Chatbot: definition, limits and use
Why the Term Matters
Chatbot is the oldest and most stretched term in the field. Some expect a rigid FAQ bot, others a full assistant. Clarifying it settles the expectation. A chatbot is strong at frequent, clearly bounded questions. As soon as real actions and system access are needed, the fitting term shifts to an AI assistant with the right permissions.
When It Is Relevant
- Many similar questions reach a single channel, for example the website.
- Users should get a first answer around the clock.
- The topics are manageable and rarely change much.
- A fast, visible relief of the first contact is the goal.
When Not
When requests are individual and need access to customer data or systems, a pure chatbot hits its limit. Then the path leads to an AI assistant. And when the knowledge base is thin or contradictory, even a language model produces no reliable answers. Clean knowledge first, then a surface on top.
Industries That Use It
- Online retail and e-commerce: answer shipping, return and availability questions directly on the product page.
- Travel and tourism: reliably catch booking and travel questions at peak, without a queue.
- Telecommunications: bundle standard questions on tariffs and outages as a first filter.
- Education and EdTech: guide learners with organisational questions and point them to fitting content.
Typical Use Cases
- An FAQ bot on the website that answers the most frequent questions immediately.
- A support bot in a messenger that pre-qualifies requests and routes them.
- A product advisor that suggests fitting options from catalogue data.
- A bot for appointment prep that collects the request before a human takes over.
- An onboarding bot that guides first-time users through the key steps.
Decision Guide
| Signal | Chatbot fits | Alternative |
|---|---|---|
| Many similar questions, one channel | lean chatbot | AI assistant when action is needed |
| Access to customer data required | rather not | AI assistant with permissions |
| Answers from expertise needed | chatbot on a company brain | company brain as the base |
| Channel is the phone line | does not fit | voice agent or voicebot |
Example
An online retailer receives many similar shipping and return questions at peak. In the workshop a narrow topic area is chosen and a chatbot is tested as a prototype on real requests, including cases where it hands over to the support team. Only after the test does the team decide on a fixed embedding into the website. This describes the method, not an asserted metric.
Service Connection
The entry is the AI workshop to choose the topic, then AI development and, for the fixed embedding, AI implementation.
Sources
This framing describes common practice when building chatbots. Scope, knowledge base and integrations depend on the concrete project and should be checked there.