In recent years, the rise of big data has popularized data-driven decision-making. However, the interpretability shortcomings of artificial intelligence (AI) models limit their reliability for critical decisions. This paper proposes a paradigm shift from conventional data-driven to causal model-driven decision-making, leveraging advancements in causal inference and large language models (LLMs). By applying the capabilities of generative AI, this paradigm shift enables the fusion of extensive domain knowledge, facilitating the development of causal models that capture the complexity of real-world systems and problems. The model-driven approach provides a better understanding of the causal mechanisms, relationships, and dynamics compared to correlational data-driven methods. Moreover, we introduce the concept of a composable business model based on modular causal components. We present a methodology for constructing robust causal model-driven decision frameworks, emphasizing the comprehensive utilization of generative AI to incorporate domain knowledge with causal inferences. Through an in-depth analysis of case studies across multiple domains, this model-driven approach empirically showcases its potential to improve decision quality, optimize resource allocation, and enhance process efficiency. Additionally, we critically discuss future research directions and challenges in the evolutionary trajectory of causal model-driven decision-making.