人工智能驱动的审计效能提升：基于审计监督职能数字化转型的实证探索

AI-Driven Audit Effectiveness Enhancement: An Empirical Exploration Based on the Digital Transformation of Audit Supervision Functions

摘要: 在数字经济的大背景之下，财务数据的规模和复杂性不断增大，传统的审计方式以抽样核查、人工判断为主，在审计覆盖面、风险识别及时性、资源配置效率等方面已经逐渐暴露出不足。文章从审计监督职能的角度出发，探讨人工智能嵌入审计流程之后对审计效能产生的影响机制。研究采用案例型准实证方法，选择2家已部署人工智能审计工具的大型企业内部审计部门作为研究对象，采集2022年到2024年三年期间36个财务审计项目中的相关数据，从审计周期、样本覆盖率、异常识别准确率、人工工时占比等几个方面展开分析。经过研究得知，人工智能依靠全样本分析、智能风险评估、异常识别等手段，大大提高了审计效率和风险识别能力，但其效能发挥有赖于审计人员的职业判断和人机协作来实现。基于审计学视角，进一步分析人工智能的应用边界问题，认为其只能起到辅助决策作用，审计的最终结论要经人工复核才可被当作审计依据。研究结果为审计智能化规范应用、审计准则适应完善提供了经验证据和理论参考。

Abstract: Against the backdrop of the digital economy, the scale and complexity of financial data continue to expand. Traditional auditing methods, which primarily rely on sampling checks and manual judgment, have gradually revealed limitations in audit coverage, timeliness of risk identification, and resource allocation efficiency. From the perspective of audit supervision functions, this paper investigates the mechanisms through which the integration of artificial intelligence into audit processes impacts audit effectiveness. Employing a case-based quasi-experimental research design, this study selects the internal audit departments of two large enterprises that have adopted AI audit tools. Data were collected from 36 financial audit projects conducted over the three-year period from 2022 to 2024, examining aspects such as audit cycle duration, sample coverage rate, anomaly identification accuracy, and proportion of manual working hours. The findings indicate that artificial intelligence, through full-sample analysis, intelligent risk assessment, and anomaly detection, significantly enhances audit efficiency and risk identification capabilities. However, its effectiveness relies on auditors’ professional judgment and human-machine collaboration. From an auditing perspective, this paper analyzes the boundaries of AI application, concluding that AI should only serve as a decision aid; final conclusions must undergo manual review before being admissible as audit evidence. The research results provide empirical evidence and theoretical foundations for the standardized application of intelligent auditing and the adaptive improvement of auditing standards.