E-Prosiding Persidangan Serantau Manasik Haji 2025 (MANASIK2025)

MANASIK2025 Putrajaya, Malaysia, 7-8 Oktober 2025 653 Second, the Design and Development phase is expected to produce a structured knowledge base built from authentic Islamic references and official Hajj guidelines. By combining Generative AI with a Large Language Model (LLM), the chatbot is expected to deliver accurate, interactive, and multilingual responses, making the learning process more engaging compared to traditional text- based resources. Figure 3. Preliminary results and expected outcomes Third, the Prototype Implementation aims to improve accessibility by supporting both mobile and web deployment, with added offline functionality. The expected outcome is that pilgrims will be able to practice and review ritual steps more conveniently, assisted by interactive features such as step- by-step guides, FAQs, and scenario-based help. Finally, the Evaluation Phase, once completed, is expected to show positive results in terms of user satisfaction, response accuracy, and cultural appropriateness. Feedback from prospective pilgrims and trainers will guide further refinements, particularly in areas such as voice-based support for elderly users and broader multilingual coverage. In summary, while the technical development is still ongoing, the preliminary analysis suggests that the proposed chatbot has strong potential to enhance Hajj and Umrah learning by improving accessibility, deepening understanding of rituals, and serving as a trusted source of Shariah-compliant guidance. A simple prototype was developed using Copilot Studio, supported by a knowledge base derived from Nota Kursus Asas Haji (KAH), Musim Haji 1444H (Lembaga Tabung Haji, 2023), to demonstrate the chatbot’s ability to support pilgrims in real ritual situations. For example, when a pilgrim forgot the number of tawaf rounds the chatbot guided them step by step, advising to assume the smaller number when unsure, complete the remaining rounds, and recall the conditions for a valid tawaf (see Figure 4).