A Bilingual Retrieval Augmented Generative AI Chatbot Architecture for Inclusive Retail Banking in Egypt

Abstract

This paper presents Cash Compass, a bilingual Modern Standard Arabic - English generative AI chatbot designed to advance ethical financial inclusion in Egyptian retail banking. The system provides real-time, context-aware guidance on deposits, credit, cards, savings, and investment products while scaffolding financial literacy for underserved users. Ethical design principles are embedded across privacy, fairness, transparency, accountability, accessibility, and user protection. The modular architecture integrates intent and entity parsing, a curated banking knowledge base, retrievalaugmented generation, compliance filtering, and adaptive dialogue management optimized for literacy and bandwidth constraints. Evaluation demonstrates improved accuracy, reduced latency, enhanced bilingual clarity, and increased user satisfaction, indicating that Cash Compass can enable trustworthy, inclusive, and scalable digital financial advisory in emerging markets.