The challenge
The client's customer-service team was overwhelmed by a steady stream of repetitive WhatsApp and web inquiries — mostly about store hours, returns, gift cards, and order tracking. Response times had crept past four hours during peak season, and CSAT was slipping. Their existing rule-based bot couldn't handle Arabic well and routinely escalated trivial questions.
The solution
We designed a RAG-grounded GPT assistant with the client's full knowledge base — store policies, product catalogue, FAQs — embedded as searchable context. The bot greets in the user's language automatically, handles 80+ common intents, and hands off to a human with full conversation context when it can't help. It writes back to the CRM and Slack so the human team always sees the latest state.
The outcome
Within eight weeks the bot was deflecting 62% of incoming tickets, the average response time dropped from 4h to under 30s, and CSAT lifted by 14 points. The customer-service team now focuses on complex returns and high-spend customers — work that actually moves the business.