A Chat Widget That Does More Than Answer
Most website chat widgets are a search box wearing a friendlier font. This is the same familiar chat interface, wired to the systems your team already uses, so it can finish the task inside the conversation instead of pointing a visitor somewhere else to do it.
The Same Agent, Three Common Surfaces
The interface is almost always a familiar chat window. What differs is where it sits and what context it has when a conversation starts.
Website Widget
The launcher in the corner of a marketing site or app — usually the first thing a visitor tries before searching or emailing.
In-App Panel
Inside a logged-in product, where the agent already knows who is asking and can act on their account directly.
Help Centre Search
Replacing a keyword search box with something that can act on the answer it finds, not just link to an article.
What a Chat Agent Needs Access To
A Real Knowledge Source
Your actual documentation, help centre or policy pages — not a rewritten summary that can drift out of date on its own.
Write Access to the Right System
A CRM, order system, helpdesk or booking tool it can update, scoped to exactly the fields the task needs and nothing wider.
A Way to Identify the Visitor
Where one exists — an account login, an order number, a support ticket ID — so it can act on a specific record instead of asking generic questions.
A Live Handoff Path
A real human channel to hand off to, with the conversation history attached, for the cases it should not resolve alone.
What It Handles Alone vs. What It Hands Off
Routine, Reversible Actions
Answering from documentation, looking up an order or account status, updating low-stakes fields, logging a request.
Ambiguous or High-Stakes Requests
Refunds above a set threshold, account changes with security implications, or a visitor who is clearly frustrated and needs a person, not another prompt.
How This Differs From a Support Agent
This page is about the channel — the chat window itself, wherever it sits. AI customer support is about the job of resolving post-sale questions and tickets, which might run through chat, email or phone. The same underlying agent can power both; this page covers what is specific to the text interface.
What a Good Brief Looks Like
Bring
The two or three questions visitors ask most often, the systems those answers live in, and which of those tasks should end in an action rather than just a reply.
We Define Together
Where the widget lives, what it can do without asking a person first, and what a clean handoff to a human looks like when it should not act alone.
Chatbot Questions
Is this just a chatbot with a different name?
No — that distinction is the point of the page. A chatbot answers from a knowledge base and stops. What we build sits in the same chat window but is wired to your systems, so it can look up an order, update a record or file a request instead of just describing what you would need to do next.
Does it replace our existing live chat tool?
Not necessarily. It usually sits in front of your existing live chat, handling what it can and hands off to a human agent, in the same conversation thread, when a case needs one.
Can it recognise a returning, logged-in customer?
Yes, if your site or app already has an authentication or account system — the agent can use that context so it does not ask a customer to repeat information you already have.
What if a visitor asks something completely unrelated to our business?
It says so and redirects the conversation, rather than guessing an answer outside its scope. Staying inside a defined scope is part of the boundary design, not an afterthought.
Tell Us What Visitors Actually Ask
We’ll tell you plainly which of those questions can end in a finished action, and which still need a person.