked with menus and keywords: if the customer did not type the expected phrase, they replied "I didn't understand". Those based on language models understand the question even if it is badly written or in another language.
That flexibility is also the risk: an AI chatbot with no boundaries can answer on anything, invent terms or promise things the business does not offer. A good chatbot has clear instructions, answers from the business's own documents and hands over to a person the cases it should not handle.
An example in an SMB
A bicycle shop puts a chatbot on its website, connected to the catalogue and the returns policy. In the evening it answers questions on sizes, availability and delivery times. When a customer asks about an unusual repair, it collects their details and books a call back.
The most common mistakes
- Letting it answer on any subject instead of only your own service.
- Hiding that it is an AI: besides being unfair, in Europe transparency is an obligation.
- Never reading the conversations: they are the fastest way to find out what customers ask.
Why it matters to you
A chatbot on your website works at night and on public holidays, when requests would otherwise go unanswered until morning. But its real value lies in the recorded conversations: they show the questions customers actually ask, in their own words, and point to which pages of your site or which procedures need clarifying.
All the words of AI explained simply are in the glossary.
Frequently asked questions
Does a chatbot have to say it is an AI?
Yes, the AI Act requires that people know when they are talking to an AI system, unless it is obvious.
Does a chatbot replace customer service?
It absorbs the repetitive questions and leaves the cases that need judgement to people.
How long does it take to get one online?
A first version covering a few frequent questions can be prepared quickly. The real work comes afterwards: reading the conversations, correcting the answers and widening the topics a little at a time.
Read next
What a chatbot's system prompt is
The system prompt is the set of fixed instructions the model receives before every conversation: who it is, how it should speak, what it can and cannot do.
GlossaryWhat is RAG: AI that answers from your documents
RAG (Retrieval Augmented Generation) is a technique where, before answering, the AI searches your documents for the useful passages and uses them as the basis for its answer.
Rules without the legaleseAI Act: what it really requires of an SMB using AI
If you use ChatGPT in the office or a chatbot on your website, the obligations are few and clear. The heavy ones concern other uses, and they have been postponed.
GlossaryWhat a voicebot is and when it makes sense on the phone
A voicebot is an assistant that talks on the phone or by voice: it listens, transcribes what you say, understands the request with an AI model and replies in a synthetic voice.
GlossaryWhat an AI agent is, and how it differs from a chatbot
A chatbot answers. An agent acts: it decides the steps, uses software and sees a task through.
Want to understand what your business really needs, without the jargon? Let's talk.
Text and graphics created with the support of AI. Method and content are AIR's.