Case study · AI shop assistant
Floora
Most website chat bots either say nothing useful or say something the owner would never sign. Floora is my answer: a shop assistant with a face. She greets visitors, answers their questions from the business’s own product list and guides, and hands anyone ready to buy to the staff. One rule is built in, not just asked for: she never states a price. Here she works as a flooring guide; for another business she gets a new name, look, voice and specialty.
Shown in her first job, as a flooring guide. The store’s name, phone number and web address are blurred.
How she works
Answers from the business, not the internet
Floora looks things up in the business’s own catalog and a set of short guides written for it. When nothing matches, she says so instead of guessing.
Never a wrong price
A price quoted by a chat bot is a promise the owner never made. Floora describes products as budget-friendly, mid-range or premium, and a guard checks every reply before it is shown: it blocks any price figure, any cost wording, and any product name she was not given.
Tested before every release
Before each change goes live, she is run against 98 tricky questions written to make her slip, and every reply is checked for leaks.
Leads go straight to the staff
When a visitor is ready, Floora offers the next step, such as a free estimate. The request reaches the business by email, marked as a Floora lead, and flagged when it arrives after hours.
A face, not a pop-up
On a computer she stands at the corner of the page, welcomes visitors, shifts her stance now and then, and opens the chat when clicked. On a phone she rides on the Ask Floora button, and the chat fills the screen. Her name, look and poses are made for each business.
Costs kept small
She runs on a fast, inexpensive model with a daily spending cap, and her answers to the quick buttons are saved for a day so the same question is not paid for twice.
More of Floora
Her poses, made for each business
Quick buttons for the questions people ask most
She asks before she recommends
On a phone
The chat
An answer
Built with Next.js on Vercel, with Claude for the answers, a Turso database for saved answers and usage, Resend for lead emails, and Veo for her video clips. The persona, poses and clips are generated per business.
Designed and developed by Montazeri AI Studio