RoboYogi Chat Order

Inspiration

KFC Vietnam's customers already live in messaging apps — but placing an order means leaving the conversation and switching to a separate app or website. That switch is where orders die. The problem statement is simple and painful: customers engage through chat, ordering happens elsewhere. We built the bridge so the conversation itself becomes the ordering channel.

What it does

RoboYogi Chat Order is a conversational ordering agent for KFC Vietnam. Inside the messaging app the customer already uses, it presents the menu, answers questions about items and combos, handles customization and quantity, confirms the order summary and total, and completes the order — without the customer ever leaving the chat.

How we built it

The agent reuses the conversational-commerce engine we built for our Retail track sales agent, retargeted to the QSR ordering flow. It connects to menu, pricing, and promotion data so item availability and combo pricing stay accurate, manages a structured order state through the conversation (items, modifications, quantities), and produces a confirmed order summary before handoff to fulfilment. Ordering intent, clarification, and upsell suggestions are handled conversationally, but the order object itself is always structured and validated — the model never freehands prices or totals.

Challenges we ran into

Order accuracy in free-form chat. A conversational interface invites ambiguity — "make that two, no ice, the big one" — so the hardest work was mapping messy natural language reliably onto a structured, priced order object, and always confirming the exact summary with the customer before completion. We also had to keep upsell suggestions helpful rather than pushy, since chat is a more personal channel than a kiosk screen.

Accomplishments that we're proud of

One conversational-commerce core now genuinely powers two different industry applications — electronics retail product discovery and QSR food ordering — with the domain layer swapped, not the whole build. That's honest reuse, not copy-paste with a new logo.

What we learned

Chat-based ordering lives or dies on confirmation discipline. Customers forgive a clarifying question; they don't forgive a wrong order. Designing the agent to be conversational at the front and rigidly structured at the back was the key lesson.

What's next for RoboYogi Chat Order

Supporting order tracking and reorder-my-usual flows in the same conversation, and extending the same engine to KFC's kiosk recommendation use case.

What we learned

Built With

• node.js
• openai