A chatbot is software that carries on a text conversation with a person, usually a customer, to answer questions or complete simple tasks without a human on the other side. It typically appears as a chat window on a website or in an app, or inside a messaging channel such as SMS or a social messaging service. Some chatbots follow fixed scripts and buttons; others use AI to understand free-form questions and compose replies.
At a glance
- Chatbots handle text conversations; voice equivalents are usually called voice bots, IVR or virtual agents.
- They range from simple decision trees to AI-driven bots that generate answers from a knowledge base.
- Common jobs include answering frequent questions, checking order or account status, collecting details and routing to a person.
- A good handoff to a human agent, with the conversation history attached, matters as much as the bot itself.
- Results depend heavily on the content and systems the bot can reach and on ongoing tuning.
What problem it solves
Many customer questions are repetitive: store hours, password resets, order status, how to update a payment method. Answering each one with a person is expensive, and customers who just want a quick answer at night or on a weekend may have to wait. A chatbot can answer these around the clock and let human agents spend their time on conversations that need judgment.
A chatbot can also gather information at the start of a conversation, such as an account number or the reason for contact, so that when a person does take over, they don’t have to start from zero. For buyers, the goal is usually fewer simple contacts reaching agents without making customers work harder to get help.
How it works
Understanding the request. A scripted chatbot presents menus or buttons and follows a decision tree. An AI-driven chatbot uses conversational AI to work out what the customer is asking from free text. Many newer bots use a large language model (LLM) to interpret questions and draft answers, a form of generative AI.
Finding the answer. The bot looks up an approved answer, searches a knowledge base, or calls other systems through integrations, for example to check an order or reset a password. What a bot can do is limited by what it is connected to.
Responding and handing off. The bot replies in the chat window. When it can’t help, or the customer asks for a person, it should pass the conversation to a live agent, ideally in the same window and with the history attached so the customer doesn’t repeat themselves. In a contact center, that handoff goes into the same queues as other chats.
Improving. Teams review transcripts, see where customers got stuck or gave up, and update content, flows and integrations. Bots that are launched and left alone tend to get worse as products and policies change.
When it matters for buyers
- When a large share of contacts are simple and repetitive, and those contacts are taking up agent time.
- When customers expect help outside business hours.
- When adding chat to your website, because deciding the bot’s role, and its limits, early shapes the whole experience.
- When a vendor promises a resolution rate. Ask how resolution is defined; a customer who gives up is not a resolved customer.
- When the bot will touch personal or regulated data, which raises questions about data handling, retention and where AI processing happens.
Our contact center as a service page covers how bots fit into a wider contact center platform.
Questions to ask vendors
- Is the bot scripted, AI-driven or a mix, and can we control exactly what it is allowed to say?
- What systems can it connect to out of the box, and what requires custom work?
- How does handoff to a live agent work, and does the agent see the full conversation?
- How do you define a “resolved” or “contained” conversation in your reporting?
- How is it priced: per conversation, per resolution, per message or flat?
- Where is customer data processed and stored, and is it used to train models?
- What tools do we get to review transcripts and improve the bot over time?
How it differs from an intelligent virtual agent (IVA)
An intelligent virtual agent (IVA) is usually a more capable, AI-driven assistant that can work across voice and text, keep context through a longer conversation and complete transactions in back-end systems. “Chatbot” is the broader, everyday term for any bot in a chat window, including simple scripted ones. In practice the line is blurry and vendors use both terms; judge a product by what it can actually do in your environment, not by its label. A chatbot also differs from agent assist, which helps human agents rather than talking to customers.