Lian Yun Gang City Audit Bureau Liu Jianbo

Abstract: This paper presents a comprehensive analysis and research on the history, development trends, and practical implementation of intelligent audit. It demonstrate the necessary of study of digital artificial intelligent audit. It list a brief history of the development in digital and intelligent audit, and show how the technology be implemented in this field. Finally, It concludes the great benefit of the of integrating of these digital and artificial intelligent audit tools.

Key words: Research Review of Digital Artificial Intelligent Audit, Artificial Intelligence,integrating of these digital and artificial intelligent

 I. Research Background of Digital Artificial Intelligent Audit

In February 2021, the Central Committee of Chinese Communist Party convened a conference on the “14th Five-Year Plan” (2021-2025), emphasizing that economic and social development during this period should be centered around promoting high-quality development. Since then, China’s economic and social development has entered an era of high-quality development. This new stage encompasses key elements from previous planning periods, including: Industry 3.0, supply-side structural reforms, and Internet Plus strategies, demonstrating a continuity with recent industrial development policies. This new stage is characterized by the extensive application of technologies such as big data, cloud computing, artificial intelligence, and automated control. It is defined by industrial digitalization and the rise of digital industrialization.

In 2024, both the Central Committees of the Communist Party and the Central Government consecutively released action plans for “AI Plus”, along with various levels of government subsequently implementing specific policies to support this initiative. This government-led accelerated the arrival of a digitalized era. Concurrently, national administrative management, enterprise management, and grassroots organizational management models have all undergone varying degrees of development and innovation in the direction of digitalization. As a crucial component of state governance capacity and possessing the important function of economic supervision, national audit acting as a multi-sensor system within the state apparatus, capable of rapid response. Mastering the analytical tools of digitalization is an essential pathway for auditors to enhance their capabilities and professional development.

At the beginning of 2025, General Secretary Xi Jinping emphasized that the audit department should safeguard high-quality economic and social development through rigorous oversight, making new contributions to the comprehensive advancement of national rejuvenation and the construction of a strong nation. This constitutes a clear positioning and directional guidance for audit work in the new era. In response to the imperative of effective governance, embracing digitalization as a means to address the challenges of digital governance represents the permanent mission and purpose of audit work. On an international scale, the significance of digital audit has become increasingly evident following recent USA Department of Government Efficiency audits focused on government efficiency and budget allocation. Digital information serves as the dynamic lifeline connecting macro and micro economic perspectives in contemporary society. Facing the realities of computer auditing, networked auditing, and big data auditing – three key digital auditing methodologies – auditors are confronted with the practical challenge of acquiring the necessary skills for effective digital intelligent audit.

2. A brief history of the development of digital and artificial intelligent audit

In our audit practice, the concept of digital artificial intelligence auditing originates from the integrated development of computer technology and is not a simple mixture of big data and artificial intelligence concepts. The process of human‘s pursuing modern electronic computing is also an exploration of simulation to human intelligent computing capabilities: the emergence of computer artificial intelligence has gradually evolved from inductive expert systems and algorithmic functions to artificial intelligence that integrates multiple algorithms and multi-layer coding and decoding. In specific electronic auditing processes, this evolution has progressed from specialized digital audit tools to networked database analysis audit, and then to the big data analysis audit proposed in recent years, marking a shift from dedicated data analysis tools to the use of general purpose analytical tools; From the perspective of database usage scope, it has evolved from standalone systems to local area networks and then to the internet; From the application scenarios of databases, it has progressed from ordinary relational databases to online transaction processing (OLTP, focusing on realtime transaction data processing), online analytical processing (OLAP, emphasizing multidimensional and historical data comparison), distributed data processing, and graph database processing.

In contrast, the development of intelligent auditing in finance and accounting must be closely integrated with the advancement of database technology. Intelligent auditing has evolved from RPA (Robotic Process Automation) audit that connects ERP, CRM, and other business systems to the era of digital intelligence. In this era, large language models serve as the foundational model for auxiliary analysis, forming a digital artificial intelligence audit system based on realtime databases (including vector databases) that operate in the form of separate tables, clusters, and clouds. This system uses message middleware (Message-oriented middleware, abbreviated as MOM) to obtain data from ERP, CRM, and other business systems in real time, enabling realtime analytical functions. Historically, databases and large language models (LLM) have always been intertwined. According to media reports, the USA Department of Government Efficiency (DOGE) large scale budget cutting activity involves dispatching four person data analysis and audit teams to various departments to conduct realtime audits of fiscal and business systems. This approach not only identifies issues but also ensures that national government operations remain unaffected, highlighting the significant role of digital artificial intelligence auditing technology.

