← Home

An AI sales assistant for shops with large catalogues

An assistant that answers customer questions on WhatsApp and the shop’s website, finds the right product in the shop’s own catalogue, and hands the chat to staff whenever it is unsure.

Status

In development. Next: 30-day pilots with specialty retailers in Kerala. We are starting with a small number of shops, and will publish results only once pilots have produced them.

The problem

Shops that sell car care, spare parts, electrical or hardware items carry hundreds or thousands of products. Customers rarely know a part number. They ask in everyday words, in Manglish, as a Malayalam voice note or with a photo of a worn label, often in the evening or during a counter rush.

Most WhatsApp tools are built around keyword menus and broadcasts. Website chatbots stay on the website, and some answer “we don’t have it” for products the shop does stock but has run out of. Every one of those answers is a lost sale.

Rules it follows

Never says “we don’t have it” for a listed product

Out-of-stock items stay searchable. The assistant says the shop stocks it but it is out right now, offers the closest in-stock alternative, and offers a back-in-stock message.

Exact words and meaning

Words like “matte”, “ceramic” or a part number must match exactly, not only by meaning. Each shop gets its own list of trade terms and local synonyms during setup.

Prices and stock come from the shop

They are read from the catalogue at reply time. Discounts and offers come only from the shop’s own data.

Replies the way the customer wrote

Manglish in, Manglish out. Malayalam script in, Malayalam script out.

Hands off instead of guessing

When it is unsure, the customer asks for a person, or the order is large, the chat goes to staff with a summary attached.

Stays on the shop’s business

It answers about the shop’s products, orders and policies, ignores instructions hidden in customer messages, and politely declines anything else.

Working today

  • Receives WhatsApp messages and checks they really come from Meta
  • Queues each message once, even when WhatsApp sends it again
  • Matches each message to the right shop and keeps the conversation
  • Catalogue database with exact-word and meaning search, out-of-stock items included

Building next

  • Catalogue import from spreadsheets and WooCommerce
  • AI answers, in the customer’s own language and script
  • Malayalam and mixed-language voice notes
  • Handoff to shop staff with a summary
  • A test set of real questions for each shop, run before every change

Data and WhatsApp

Each shop connects its own WhatsApp Business number, and we act as its solution provider. The assistant answers only about that shop’s products, orders and policies.

WhatsApp conversations are never used to train or improve shared AI models. Customer data and AI processing stay in India.

Run a shop with a large catalogue?

Tell us what you sell and how customers reach you today. We will tell you honestly whether a pilot makes sense for your shop.

Ask about a pilot