Over the past four decades, Sweden's education system has undergone a profound transformation, shifting from a centralised structure to a market-oriented model characterised by independent schools, deregulation, and competition. This paper introduces an innovative methodological approach to studying this transformation by applying computational text analysis with large language models (LLMs) to 45 years of parliamentary debates. By leveraging these methods and extensive parliamentary open data, we identify thematic patterns, ideological shifts, and policy discourses that have shaped the marketisation of Swedish education. Our methodological contribution lies in demonstrating how LLMs can be employed to scale up traditional discourse analysis, bridging the gap between computational methods and qualitative interpretative approaches. We engage critically with the challenges of algorithmic opacity, validation strategies, and interpretative transparency, addressing concerns about bias and the risks of black-boxed analyses. Combining machine-assisted text analysis with traditional qualitative methodologies, we present a scalable yet nuanced framework for studying education policy debates over time.