Conversational commerce · Delivered

Customers buy by sending a photo

An auto parts manufacturer was losing sales because customers could not find the right part. They did not know part names, spent a long time comparing options, and abandoned carts. We built a WhatsApp bot that knows the catalogue, advises like a mechanic, and writes the order into the ERP.

1

channel the customer already has open

2

ways to search: a photo, or their own words

0

orders re-typed into the ERP by hand

Plain paper carrier bags in a row of colours
The problem we solved

The buyer knew the symptom, not the part number

Our automotive parts client was losing sales because customers couldn't find the right parts. They didn't know part names, spent hours comparing options, and abandoned carts frequently.

Before

  • Customers could not find the right part on the website
  • A complex buying process, and carts abandoned inside it
  • No expert guidance — the customer had to be sure before ordering
  • Staff entering every order into the ERP by hand

After

  • Image search: a photo of the part is the query
  • A conversation on WhatsApp, the channel they already use
  • A mechanic agent that diagnoses the problem and explains the choice
  • Confirmed orders written straight into the ERP
What we built

As easy as asking the person behind the counter

Every capability exists so a customer who cannot name what they need can still buy the right thing.

Natural language shopping

Customers describe what they need in their own words. The system understands intent and context, so no part numbers are needed.

Image-based search

Send a photo of any part and get matches back. Vision identifies parts even from damaged or dirty components.

Smart product comparison

Products compared on reviews, specs and use cases, with a recommendation for the customer's specific need.

Expert mechanic agent

Diagnoses the car problem from the customer's description, recommends the right part, and explains why it is the right one.

Real-time ERP integration

Live inventory, instant orders and automatic invoicing — the conversation writes into the system of record rather than into a spreadsheet.

The connection

A conversation that writes into the system of record

Live inventory, instant orders, automatic invoicing.

A chatbot that cannot see stock will eventually sell something the warehouse does not have. This one reads availability at the moment it answers, and the confirmed order lands in the ERP through the same path the business already trusts.

The stack

Built with

WhatsApp Business APIGPT-4 VisionLangChainPineconeFastAPIPostgreSQLRedis

Bring the photo a customer sent you

A working session on your own catalogue: send a real customer message — the blurry photo, the description with no part number — and watch it become a matched product with a price and a route into your ERP.
Our second practice

This is our AI engineering practice

It is real work and it is where our four products came from. But what Cognilium leads with is narrower: optimization apps that run in tandem with Microsoft Dynamics 365, computing the decisions the ERP records but does not derive — the optimal price, the optimal pick path, the optimal stock level. See the optimization apps · How we build inside the ERP.