This research introduces a holistic innovation spanning process, product, and technology. It reimagines traditional IT decision-making by leveraging Large Language Models (LLMs) to create an intelligent, enterprise-aware decision-support system grounded in organizational architecture artifacts. At the process level, it integrates a mixture-of-experts approach with a multi-agent architecture, enabling coordinated reasoning among domain-specialized agents while managing conflicts during recommendation generation. The resulting product delivers not only high-level strategic advice but also traceable, evidence-based justifications from individual expert agents—a capability absent in current literature and practice. Technologically, the work advances practical LLM deployment through a synergistic combination of quantization, security hardening, and domain-specific fine-tuning via Retrieval-Augmented Generation (RAG), enabling secure, efficient, and on-premises operation within enterprise environments. Together, these contributions redefine how strategic technology decisions can be supported in complex organizations.