The problem wasn't that Panda didn't have customer support. It's that the support they had wasn't working the way customers actually shop.
When someone opened the Panda app and searched for "chicken," they got chicken. But when they typed "I want to make a chicken biryani for six people," the app had no idea what to do. The intent was there. The catalog was there. The connection between the two wasn't.
At the same time, customer support flowed through a call center and a generic web form with no triage, no self-service, and no way for a customer to get a fast answer without waiting for a human. Millions of users, thousands of daily interactions; all of it handled the hard way.
The app worked, the experience didn't.
We built an AI conversational layer that lives inside the Panda mobile app, on both Android and iOS, and handles four things through one system.
Natural-language product search in Saudi Arabic dialect. Recipe-to-basket conversion: a customer describes a meal and the agent assembles a shopping basket from real-time catalog availability. A support assistant that answers questions from Panda's own knowledge base without routing to a human. And a smart escalation flow that hands off to a live agent through Genesys Cloud CX when the AI isn't confident enough to answer.
The system was built so the catalog updates daily without affecting live conversations. Products refresh in the background. The chat never waits.
Eight weeks, from start to integrated.
The build ran for eight weeks. Backend in Python, AI via Google Gemini 2.5 Flash with multilingual reranking, vector retrieval targeting under 200 milliseconds, and custom Kotlin and Swift SDKs wrapping a single Unified Chat API for both platforms.
The architecture was designed to extend. The same AI brain can be connected to voice channels and the e-commerce site without being rebuilt, only the channel changes.
Where it stands
The agent is live and integrated. The engagement is in an iteration phase with Panda's mobile and operations teams. Hard adoption metrics, containment rate, basket conversion from recipe suggestions, are being tracked on Panda's side and will be published once the post-integration review is complete.
“A customer can describe a meal in Arabic and leave with a full basket, without talking to anyone"
- Natural-language search in Saudi Arabic dialect, across a live product catalog
- Recipe-to-basket conversion from a single text message
- Customer support answered from Panda's own knowledge base
- Smart handoff to a human agent when confidence is low
- Built and integrated across Android and iOS in 8 weeks
