Managing Information Volatility and Density in Academic and Vocational Guidance: Modeling a RAG-based Chatbot for Guidance Counselors in Morocco

In the Moroccan educational landscape, characterized by the rapid expansion of training programs and the extreme volatility of regulatory norms, guidance counselors are facing an increasing cognitive load. The real-time mastery of administrative information (ministerial memos, procedures, admission thresholds) often conflicts with the mental availability required for the counseling relationship. While generative Artificial Intelligence offers promising prospects for assistance, standard Large Language Models (LLMs) suffer from factual hallucinations and data obsolescence, making them unsuitable for critical professional use.

This article proposes a conceptual modeling of a mobile AI assistant specifically designed for guidance counselors. Through a comparative analysis methodology, we demonstrate the superiority of the Retrieval-Augmented Generation (RAG) architecture over Fine-Tuning for managing dynamic knowledge updates. The proposed model couples a generic LLM with a vector database of regulatory documents, thereby ensuring reliability, source traceability, and a significant reduction in the practitioner's extrinsic cognitive load. This work contributes to the Design Science Research (DSR) field by offering a robust architectural framework for the transition toward a human-AI symbiosis in educational guidance.