A chatbot that really helps your customers.

What is a chatbot?

A chatbot is a program that talks to visitors on your website, in an app or through messages. The visitor types a question and the chatbot answers straight away. That way customers get help without waiting for an email or a phone call.

There are two kinds of chatbots, and the difference decides what you can do with one in practice.

The difference between a regular chatbot and the new generation

A classic chatbot works with fixed scripts. The visitor picks a button or types a keyword, and the bot follows a pre-programmed path. That works for simple questions, but as soon as someone phrases it differently or asks something unexpected, it gets stuck and says it does not understand.

The new generation, an AI chatbot, is built on a large language model. It understands everyday language, including typos, casual phrasing or a question in two parts, and writes its own answer. That makes it feel like a conversation instead of a menu.

An AI chatbot can also make mistakes if you do not set limits. That is why I only let it answer from your own information, and it refers visitors to you when it does not know something.

One step further are AI agents, which do not only answer but also take actions, such as forwarding a request or suggesting an appointment. We discuss that when it fits your business.

How did we get here? A short history

Chatbots are older than you think. In 1966 Joseph Weizenbaum built ELIZA at MIT, a program that imitated a therapist by responding to keywords. It understood nothing, yet people talked to it anyway. That is how the first chatbots came about: rules and scripts.

For decades chatbots kept working that way, with menus and fixed answers, on websites and later in messaging apps and voice assistants.

The turning point came in 2017, when Google researchers described the transformer architecture, the technique modern language models run on. At the end of 2022 OpenAI released ChatGPT, and everyone could experience for themselves that a computer understands everyday language and writes sensible answers. Since then the models have followed each other quickly, including open models that you can run on your own computer.

Where does AI come from and how does it work?

Modern AI comes from machine learning: instead of giving a computer rules, you let it learn patterns from huge amounts of examples. A language model learns this way from a gigantic amount of text.

At its core it keeps predicting which piece of text logically comes next. By practising that billions of times, the model learns grammar, facts and writing styles. So it does not look anything up, it writes an answer that fits. That is also why it can sometimes say something wrong with great confidence.

For a business I make that reliable by having the model first look up the right passage in your own documents for every question and answer from it. Specialists call that retrieval-augmented generation, RAG for short. For you it means: answers from your own price list and terms, not from the internet.

What does an AI chatbot do for your business?

You get the same questions every week: are you open on Saturday, roughly what does it cost, do you also deliver to Geel? An AI chatbot catches those questions, also in the evening and at weekends. Visitors get an answer immediately instead of waiting for an email, and you get time back.

On the demo page you can test a classic bot and an AI bot side by side yourself.

Trained on your information, not on the whole internet

The bot only answers from what you provide: your website, a price list, a manual, frequently asked questions. That way it does not make things up about your business. Before it goes live, we test it together with the hardest questions your customers ask.

If a visitor leaves an email address or phone number, that enquiry lands straight in your inbox. That is how a question at 11 pm becomes a new customer the next morning.

In the cloud or fully local

By default the chatbot runs through a fast, affordable AI service. If you work with sensitive data, such as in a practice or law firm, it can also run entirely locally on a server at your place, so no question ever leaves your building.