Learn AI /AI Agents and Automation /Chatbot
Chatbot
- Starter
- AI Agents and Automation
- Programme & data teams
Also called: conversational agent.
Back to the milestone you came from
In five seconds
What it is
A chatbot is a system that takes messages in ordinary language and returns replies in a back-and-forth conversation.
In plain words
A chatbot is a programme you chat with in ordinary language to get answers or help with simple tasks.
An everyday example
Imagine a reception desk in a large community centre. Visitors come up one by one and ask questions such as opening hours, room bookings, or where to find a form. The receptionist listens, looks at a guidebook, and gives an answer in conversation. For some requests, the receptionist can also phone another office to confirm a booking or transfer the visitor to the right team. The exchange feels like a conversation, but the quality depends on the guidebook, the phone list, and whether the receptionist is allowed to take further action.
A human receptionist can use broad real-world judgement, notice social context, and take responsibility in ways a chatbot cannot. Many chatbots do not truly check facts or understand the situation beyond patterns in text. Also, a receptionist is a person with continuous awareness, while a chatbot is software that processes each turn through programmed components.
Remember thisA chatbot is defined by its turn-by-turn conversational interface. It may retrieve set replies, generate text or call other tools, but that alone does not make it an autonomous agent or a reliable source.
Understand it
Understand it
Why it matters
Many people now meet AI through chatbots, so it helps to know what they can and cannot do. If you understand the term, you can ask better questions before using one: Is it only answering, or can it also take actions? What data does it use? Is a human checking the result? That matters for procurement, risk, privacy, accessibility, service quality, and accountability. A chatbot can speed up routine communication and first-pass support, but if people mistake it for a reliable expert or a fully autonomous agent, they may trust it too much or give it tasks it should not handle.
How a chatbot turns a message into a reply
After viewing, the reader should be able to describe the basic flow: user message in, processing and optional tool access, reply out.
01
1. User message
A person types or speaks a question, request, or instruction in ordinary language.
Draw a message bubble icon entering the first box labelled 'User message'.
02
2. Chatbot processing
The system interprets the message using its rules, prior conversation, and possibly a language model that generates likely wording.
Second box labelled 'Process message'. Include small stacked layers labelled 'instructions', 'chat history', and 'model'.
03
3. Optional tool step
If configured, the chatbot may query a database, search approved content, or use another software function before answering.
Add a branch from the processing box to a smaller box labelled 'Optional tool or data source', then back to the main flow.
04
4. Reply to user
The chatbot returns a conversational answer, suggestion, question, or action prompt.
Final box labelled 'Reply' with an outgoing speech bubble icon.
Show a left-to-right flow with four boxes connected by arrows. Read it from the user message to the final reply. A smaller optional branch shows that some chatbots call a tool before replying, while others answer directly from their built-in instructions and model.
Hover or tap each box to reveal a short example such as a user question, an internal step, a tool lookup, and the final reply.
Describe this without the diagram
The diagram is a four-step flow read from left to right. Step 1 shows a user sending a message. Step 2 shows the chatbot processing that message using instructions, earlier chat turns, and possibly a language model. Step 3 is optional: some chatbots connect to a tool or data source, such as a database or search system, and feed the result back into processing. Step 4 shows the chatbot returning a reply. The main lesson is that a chatbot is a conversation interface, and some, but not all, also use tools before answering.
With reduced motion: A static four-box flow replaces animation. Solid arrows show the main path from user message to reply, and a dashed side loop shows the optional tool step.
A chatbot lets a person type or speak a message in everyday language and get a conversational reply. The input is usually a question, request, or short instruction. The output is usually text, and sometimes voice, buttons, or links.
At the simplest level, a chatbot matches a message to prepared answers or uses a language model to generate a reply. A language model predicts likely next pieces of language from patterns in large amounts of text. Some chatbots only answer. Others can connect to tools such as a calendar, database, or ticketing system.
A chatbot is not automatically an AI agent. It is a conversational system defined by turn-by-turn interaction. An AI agent can pursue a goal by choosing and carrying out multiple steps, often using tools. A chatbot may include those features, but many do not.
A chatbot is also not the same as a large language model. The model is one component that may generate words. The chatbot is the user-facing system around it, including the chat window, instructions, safety rules, memory settings, and sometimes connections to other software.
An AI assistant is defined by helping with tasks, while a copilot works under human direction in digital tasks. Either may use chat, but not every chatbot has those roles.
A chatbot can be useful for routine questions, drafting, and first-pass sorting. Its replies still need checking, especially when accuracy, rights, safety, or sensitive information matter.
How “A reception desk with a script and a phone list” maps onto the real thing
| In the story | In the AI system |
|---|---|
| visitor questions at the desk | user messages in a chat window |
| guidebook and prepared notes | rules, instructions, and stored content used to form replies |
| receptionist speaking back | chatbot response in text or voice |
| phone list to contact other offices | connections to tools, APIs, or other systems |
A worked example
Resolving an indicator discrepancy while drafting the quarterly donor report
Starting from: Statistician: "Why does the narrative use a different total from the results table?" Follow-up: "Which figure should I use for this quarter, and can you draft two sentences explaining the change?"
- Through an approved connector, the chatbot retrieves the relevant aggregate values, indicator definitions and reporting-period labels.
