HKUST Business School Magazine

Use cases of AI in auditing and accounting Processing and standardizing information In traditional accounting, accountants typically engage in extensive information retrieval, sifting through vast search results to extract relevant content for benchmarking competitors or confirming accounting standards and audit methodologies. This process is then enhanced by accountants’ professional expertise and experience, enabling them to make informed judgments and distill valuable insights. AI has revolutionized how accountants access information and knowledge. By incorporating industry expertise, regulatory standards, policies, guidance and company-specific audit methodologies, auditors can create a proprietary knowledge graph for their enterprise. Accounting professionals can then interact with this knowledge graph, which serves as the professional’s “brain” through Large Language Models (LLMs). This natural language Q&A capability shortens the path from “question” to “answer,” minimizing retrieval time and allowing professionals to focus more on judgment and analysis. Generative AI, in particular, has transformed how accountants interact with information. Traditionally, data is categorized into structured formats like online tables and bank statements, as well as unstructured content such as receipts and contracts. Processing these various types of data requires multiple tools and presents challenges due to the large volume of structured data with inconsistent quality and the complexities of dealing with unstructured data in bulk. With generative AI, accountants can bridge the gap between different data types, extracting key information with unprecedented efficiency and uncovering hidden facts. For instance, AI can assist in data preprocessing by applying personalized methods for different data samples and automatically identifying and refining low-quality text through operations like string standardization and noise removal. 4 Moreover, generative AI can enhance data reconciliation processes, such as matching loan portfolios with corresponding payments and related data. By ensuring that payments align with loan agreements, AI accelerates processes and improves the end-to-end accuracy of transactional records, linking back to the legal contract. 5 Analyzing information A key function of the accounting and auditing profession is to detect and flag accounting issues, enhancing the quality of financial information used internally or presented to investors. However, AI is still not fully mature in performing quantitative financial analyses or calculations to identify accounting frauds. In accounting academic research, scholars have traditionally relied on linear models to predict fraud and identify anomalous accounting entries among large samples of firm-year observations. However, these linear models often face significant challenges regarding prediction quality. The advent of AI offers the potential to enhance the accuracy of fraud detection in empirical research. For instance, studies in bookkeeping have demonstrated that advancements in graph machine learning can effectively identify anomalous entries, aiding accounting professionals in detecting errors within the accounting system. 6 Additionally, research indicates that employing methods like random forests or gradient boosted regression trees can lead to significant improvements in fraud detection. 7, 8 Moreover, the use of automated machine learning also allows for real-time fraud detection, which would vastly improve the timeliness of fraud detection. 9 These academic insights suggest a substantial opportunity for AI to improve error detection in auditing and accounting tasks. Synthesizing information Finally, AI has the potential to revolutionize how information is synthesized in auditing and accounting. In these professions, presenting work often involves compiling scattered insights into a structured document while adhering to complex regulatory requirements. Generative AI can assist professionals in quickly establishing a starting point by decomposing tasks, systematically extracting information from various sources, and integrating it into a draft. By understanding context, AI aids in interpreting and compiling information, supporting professionals in documenting the process. Additionally, AI can function as a compliance checker, ensuring that documents align with industry regulations, company policies, and methodologies. It flags errors and omissions, providing references to guidelines that enable professionals to make necessary corrections and revisions. The advent of generative AI thus holds the potential to radically restructure workflows in the accounting and auditing professions, shifting from a copilot to an agent model and achieving a high level of human-machine integration. This transformation is primarily reflected in the reallocation of professionals’ time, reducing the effort spent on fact-gathering and documentation and allowing them to focus on framework design, review, and judgment, ultimately delivering better insights. Biz@HKUST 39