The product, one screen at a time
Stories are short visual walkthroughs of how the assistant behaves on a real store: what it retrieves, what it refuses, and what a shopper actually sees. Each is a handful of full-screen panels you tap through, rather than an article about one. They are built as plain HTML and CSS, with no tracking and no third-party script, and they work with JavaScript switched off.
The first stories being planned
None of these is published yet. Each goes up only once it is built from a real conversation. They are written in the same place as the notes, with no AMP runtime, and appear here at the next build.
-
A sale, start to finish
It will follow a shopper who asks for cooking oil for deep frying, is shown two products with their live prices, picks a size on the card, adds it to the cart and is handed to the store's checkout.
-
A question it refused
It will show a shopper asking about a product the store does not sell: the assistant says so, offers a person, and the question lands in the merchant's review queue with its trace.
-
A chat handed to a person
It will show a shopper asking for someone from the store: an operator takes over from the console, replies under their own name, and hands the chat back.
-
A wrong answer, diagnosed
It will show a merchant opening the trace behind a wrong price, seeing that the right page scored a little below the threshold, and fixing it with a verified answer.
-
In Tamil
The same sale, with the shopper writing in Tamil and the assistant answering in Tamil script.
Why stories rather than testimonials
We do not publish invented testimonials, customer counts or results. A story shows the product doing something specific, on screen, so you can judge it yourself. When a merchant is happy for their store to be shown, their story will say so and link to it.
How they are made
Each story is built from a real conversation on a real store, with the shopper's details removed. The panels show what the shopper saw and, where it helps, what the merchant saw beside it in the console: the trace, the review queue or the handover. Nothing is re-enacted or written for the camera. If a story shows the assistant getting something wrong before it is fixed, that is the point of it.
What each story shows
Every story ends on the evidence: the passage the answer came from, the line it had to clear, or the moment a person took over. That is the part a testimonial cannot show you, and the reason we would rather publish five short, specific stories than fifty lines of praise.
Until the first one is published
The live demo shows the same behaviour on a sample store. Ask it something and watch what it retrieved.
The thing the panels are about
A retrieval that worked, a question it refused, and the line between them. The trace panel shows all three on every real answer.