- It compares the narrative and table, then asks whether the report should follow the current-quarter cut-off or the cumulative reporting period.
- When the statistician replies, "Use the current quarter", the chatbot retains that context rather than requiring the full question again.
- It identifies the validated value, drafts two sentences and points to the source cells for human checking. It does not alter the table or send the report for clearance.
Result: Reply: "The difference comes from the reporting cut-off: the narrative is cumulative, while the table is for the current quarter. Use the value marked as validated for the current quarter. I have drafted two sentences and flagged the source cells for checking before clearance."
In a United Nations workplace
Before the quarterly report to a donor goes for clearance, a statistician in the Resident Coordinator's Office uses an approved internal chatbot to query a validated aggregate results table. The statistician asks follow-up questions about indicator definitions, reporting periods and apparent discrepancies, then requests draft language for the results section.
The conversation lets the statistician refine the request turn by turn. The chatbot can retain short-term context, retrieve authorised material through a connector and draft text, but it does not make the figures authoritative or clear the report.
Before you try thisDo not upload row-level records, personal data or restricted donor material to a chatbot unless the tool is expressly approved for that information and use. Connect only sources it is authorised to access. Check every figure, definition and drafted claim against the validated table before the report enters clearance.
Where you may hear it
- A vendor demo during an ICT governance board meeting
- A service desk review at Headquarters
- A reporting discussion before close of business
- A procurement clarification meeting for an internal helpdesk tool
Avoid confusion
Avoid confusion
Often confused with
AI agent
An AI system that can pursue a goal by selecting and carrying out multiple steps, often using software tools, with oversight matched to its risks.
Large language model
A very large text-processing model predicts likely next pieces of language from a prompt and uses that pattern-matching ability to generate or transform text.
Copilot
A tool that helps a person complete digital tasks by drafting, searching, and sometimes taking limited actions under human direction and review.
AI assistant
Software that uses AI to help with tasks by taking instructions, sometimes using connected tools, and returning a draft, answer or proposed action for a person to review.
Side-by-side
Common misconceptions
Common misconception: A chatbot is always an advanced AI agent.
Correction: Not necessarily. Some chatbots only follow scripts or generate replies without taking further steps. An AI agent usually refers to a system that can plan, use tools, and act across multiple steps. A chatbot may be agentic, but the term itself mainly describes the conversational interface.
Common misconception: A chatbot and a large language model are the same thing.
Correction: They are related but different. A large language model is one possible engine for generating text. A chatbot is the whole application a person interacts with, including the chat interface, instructions, safety settings, memory choices, and sometimes tool connections.
Common misconception: If the reply sounds fluent, it is probably correct.
Correction: Fluent wording is not proof of accuracy. Many chatbots generate plausible text from patterns and may produce mistakes, omissions, or invented details. Important outputs still need verification against trusted sources and human judgement.
Quick check
Which statement best describes a chatbot?
The core idea is the conversational interface: a person sends a message and gets a reply. Some chatbots are simple, and some can connect to tools, but not all are autonomous agents. A large language model or database may be part of the system, but neither is the chatbot by itself.
Go deeper
Go deeper
The simple version
A chatbot is a programme you chat with in ordinary language to get answers or help with simple tasks.
The technical version
A chatbot is a conversational interface that accepts natural-language input and returns responses turn by turn. The response mechanism may be rule-based retrieval, template selection, or generation by a large language model. More capable chatbots may maintain short-term conversation state, apply system instructions, and invoke external tools through APIs or connectors. However, the term 'chatbot' describes the interaction mode more than the underlying architecture. That is why it should be distinguished from a language model, which is only one component, and from an AI agent, which usually implies multi-step action, tool use, and some degree of autonomy within set limits.
Reviewed external resources
Primary sourceModel Context Protocol (opens in a new tab)Model Context Protocol
This link explains the Model Context Protocol, which suits readers who want to understand one way chat systems can connect to tools and data sources.
Deeperlink checked 2026-07-31
EncyclopaediaChatbot (opens in a new tab)Wikipedia
This link gives a broad overview of the term, its history, and common examples, and it suits readers who want a general starting point.
Intermediatelink checked 2026-07-31
ArticleBuilding effective agents (opens in a new tab)Anthropic
This link explains agents and helps readers see how a chatbot differs from a system that can take multi-step actions, which suits readers comparing terms.
Intermediatelink checked 2026-07-31
VideoVideo explanations: Chatbot (opens in a new tab)YouTube search
This search results page surfaces multiple video explanations of chatbots and suits readers who prefer to compare several visual introductions rather than read one article.
search results, not one pagelink checked 2026-07-31
DiscussionCommunity discussion: Chatbot (opens in a new tab)Reddit search
This search results page surfaces community discussions about chatbots and suits readers who want to see common questions, practical experiences, and differing views.
search results, not one pagelink checked 2026-07-31
Where this came from
Drafted with a language model through an offline pipeline, checked against this site's schema, and every external link above fetched to see whether it answers. Treat it as a starting point and check anything that matters against the primary sources. How this site is built and what that means.
The one thing to remember
A chatbot is defined by its turn-by-turn conversational interface. It may retrieve set replies, generate text or call other tools, but that alone does not make it an autonomous agent or a reliable source.