Enterprise AI Support Chatbot for DIFC FinTech

The Challenge

Our client — a regulated financial services company operating in Dubai's DIFC — was handling hundreds of repetitive customer support queries daily across email, WhatsApp, and their website. Their support team was stretched thin answering the same questions about products, onboarding steps, compliance requirements, and account management. They needed an intelligent, compliant solution that could handle queries in both Arabic and English without compromising accuracy or tone.

Our Solution

CodeOrAI designed and built a Retrieval-Augmented Generation (RAG) chatbot powered by GPT-4o, grounded entirely in the client's own documentation — product guides, compliance FAQs, onboarding flows, and internal knowledge base articles. Key capabilities included:

• Semantic search across 200+ internal documents using vector embeddings
• Strict context grounding — the model only answers from verified client content
• Arabic and English bilingual responses with automatic language detection
• WhatsApp Business API integration for mobile-first customer engagement
• Embedded web widget for the client's customer portal
• Escalation routing — seamlessly hands off to human agents when confidence is low
• Admin dashboard for monitoring conversations, flagging gaps, and updating knowledge base

Technology Stack

Compliance & Data Privacy

Operating within DIFC's regulatory environment required careful attention to data handling. All data is processed and stored within Azure UAE North, ensuring UAE data residency compliance. The system logs no personally identifiable information from users by default, with full audit trails available for compliance review. The knowledge base ingestion pipeline includes a human review step before any document is made available to the model.

Results

Within eight weeks of deployment, the chatbot was handling 60% of all inbound support queries without human intervention. Average first-response time dropped from 4 hours to under 10 seconds. The support team refocused on complex, high-value client interactions — improving overall service quality. User satisfaction scores collected via post-chat surveys averaged 4.7 out of 5.

The system now processes thousands of conversations per month in both Arabic and English, with the client independently maintaining their knowledge base through the admin dashboard CodeOrAI built alongside the chatbot.