What it decides on its own
Autonomy is only useful if you know where it stops. Here is the honest split.
It decides on its own:
- What a question is really asking, including when it is phrased badly
- Which part of your content answers it
- What language to reply in — it matches the visitor
- Whether the moment calls for a button, a contact form, or just an answer
- Whether it has enough information to answer at all
It comes to you:
- When it could not answer, the gap is logged for you to fill
- When a visitor asks for a person, it shows your contact card
It never:
- Invents facts that are not in your knowledge base
- Makes changes in your other systems mid-conversation
- Passes the conversation to a shared inbox — there is not one
See how it builds itself
Is an AI agent the same as a chatbot?
Mostly, yes. The words are used interchangeably by most people searching for either.
Where there is a difference, it is about how much the thing works out for itself. A chatbot in the older sense answers from a script somebody wrote. An agent is given a role and a set of knowledge, and decides how to respond within that.
This product is both, depending on which word you use. It appears on your site as a chat widget. Behind it is an agent that reads your content, judges intent, and picks its own response.
How to add an AI agent to your website
Three things it is not, so nobody finds out the hard way.
- It is not a shared inbox. There is no queue of conversations for a team to work through, because the point is that nobody has to.
- It does not transfer a live conversation to a person. When someone wants a human, it shows your contact card with a WhatsApp, call or email button and they carry on there.
- It is not a flow builder. There is no canvas of branching logic. You steer it with written rules instead.
If you need a shared inbox with agent handover, Tidio and Intercom are built for that and we are not going to pretend otherwise.