3. Research review of the implement in digital and artificial intelligent audit

Because the concept of digital and intelligent auditing is still in its formative and developmental stages, there are different understandings of this concept and its application within the Chinese theoretical community. This study reviews the practical applications and theoretical developments of related concepts based on audit practices in both China and Western countries. Computerized accounting audit matured in the 1970s and 1980s, with EDP technology led by the Electronic Data Processing Auditors Association (Electronic Data Processing Institute) in the United States and CAAT technology (Computer-Assisted Auditing Techniques) in the United Kingdom being the pioneers. In the automation era, RPA (Robotic Process Automation) technology, as software that mimics human behavior, has been widely applied across various fields and gradually expanded into the economic field, leading to the formation of concepts and theories in the audit field. With the promotion of RPA applications and the maturation of specialized applications in the LLM domain of artificial intelligence, digital and artificial intelligent audit has entered the stage of theoretical development and industrial practice.

3.1. Early concept of RPA

In 2011, the Institute of Internal Auditors (IIA) defined technology based auditing techniques in the “Global Technology Audit Guide (GTAG) 16 Data Analysis Techniques” as “including all automated audit tools such as general audit software, test data generators, computerized audit programs, expert audit applications, and CAATs,” and emphasized training auditors to master highly automated audit routes. In 2016, McKinsey London interviewed Professor Leslie Willcocks from the London School of Economics, who described RPA in the economic domain as suitable for handling simpler audit tasks. It is primarily used for performing entity tasks that do not require additional knowledge, understanding, or insight, which can be accomplished through the compilation of rules, procedures, and instructions to terminal or software actions. Leslie Willcocks believes that future cognitive automation will be capable of processing natural language, reasoning, and judgment, establishing context, and possibly even understanding the meaning of comprehensive concepts and providing conclusion insights. Therefore, there is a significant difference between RPA and cognitive automation. The concept of cognitive automation is actually an early idea closely related to the current notion of intelligent auditing. The American Association of Chartered Accountants (AACA) reported that RPA has been applied to banking and mortgage apply Processes, sales Processes, general big data application processing programs (such as OCR optical character recognition, data extraction), and used in more and more different field audit. A Price WaterhouseCoopers (PwC) article in 2017 estimated that 45% of global labor tasks could be automated. In November 2019, Deloitte disclosed the widespread use of RPA in its financial audit processes. In January 2017, the International Institute of Internal Auditors (IIA) revised, the China Institute of Internal Auditors (CIIA) compiled and translated “the International Professional Practices Framework” (IPPF) , which confirmed audit automation in sections on audit concerns, control procedures and methods, and ongoing supervision. In 2019, the Chinese National Audit Office website reported that China Mobile introduced multiple models including OCR to implement automated audit, achieving full coverage of auditing.

3.2. Standardization of RPA

According to the 2018 CPA Journal report, K. Moffit, A.M. Rozario and M.A. Vasarhelyi proposed a roadmap for implementing RPA (“Robotic Process Automation for Auditing,” Accounting Emerging Technologies Journal). According to this roadmap, RPA implementation includes three main stages: (1) process understanding; (2) audit data standardization; and (3) execution of automated audit testing (such as some audit applications). The audit tasks that benefit most from RPA are those where processes are well-defined, repetitive, time-consuming, and do not require additional audit judgment.

In February of 2020 Vasant Raval and Erica Smith, “The Practical Aspect: Organizational RPA Adoption and Internal Auditing” was published in the ISACA (Information Systems Audit and Control Association) e-journal, highlighting that if business processes are mature, there is an opportunity to transform these processes into more advanced RPA workflows, thereby transferring some or most manual tasks to computer processing. In November 2020 Hassan Toor, “Robotic Process Automation for Internal Audit” was published in the ISACA e-journal, emphasizing that RPA can assist internal auditors in generating and standardizing data for custom analysis, automatically initializing data acquisition and categorization for annual risk assessment processes, testing and verifying the detailed structure of fields from different sources of data, and automatically controlling test automation procedures. 

3.3. The rising of digital and intelligent auditing

In March 2023, Bloomberg Shijie Wu, Ozan Irsoy, and Steven published a paper on LuBloombergGPT as a specialized training LLM for finance, marking the first step in the development and application of NLP technology in the financial sector. The model uses a 70-layer transformer decoder structure and has been trained on 1.3 million hours of GPU time at 40 GB A100 GPU. It leverages Bloomberg’s 40 years of accumulated financial information, data, files, and other web-based datasets for training. This model has already been applied to Bloomberg’s existing financial NLP tasks, such as market sentiment analysis, named entity recognition, news classification, and question answering.

