Implementation of Artificial Intelligence and its Impact on Human Auditors


This research paper has an objective to answer the question of possibility of auditors being replaced by Artificial Intelligence (AI). Emergence and rapid development of technologies lately possibly disrupts many fields of work, including auditing. Technology such as generalized audit software has been used in current day audit, and AI advancement can also be utilized in the process. Early functions of AI are already adopted in big audit firms, and possibilities of implementing deep learning and AI-based automation in audit in the future exist. This paper supports the idea that human auditors are irreplaceable by AI due to the lack of human-like professional judgement and skepticism. Current and future auditors as well as students should however still prepare for the technological breakthrough of AI by improving their IT proficiency and database knowledge.

Chapter 1 Introduction

In the recent years, newly emerging sophisticated technologies such as artificial intelligence has been slowly implemented into many lines of work, and that also includes financial auditing. The possibility of machine that can think and learn like a human being raises a question of whether auditors can will be replaced by artificial intelligence or not. But technology is already a part of current day auditing, as it is not rare for auditors to work alongside systems or software that can assist audit procedures.

1.1. Current Role of Auditors and Technology

The purpose of audit is to provide reasonable assurance that a company’s financial statements are fairly presented, hence the method of sampling selection is done with an assumption that those samples will reflect the financial statements as a whole when tested (International Auditing and Assurance Standards Board, 2009). Currently, selection of computerized data can be done with the used of generalized auditing software, shortened as GAS, with some examples being programs such as Audit Command Language (ACL), IDEA, and ProAudit. According to Debreceny (2005), other than sample selection, GAS is utilized to detect misstatements and is also able to thoroughly analyze a company’s financial data in their accounting system.

Although GAS can do various functions to assist audit procedures, it still needs to be operated by human auditors. According to Ahmi (2013), functions of GAS can fulfil requirements in auditing standards issued by American Institute of Certified Public Accountants (AICPA) and Auditing Practices Board from United Kingdoms. However, the use of GAS, according to Janvrin (2009), requires consideration of other factors to determine whether it is applicable or not. Janvrin (2009) mentioned that if a client has complex IT structure and the assessed control risk is low, which indicates that the auditor relies on the internal control, computer-related audit procedures or other GAS are more likely to be used. To obtain understanding of internal control during audit planning stage, professional judgement of auditors is needed to determine the client’s control risk. Other than that, the size of the auditing firm also affects the use of GAS. From a survey conducted by Janvrin (2009), auditors from big 4 firms are “more likely than smaller firms and national firms to use computer-related audit procedures”. Both factors mentioned to affect GAS implementation are caused by numbers human force in a firm and judgement from human auditors.

The implementation of GAS is encouraged by auditing standards, but there are still factors that affects auditors to not use it. Singleton (2006) stated that computer assisted auditing tools and techniques, known as CAATTs, are already adopted but underutilized due to the computer skills, such as databases and data management knowledge, needed. Technological proficiency requires general training that incurs cost for auditing firms that possibly outweigh the benefits of implementing use of GAS. Underutilization aside, current adoption of GAS and CAATTs in auditing still heavily relies on intervention from humans, as these systems are not yet able to operate on their own. Current auditing practices utilize technology to ease several audit procedures, such as selection of sample, but the role of human auditors in collecting evidence and for controls and dollar misstatement is still a prominent part in auditing.

Chapter 2 Findings and Discussion 

Artificial intelligence is defined as a computer system with the ability to imitate human behavior and intelligence to perform tasks (Marr, 2018). In the field of auditing, such technology can be utilized in performing audit procedures or judgement-making decisions, which will be covered in this chapter.

2.1. Artificial Intelligence in Audit

As technological advancement happens worldwide in a fast rate, the capability of programs to run and think on their own might also change, especially with rapid development of artificial intelligence. Artificial intelligence, shortened as AI, is usually implemented into systems to execute data processing and judgement-making, and significant progress has been made to apply it in business situations in the last few decades, which includes public accounting firms that offers auditing services (Vasarhelyi, 2016).

The intelligent computer capable of making decisions are in line with the main purpose of auditing as an assurance service, which involves gathering and evaluating data for samples and generate a reliable opinion on the presentation of the financial statements, as AI are able to aid auditors in making judgements (Omoteso, 2012). According to Carlson (1983), as cited by Omoteso (2012), there are three main phases of a decision-making process: “intelligence (which involves gathering data, identifying objectives, diagnosing problems, validating data and structuring problems), design (which comprises manipulating data, quantifying objectives, generating alternatives and assigning risks or values to alternatives) and choice (which involves generating statistics on alternatives, simulating results of alternatives, explaining alternatives, choosing among alternatives and explaining choice)”. An AI-based system with the ability to make decisions in a similar fashion to a human being by fulfilling the phases of decision-making can thus make better judgements by removing unknown biases that might arise in the work of human beings. Other than decision-making, AI can also assist investigative procedures such as analytical procedures, fulfil the classification audit objective, and assess and evaluate risk and internal control.

