Chatbot
A program that talks to people in natural language
What is Chatbot?
A chatbot is a computer program that holds a conversation with people in natural language, usually in writing in a chat window on a website, in an app or in a messenger, and sometimes by voice. The user types a question and the chatbot answers it, offers a next step or hands the conversation over to a person. Older chatbots follow predefined scripts and buttons. Modern AI chatbots understand free-form text thanks to large language models (LLMs) and build their answers from the information a company gives them. Businesses use them mainly in customer support, in sales and for internal employee questions.
Why It Matters
Customers expect an answer straight away, including evenings and weekends. A chatbot handles repetitive questions such as opening hours, order status, prices or how to file a complaint without waiting and without tying up the team, so people can focus on more complex cases. A well-configured chatbot also collects enquiries outside office hours and shows what customers ask about most. A badly configured one drives customers away: it does not understand, goes round in circles or makes things up. What matters most is good source material, clear limits on what it may answer and an easy hand-over to a person.
Real-World Examples
- 1Online shop: after the customer enters an order number, the chatbot checks the status in the shop system and says when the parcel will arrive.
- 2Service company: the chatbot asks for the device type and a description of the fault and creates a ticket with contact details in the CRM.
- 3Accommodation: answers to common questions about check-in, parking or pets in several languages.
- 4Internal helpdesk: an employee asks how to request leave and the chatbot answers from the internal policy.
- 5Lead qualification: the chatbot asks three or four questions (scope, deadline, budget) and passes a ready-made enquiry to the sales team.
How a chatbot works
Every chatbot, simple or AI-powered, goes through the same four steps:
- Input: the user types a message, clicks a button or speaks.
- Understanding: a rule-based chatbot looks for keywords or waits for a menu choice. An AI chatbot uses a language model to grasp the meaning of a sentence, even with typos or different wording.
- Source of the answer: a predefined script, the company's knowledge base (FAQs, price list, terms, documentation) or live data from other systems connected via an API, such as an online shop, booking system or CRM.
- Answer or action: the chatbot replies, fills in a form, creates a request or, when it cannot help, hands the conversation over to a person.
Types of chatbots
- Rule-based (scripted) chatbot: guides the user through a predefined tree of questions and buttons. Predictable and cheap to run, but it cannot answer anything outside the script.
- AI chatbot: built on a large language model. It understands free text and copes with questions its author did not anticipate. You need to make sure it answers only from verified company material.
- Hybrid chatbot: AI for understanding and routine questions, a fixed script for sensitive steps such as orders, payments or complaints. In practice the most common choice for businesses.
- Voice chatbot (voicebot): the same principle over the phone or in a voice assistant.
By audience, there are customer chatbots (website, online shop, social media) and internal chatbots for employees (helpdesk, policies, onboarding).
Chatbot vs AI assistant vs AI agent
The terms are often mixed up. A chatbot is a conversational interface, typically for one company's customers and limited to its topics. An AI assistant helps an individual with their work: writing, summarising, analysing. An AI agent goes further and completes multi-step tasks on its own using other tools. What AI agents can do in a business is covered in our article on AI agents.
A short history of chatbots
The first chatbot is generally considered to be ELIZA, created by Joseph Weizenbaum at MIT between 1964 and 1967. It worked only with simple text patterns, yet some users felt it understood them. Scripted, button-based chatbots then dominated for decades. The turning point came on 30 November 2022, when OpenAI released ChatGPT: since then AI chatbots that understand everyday language have spread to smaller businesses too.
How to introduce a chatbot in your business
- Collect the most common questions: go through emails, calls and messages from recent months and pick the recurring ones. They form the core content.
- Prepare material and limits: FAQs, prices, terms. Also decide what the chatbot must not answer, such as custom quotes or legal and medical advice.
- Decide on integrations: should the chatbot only answer, or also check order status, book appointments and create enquiries in the CRM?
- Set up a hand-over to a person: a button to reach staff or a form sent to the team together with the conversation history.
- Test, launch and measure: try the chatbot on real customer questions; after launch, track the share of resolved conversations, the number of leads and the questions it could not answer, and keep adding material.
Chatbots and the law: AI Act and GDPR
Since 2 August 2026 the transparency obligations of Article 50 of the EU Artificial Intelligence Act (AI Act) apply. An AI system that interacts directly with people must be designed so that users know they are dealing with AI, unless this is obvious from the circumstances. For a website chatbot, a clear notice, for example in the opening message, is enough.
A chatbot also processes personal data: name, email, phone number and the content of the question. The same rules apply as for a contact form: inform people about the processing, keep only what is necessary and have a data processing agreement with the tool provider. See GDPR and personal data protection. This overview is not legal advice.
When a chatbot pays off
A chatbot makes sense when a business receives many repetitive questions, customers write outside office hours or you communicate in several languages. If you get only a few questions a week, or each one needs individual judgement, a well-written FAQ page and a contact form are enough.
We design website chatbots, train them on your material and connect them to your online shop, bookings or CRM as part of our AI solutions for businesses. Connections to other systems and the follow-up processes are part of digitalisation and process automation.
Frequently asked questions
What is a chatbot?
A chatbot is a program that communicates with people in natural language, usually in a chat window on a website, in an app or in a messenger. It answers questions, offers next steps and hands the conversation to a person when needed. Businesses use it mainly in customer support and to capture enquiries.
What is the difference between a chatbot and an AI chatbot?
A classic chatbot follows a predefined script and buttons, so it only answers what its author planned for. An AI chatbot uses a large language model, understands free-form text and builds answers from the company's material. It is more flexible, but the accuracy of its answers needs to be monitored.
What was the first chatbot?
The first chatbot is generally considered to be ELIZA, created by Joseph Weizenbaum at MIT between 1964 and 1967. The program reacted to patterns in the text and imitated a conversation with a psychotherapist. It did not understand the actual meaning of sentences.
Does a chatbot have to say that it is AI?
Yes. Since 2 August 2026, Article 50 of the AI Act requires AI systems that interact with people to be designed so that users know they are dealing with AI, unless this is obvious from the circumstances. For a website chatbot, a clear notice in the opening message is enough.
Can a chatbot give wrong answers?
Yes. An AI chatbot sometimes produces a convincing but false answer (a so-called hallucination), while a rule-based chatbot fails to understand questions outside its script. Limiting answers to verified company material, ruling out sensitive topics, testing on real questions and offering an easy hand-over to a person reduce the risk.