As the call to open up the AI industry was sounded across China in 2024, by March 2025, according to incomplete statistics, there were 245 large language models of China publicly available on the internet, among which 23 were related to finance and accounting auditing industries. Based on public information, the main types of model architectures include Llama, Llama3, and Bloom. Deepseek V3, an open-source project, has shared some training methods and code. In their technical paper, they pointed out that inspired by low-precision training, they developed a hybrid precision training method using FP8 format for most data and using original precision format for key data. They also optimized the transformer structure, achieving a good balance between efficiency and data accuracy. By early 2025, it emerged as relatively advanced in certain fields, gaining widespread use in government systems. When developed with middle applications, it can achieve OCR functions, process basic financial statements, and perform data visualization. There are reports indicating its application used in assistance in legal classification analysis for frontline audit. Hundsun Electronics, specializing in financial services, developed a text-to-SQL model called LightGPT based on ChatGPT in July 2023. This application achieved an execution accuracy rate of 82.3% on the Spider holdout data set, becoming the most advanced Zero-Shot Learning on text-to-SQL method (C3) in the Spider challenge. The paper describes this feature as consisting of three key components: Clear Prompting (CP), Calibration with Hints (CH) and Consistent Output (CO) correspond to model input, model deviation and model output respectively. It provides a systematic solution for Zero-Shot Learning on text-to-SQL with prompt engineering as the starting point.

  • Conclusion of the study of digital and intelligent auditing

So far, the current large language model (LLM) technology has a feasible route for preliminary financial data analysis, compliance review, knowledge base, question answering, assistance in legal classification analysis, and assistance in formulating audit conclusions and recommendations. Considering the RPA technology used in some banking、finance workflow prior to the LLM era. It can be seen the great benefit of integrating of these technology tools in the future.

References:

  •  Altus J. Lambrechts, Jacques E. Lourens, Peter B. Millar,Donald E. Sparks.Global Technology Audit Guide (GTAG®) 16 Data Analysis Technologies [EB/OL]. (2011-08)[2025-03-01].https://www.theiia.org/globalassets/documents/content/articles/guidance/gtag/gtag-data-analysis-technologies/gtag16cx2.pdf.
  • Xavier Lhuer.The next acronym you need to know about: RPA (robotic process automation)[EB/OL].(2016-12-06)[2025-03-01].https://www.mckinsey.com/capabilities/mckinsey-digital/our-insights/the-next-acronym-you-need-to-know-about-rpa.
  • American Association of Chartered Accountants (AACA).Understanding Robotic Process Automation in Audits [EB/OL].[2025-03-01].http://www.usicpa.org/Item/Show.asp?m=1&d=3153..
  • Lianyungang Audit Bureau Research Group. Report on the Application of Big Data Technology in the Field of Livelihood Audit by Jiangsu Provincial Audit Office (2021JSSJY005) [R]. Nanjing: Jiangsu Provincial Audit Office (Lianyungang Audit Bureau), 67-72.
  • Liu Jianbo. Development and Implementation of Audit Management Information System for Municipal Government Investment Projects in L City [D]. Business School, Hohai University, 2015.
  • Institute of Internal Auditors International. International Professional Standards for the Practice of Internal Auditing [EB/OL].(2017-01)[2025-03-01].https://www.theiia.org/globalassets/site/standards/mandatory-guidance/ippf/2017/ippf-standards-2017-chinese-simplified.pdf.
  •  Miklos A. Vasarhelyi,Andrea M. Rozario.How Robotic Process Automation Is Transforming Accounting and Auditing [EB/OL].(2018-07-02)[2025-03-01].https://www.cpajournal.com/2018/07/02/how-robotic-process-automation-is-transforming-accounting-and-auditing/.
  • Vasant Raval,Erica Smith.The Practical Aspect: Organizational RPA Adoption and Internal Auditing[EB/OL].(2020-02-28)[2025-03-01].https://www.isaca.org/resources/isaca-journal/issues/2020/volume-2/organizational-rpa-adoption-and-internal-auditing.
  • Hassan Toor.Robotic Process Automation for Internal Audit[EB/OL].(2020-11-11)[2025-03-01].https://www.isaca.org/resources/isaca-journal/issues/2020/volume-6/robotic-process-automation-for-internal-audit.