Applying AI systems to auditing brings various benefits, such as reduction of cost and efficiency of audit due to its capability to process large amounts of data efficiently (Bizzaro, 2017). The client being audited are using various documents in their business cycles that supports the numbers presented in financial statements that can be analyzed by AI, and other information stored online are easily accessible as well. AI is able to scan and test the large amount of complex data, as traditional audit procedures might be less efficient compared to newer inventions. This more efficient and less time-consuming method that can be learnt by AI to process data containing information can also be cost saving.

2.2. Future Implementation of Artificial Intelligence in Audit

With the rapid development of AI, it is not too far ahead from the future until a time where public accounting firms adopt technologies to incorporate audit and more sophisticated AI together. Given the known benefits and use of AI currently, the advancement of technology can be more useful in an audit process. In a few decades from now, it is possible for simpler auditing tasks to be assisted by AI.

2.2.1. Deep Learning

One specific form of AI technology that can be applied to auditing in the future is the deep learning technology enabled by convolutional neural networks. Deep learning is defined by LeCun (2015) as “computational models composed of multiple processing layers to learn representations of data.” Deep learning systems can learn structures of large data using algorithm to influence machines to react to representations.

Varashelyi (2016) mentioned that the features of a deep learning AI can utilize visual recognition to learn from existing images presented, which are connected to tags according to the visual content and relate the picture with specific logical classes. For example, Applebaum (2016) stated that such deep learning AI can scan current product catalogues, then scan a real-time recording of warehouse through automated drone camera for inventory counts that replaces a physical one. Other than that, deep learning AI can inspect written statements using linguistics analysis to identify patterns or trends which can be used to analyze conference transcripts and management discussions to be used in audit analysis (Sun, 2016). The AI can also be taught to classify texts from conference transcripts into “fraudulent” and “non-fraudulent” based on its interpretation that has been taught, where the lines deemed as “fraudulent” and possible accounts associated with it would be given more substantive attention by an auditor during testing (Varashelyi, 2016). The endless possibilities persist in the field, and development of machine learning to help auditors make better judgements in the future is technologically feasible.

2.2.2. Artificial Intelligence-Driven Automation of Audit

The use of AI and deep learning techniques can also help automatize various stages in an audit in the future. Louwers (2015) mentioned that there are general phases of an audit, which all can be automated with support of AI systems.

As an example, the first phase involves planning and designing the audit, where the next step after accepting a client is continued by gathering initial knowledge of the client and the industry. Varashelyi (2016) suggested the use of AI to compile data of the business from external sources, learn the operational methods and accounting system of the client, and then estimate initial risks.

Next, to understand the internal control of the client, the AI can utilize deep learning in the form of visual recognition previously covered to analyze and recognize patterns in identifying risk (Varashelyi, 2016). The audit evidence of observation can be replaced with automated drones, where the captured images or videos will be processed by the AI. During the tests of controls, a monitoring system supported by AI is suggested by Issa (2014) to search for exceptions in control, and each violation is compiled into a database that will indicate lacking key controls that might lead to material misstatements for the auditor to prioritize in testing.

In performing substantive tests, the implementation of AI auditing will enable firms to examine documents in real-time continuously. Rather than an interim or year-end test for transactions using samples, Varashelyi (2016) stated that the AI will be able to perform inspections on 100 percent of the population continuously. With this suggested system, Issa (2014) also stated that the constant, thorough examination of documents and records can detect misstatements caused by error or fraud and prevent it to be inputted to the financial statements. After all required audit procedures has been done, all data and information collected by the AI together with the professional judgement of an auditor will be used as a basis to issue an audit report containing an opinion on the financial statements.

2.3. Impact of Artificial Intelligence on Auditors

The possible implementation of AI features and deep learning technology in the future for an audit process might make the impression that human auditors are soon to be replaced, but it is not quite the case. Brenna (2017) mentioned that AI lacks human characteristics that can justify reason, exercise professional judgement, and apply skepticism. In processing large amounts of data in a shorter timeframe, it might be superior, but the skills of auditors to uncover actual reasons behind findings of errors or misstatements are still irreplaceable. Other than that, the ability of an auditor to exercise judgement and the intuition for incomplete or missed things are unable to be done by AI. EY digital leader Herman Sidhu (2017), as cited by Murphy (2017), stated that AI or machines can only work with data that they are given. In a scenario where client purposefully do not record transactions and it does not get recognized by the system, the AI might not know that the transaction ever happened because it is not in the data, and it cannot fulfil the audit objective of completeness.

