This study focuses on how the digitalization of key audit matters can provide high-quality audit signal data for AI-assisted decision-making. AI-assisted decision-making is divided into two functions: “machine identification” and “assisted judgment”. High-quality audit signal data should possess both signal sufficiency and semantic consistency in order to effectively support these functions. Accordingly, the study follows the logic of “data point model construction, implementation path development, and audit supervision application expansion”. First, it constructs a data point model for key audit matters, decomposes the fully disclosed key audit matter text into 15 data points, and clarifies their semantic boundaries. Second, it proposes a demand-oriented semantic structuring path and advances digitalization at the institutional, standard-setting, and reporting levels. Finally, it points out that the digitalization of key audit matters can promote the transformation of audit supervision from subject-based collaboration to data-driven subject collaboration. Within China’s independent knowledge system, key audit matters can be understood as high-quality data resources in public governance. The digital transformation framework centered on data point modeling, semantic standardization, digital templates, and XBRL mapping can also provide reference for the digitalization of unstructured texts in fields such as finance, financial reporting, and accounting.