Ecommerce & Retail · Conversational Commerce
An AI system to automate customer service and sales across social messaging platforms, all powered by AI.
Outcome
Nearly 90% of WhatsApp and Instagram conversations closed without a human, first question to payment.
A direct-to-consumer fashion retailer in Kuwait built its business on personal connections, using WhatsApp and Instagram for sales and support. This high-touch model was central to their brand but created a natural limit to their growth.
The reliance on manual, one-to-one conversations meant operational costs scaled directly with customer volume. Support was confined to business hours, response quality was inconsistent across agents, and valuable customer data was lost in unstructured chat logs. The business could not grow without solving this operational bottleneck.
Robust Devs was brought in to design an automation layer that could handle the complexity of social commerce. We developed a central AI orchestrator that integrates directly with their existing tools, automating nearly 90% of customer interactions from product questions to final payment.
The challenge
How we solved it
Technologies
The orchestrator is the system's brain. It processes all incoming messages, manages the business logic, and routes requests to the correct internal or third-party service. This central-hub architecture ensures all interactions are consistent, logged, and easy to update, without being tied to any single front-end messaging platform.
When a conversation requires human intervention, the system does more than just transfer the chat. It analyzes the dialogue to determine the customer's goal, then packages the entire conversation history and a data summary for the agent. This allows for a fluid transition where the customer never has to repeat themselves.
How we built it
Built one orchestrator that owns the conversation, and taught it when to hand over to a person. The agent sits on a central API connected live to the WooCommerce database for stock, MyFatoorah for payment links and Respond.io for human support. Language models carry the conversation in Arabic and English, Whisper transcribes voice notes and a vision model identifies products from customer photos, so the same thread can move from a picture to a paid order. The part that protects the brand is the handoff: the orchestrator reads sentiment and flags what it cannot resolve, then passes the customer to a human with the full history and a summary of intent. That is how nearly 90% of conversations close without a person while the rest reach one properly.
In the client's words
“Most of our WhatsApp orders now close without a person. I still read the transcripts every week, and they hold up.”
Tell us what you're building. We'll show you how we'd approach it and what it would take to ship.