AI agents vs chatbots: what is the difference?
Kahu team·Updated 1 Oct 2026·5 min read
A chatbot holds a conversation and answers questions. An AI agent is given a goal and uses tools to complete it, such as booking an appointment, updating a record or sending a follow-up. If your problem is answering the same questions, a chatbot may be enough. If the problem is the work that follows the answer, you need an agent.
The difference in practice
A customer asks whether you have a slot on Friday. A chatbot says the calendar shows openings at 10 and 2. An agent checks the calendar, offers the times, books the one they choose, sends a confirmation and adds them to your customer list.
| Chatbot | AI agent | |
|---|---|---|
| Main job | Answer | Act |
| Uses tools | Rarely | Calendar, inbox, CRM, web |
| Works between messages | No | Often |
| Needs permissions set | Few | Yes, carefully |
| Risk if wrong | A poor answer | A wrong action |
Which one you need
- Most enquiries are simple questions with fixed answers: a chatbot or FAQ page may do.
- Enquiries need a booking, quote or follow-up: an agent saves the work.
- You are worried about mistakes: an agent with approval steps gives you control.
The line is blurring
Many chat products now include agent features, and many agents include a chat window. Ask a vendor what the product can do without you, which systems it connects to, and what it asks before acting. Those answers matter more than the label.
Frequently asked questions
Is ChatGPT a chatbot or an agent?
Both, depending on the feature. Standard chat answers; dots and connected tools let it act.
Are agents more expensive?
Usually, because they do more and connect to more systems. Compare the cost with the hours of work they remove.
Can an agent go wrong?
Yes. That is why approvals, limited permissions and a record of actions matter.
