Can Generative AI Support Course Design, Constructive Alignment, and Assessment in French for Specific Purposes?

This empirical study investigates whether locally executed artificial intelligence systems, augmented with Retrieval-Augmented Generation (RAG), can reliably support instructors in maintaining constructive alignment across course design and assessment in resource-constrained institutional settings. The research addresses two interconnected questions: (1) To what extent can a domain-specialized language model (HorizaEdu) running on open-source infrastructure (Ollama) with pedagogical ingest generate aligned course materials and assessments without cloud dependency or data loss? (2) Can non-technical educators use generative AI prototyping platforms (Google AI Studio) to design, implement, and deploy secure web-based examination systems? The study conducted a single-cohort empirical evaluation with 150 post-secondary learners in a Professional French (Français Professionnel) blended course targeting an engineering communication context. Using Porter’s Alignment Index, RAGAS automated evaluation, the System Usability Scale (SUS), and server-side infrastructure monitoring, we measured pedagogical alignment, algorithmic performance, technical stability, user experience, and legal compliance. Results show that the local RAG-augmented system achieved a mean alignment index of 0.88 (target: 0.80), with 92% Faithfulness and 96% Answer Relevance scores. The no-code web application maintained 100% uptime under full load and achieved a SUS score of 74.5 (Grade B / “Good”). The study shows that with careful pedagogical scaffolding, open-source AI infrastructure, and structured prompt engineering, individual instructors in modest institutional settings can operationalize constructive alignment as an automated design constraint while independently building secure assessment platforms. Limitations include single-cohort design, domain specificity, and unresolved mobile UX challenges. The work helps bridge the gap between AI capability and equitable teacher access to AI-augmented instructional design tools.