An online store with WhatsApp ordering and no commissions

The client’s problem

Small restaurants and shops in Latin America sell through WhatsApp, but they do it all by hand: the customer asks what’s available, the owner answers one message at a time, jots the order down on paper or in a spreadsheet, and sends a message when it’s on its way. Nothing gets recorded, there are no numbers to show what sells, and in countries like Venezuela prices also have to be converted between bolívares and dollars every day.

The alternative was delivery apps, which charge up to 30% commission per order. For a small business, that eats the profit.

How it works

Each business gets its own store at an address like store.pideai.com, with its own catalog, prices and payment methods.

Here’s how an order travels:

1. The customer builds the order from the catalog on their phone, with nothing to download.
2. When they confirm, the order reaches the owner instantly, with a sound alert on their dashboard. If the business prefers, WhatsApp also opens with the order already written out.
3. Every time the owner updates the status (being prepared, on the way, delivered), the customer gets an automatic WhatsApp message.
4. The delivery driver has their own screen to see the delivery and log it with a photo.

It works for table orders, delivery and pickup. The official exchange rate updates on its own, and an AI feature touches up product photos so the catalog looks professional without hiring a photographer.

The database keeps each store’s information separate with security rules, so no business can see another’s orders. Everything runs on our own servers, with three copies of the system running at once so it stays up.

The tough call and why

A flat subscription instead of a commission. Charging a percentage per order is the delivery-app model and the one that makes the most money in the short run. We chose a flat subscription because the commission was exactly what was hurting these businesses: with a fixed price, every extra order is profit for them, not for the platform.

Direct AI, no middleman. The photo enhancement started out using AI through the platform the first prototype was built on. We switched to connecting directly with the model provider: it’s 30% to 50% cheaper, and the product no longer depends on a third-party tool.

What broke in production

• WhatsApp updates weren’t going out and nobody knew. The messaging service configuration had placeholder values instead of the real ones, and the error was being swallowed without a trace. We fixed the configuration and added logging so a failure like that shows up right away.
• Some stores wouldn’t load. Two database functions were in the code but had never been installed on the production server. We installed them and added a check so the code and the database don’t drift out of sync again.
• The cart merged items with add-ons. If a customer added the same pizza twice with different toppings, the system lumped them together. We fixed it so each combination is its own item.
• Tracking links exposed internal data. The WhatsApp messages included the order’s internal number. We replaced it with a tracking code that doesn’t reveal anything about the system.

Results

It’s live at pideai.com. According to the database as of September 6, 2026:

• 1,814 orders processed: 983 table orders, 442 deliveries and 389 pickups.
• 131 stores created, 55 of them food businesses.
• 1,714 WhatsApp messages sent automatically to customers.
• 151 product photos enhanced with AI.