Chatbots vs AI agents: what is the difference

The plain definition

A chatbot waits for a question and replies to it. An agent is given a job and works it through to done.

That is the whole difference. Everything else follows from it.

Where the line sits in practice

A chatbot answers what your hours are. An agent takes the enquiry, works out what the person needs, checks the diary, books the appointment and sends the confirmation before the conversation ends.

A chatbot says an invoice is overdue. An agent chases it on a schedule until it is paid, nudges the right person at the right time, and reconciles what comes back into your records.

A chatbot points a website visitor to your services page. An agent answers the question they came with, in their language, and takes their details for follow-up.

The chatbot ends the interaction. The agent ends the task.

Why it matters when you buy

Most of the frustration people have with AI on websites comes from chatbots pretending to be more than they are. The visitor types a real question, gets a menu of links, and leaves. In some ways that is worse than having nothing, because it signals that nobody is home.

There is a second problem hiding under the hype. Chatbots get sold as automation because they remove a few emails. But a chatbot cannot act. It cannot book, chase, check or file. If your goal is to take work off your team, most of what gets sold as a chatbot is theatre.

An agent done properly is quiet about what it does and strict about where it stops. It works to your rules. Anything outside them goes to a person with the full story attached, rather than being guessed at by a language model at midnight.

Where chatbots still make sense

Honest answer: plenty of places. A small website that needs to answer four questions and capture an email is well served by a simple bot. The mistake is paying agent prices for bot work, or expecting a bot to run a booking flow.

The rule of thumb is effort. If the task has steps, it is agent territory. If it has answers, a bot is fine.

What separates a good agent from a demo

Demos run on the happy path. Real work does not. The caller interrupts. They change their mind halfway through. They ask a question nobody anticipated.

The engineering that matters is what happens at those edges: holding context across a conversation, knowing when to ask a clarifying question instead of proceeding, and stopping cleanly when something falls outside scope. That is the difference between an agent that survives a week with real customers and one that gets quietly switched off in month two.

What we build

Our agents come in three kinds. Personal agents that run an inbox or a calendar. Functional agents built for one job, like a marketing agent or an accounts agent. And call taking agents that answer the phone and book the job in.

All of them work from the same knowledge about your business, and all of them escalate rather than improvise. If someone is quoting you an agent, ask one question: what happens when the request falls outside what it knows? The answer tells you which one you are being sold.