Cloud-kitchen operations with an AI ordering line

The challenge

Website, aggregator and phone orders arrived separately and were re-keyed by kitchen staff. Sold-out items stayed available until someone updated each channel. Roughly one in four peak-time calls went unanswered, and disconnected cost records obscured the margin of each brand and kitchen.

Our approach

Orders enter one model and route by address to the kitchen with coverage and capacity, then to its station. Prep timers coordinate driver dispatch. A photo at the pass records the packed order. The driver app batches trips, captures delivery evidence and reconciles cash. The customer app shares the website codebase through a webview, with native push, location and payment bridges. The ERP connects channel pricing, recipes, stock, purchasing and daily profit by brand and kitchen. We added the Gulf Arabic and English phone agent after live menus were established. It identifies returning callers, reads orders back and sends KNET links. Complaints, allergies and uncertainty reach a person with the basket already built. We evaluated it against 1,100 recorded calls before it took live orders.

Outcome

• Stock changes reach the website, app, phone line and three aggregators within a minute
• AI completes 68% of phone orders. Exceptions reach a person, and peak missed calls fell from one in four to under one in fifty
• Kitchen readiness and driver dispatch coordinated, reducing median order-to-door time by seven minutes
• Daily margin by brand, kitchen and channel, including supplier costs and aggregator commission