Every chat is a checkout.

Chidi is commerce that lives in the conversation. Sellers run their whole business, catalog, orders, payments, and customers, inside the chats they already use: WhatsApp and Telegram. Nothing moves to a dashboard. The thread the buyer is already in becomes the system of record. I was founding designer across brand, product, and the front-end, which I built in code. That span mattered, because a conversational product fails the moment its voice and its mechanics disagree. The brand had to sound like a person and the system had to behave like infrastructure, and one designer held both.

Millions of sellers already run their business in DMs. It works until it does not: messages get lost in a flooded inbox, prices are retyped a hundred times a day, orders live in someone’s memory, and there is no record when money goes missing. The seller is fast at five orders and drowning at fifty. The job was to keep the part that works, the chat, and quietly put a system behind it. Most tools answer this by bolting a CRM onto the side, a second place to keep clean that the seller never opens. So the constraint I set was harsh: if a feature needed the seller to leave the conversation, it did not ship.

Every competitor tried to pull sellers onto a brand-new app, a dashboard to log into and abandon. The insight was the opposite: the conversation *is* the storefront. So Chidi meets buyers where they already are and keeps one commerce engine behind every channel. WhatsApp, Telegram, and a shareable web link are just surfaces onto the same business.

The heart of the product is a conversation that quietly becomes a checkout. A buyer asks “got this in blue?” and Chidi’s AI drafts the reply with the right product, price, and tone, then the seller approves or edits before it sends. The AI never speaks for the seller on its own. It removes the typing, not the judgement. That draft-then-approve step was a deliberate line. Full automation reads as a bot the moment it gets one detail wrong, and trust never recovers. An approved reply drops a product card into the thread, the buyer pays through an in-chat link, and a branded receipt lands. Customer memory carries names, past orders, and sizes between conversations, so a returning buyer is greeted, not re-interrogated.

The core decision: build a slick standalone app, or disappear into the channel the seller already trusts? A standalone app is a clean canvas, but it is also one more login to forget. Meeting sellers in the chat meant accepting the constraints of someone else’s interface, and that constraint became the product’s edge: zero behaviour change to adopt it.

That choice set the brand brief too. Inside someone else’s chat, Chidi had to feel like a trusted person, warm and human, yet confident enough to put a receipt beside real money. The branded receipt is the trust moment. When a payment clears, the buyer gets a clean confirmation with the seller’s name, the amount, and a paid marker, formatted like proof rather than a chat bubble. It is the artefact a buyer screenshots. So it leans on the mono numeral face and a steady green paid state, the register a bank uses, while the rest of the brand stays warm in Maple. Voice notes stay first-class too, because in these markets sellers reply by talking.

A commerce system sellers adopted without changing how they already work. The conversation stayed the same, the business underneath got real. 12,000 sellers onboarded across two live channels, with in-chat payments and branded receipts closing the loop where the buyer was.