Murphy (2017) also mentioned that it is possible for the standards and regulations to not catch up with rapid technological changes, and citing PwC’s Jon Andrews (2017), convincing the stakeholders such as audit committees, client, and regulators about the credibility of AI use in audit is a challenge. Auditor’s experience is also highly value to determine the individual’s competence in their specialization. With that in mind, it can be assumed that it is unlikely for stakeholders to fully trust AI in providing reasonable assurance alone. Instead of replacing the function of auditors, further development of AI will eventually be used in assisting auditors during the audit process. The use of auditor’s professional judgement will still be relied on, but the implementation of AI to be used alongside auditors instead of replacing them give users of financial statements higher level assurance because of the analytical capabilities provided by the AI.

2.4. Anticipating Technological Changes

Since AI will ideally be used in the future to help auditors make judgements in terms of optimizing work time and enabling auditors to analyze larger numbers of data to obtain reasonable assurance, current and future auditors need to prepare for the technological breakthrough. For individuals who eventually will work alongside AI, Deloitte’s chief innovation officer Jon Raphael, as cited by Ovaska-Few (2017), stated that future auditors need to prepare by improving their database and IT proficiency by joining projects, taking classes, and participating in seminars. Other than that, the professional skepticism still needs to be applied at all times, because even the most developed AI is not immune to errors, and auditors have to be able to detect when the AI’s analysis is incorrect and find ways to deal with unexpected exceptions in the system (Ovaska-Few, 2017). As mentioned by Issa (2014), tasks given to new auditors might be done by AI or automation powered by AI. In the future, individuals who are not able to contribute in technological awareness to operate the system and lack professional judgement to interpret analysis of the AI are in a position where they can be replaced, so proficiency to operate AI-based systems or exercising better judgement needs to be prepared.

Although the exact time when AI will be fully implemented in auditing is not determined yet, it will eventually be done so not only professional auditors but also university students taking accounting major also need to be exposed to the information and given general idea, as well as exercises to operate current GAS to build technological proficiency that can be useful when more complex systems are utilized in the future.

Chapter 3 Conclusion

In conclusion, the field of auditing has implemented technological tools to assist auditors in their work, and through technological changes, the main purpose of audit to provide reasonable assurance that the client’s financial statements are fairly presented remain unchanged. With emerging, possibly disruptive technologies such as AI, the job of auditor will not be fully eliminated, and the AI can be utilized to ease the audit processes by analyzing data or documents given and recognizing patterns of errors or misstatements that can be useful. At the end, the report which contains the audit opinion will still be issued by human auditors, but the process leading to it such as tests and data collection and analysis are helped by technology. To prepare for the upcoming changes, auditors need to be technologically proficient to be able to interpret analysis of the system, and professional skepticism will always be needed so auditors do not rely blindly on technological innovations.


Abdolmohammadi, M., & Usoff, C. (2001). A longitudinal study of applicable decision aids for detailed tasks in a financial audit. Intelligent Systems in Accounting, Finance & Management, 10(3), 139–154.

Ahmi, A., & Kent, S. (2013). The utilisation of generalized audit software (GAS) by external auditors. Managerial Auditing Journal, 28(2), 88–113.

Bizarro, P. (n.d.). Artficial Intelligence: The Future of Auditing. Internal Auditing., 32(5), 21–26.

Brenna, B., Baccala, M., and Flynn, M., Artificial intelligence comes to financial statement audits, CFO Magazine (Feb 6, 2017). Available at: http://ww2. t i ficial -intelligenceaudits/.

Debreceny, R., Sook-Leng, L., & Neo, W. (2005). Employing generalized audit software in the financial services sector: Challenges and opportunities. Managerial Auditing Journal, 20(6), 605–618.

Issa, H., and A. Kogan. 2014. A predictive ordered logistic regression model as a tool for quality review of control risk assessments. Journal of Information Systems 28 (2): 209–229

Janvrin, D., Bierstaker, J., & Lowe, D. (2009). An Investigation of Factors Influencing the Use of Computer-Related Audit Procedures. Journal of Information Systems, 23(1), 97–118.

LeCun, Y., Bengio, Y., & Hinton, G. (2015). Deep learning. Nature, 521(7553).

Marr, B. (2018). The Key Definitions of Artificial Intelligence (AI). Retrieved from

Murphy, H. (2017). Auditing to be Less of a Burden as Accountants Embrace AI. Retrieved from

Omoteso, K. (2012). The application of artificial intelligence in auditing: Looking back to the future. Expert Systems with Applications, 39(9), 8490–8495.

Ovaska-Few, S. (2017). How Artificial Intelligence is Changing Accounting. Retrieved from

Singleton, T. (2006), “Generalized audit software: effective and efficient tool for today’s IT audits”, Information Systems Control Journal, Vol. 2, p. 1.

Vasarhelyi, M. A. (2016). Research Ideas for Artificial Intelligence in Auditing: The Formalization of Audit and Workforce Supplementation. Journal of Emerging Technologies in Accounting., 13(2), 1–20.