i. Introduction
The beginning of auditing can be traced as far back as to ancient times, however the financial audit that we know of today is a relatively new practice and is constantly changing. As technology advances, we move further away from manual audit procedures and towards an automated audit. One of the most important technologies playing a role in automating the audit is artificial intelligence. Artificial intelligence is the development of computer systems able to perform tasks normally requiring human intelligence, such as visual perception, speech recognition, and decision-making (Lexico Dictionaries, 2019). It has already been applied in areas such as driverless cars, home-energy systems, investment portfolio management, and many more. The accounting industry is no exception, artificial intelligence is transforming the way we perform audits. This technology has the ability to process massive amounts of data and report on behavior, trends, and anomalies, making it easier for auditors to identify potential audit risks. The purpose of this report is to discuss how the deep learning function of artificial intelligence is being applied to traditional audit procedures. We will look at existing and developing technology at the Big Four accounting firms that has been created as a result of this new initiative, as well address some risks that this technology may introduce to the industry.
ii. Deep Learning
One major function of artificial intelligence is deep learning. Deep learning is a subdivision of machine learning and is concerned with algorithms that are inspired by the human brain (Brownlee, 2019). Deep learning creates artificial neural networks which can analyze unstructured data such as emails, social media, and conference calls which can provide crucial information for companies, along with auditors. When provided with sets of big data, deep learning technology can recognize patterns at a size and speed that would be impossible for humans. Deep learning technology has multiple hidden layers, the neural networks automatically “learn” from massive amounts of data, which can be structured or unstructured, received in the input layer (e.g., millions of images, years’ worth of speeches, tera-bytes of text files). From there it can recognize data patterns in more and more abstract representations as the data is processed and transmitted from one layer to the next and classify the data into predefined categories in the output layer (Sun, 2017).
Save your time!
We can take care of your essay
- Proper editing and formatting
- Free revision, title page, and bibliography
- Flexible prices and money-back guarantee
Most big data us semi-structured or unstructured, which auditors use to make decisions and to freely explore the status of their clients’ products, services, and operations, reducing their dependence on the client for information. This information needs categorizing and labeling, however, auditors cannot do this manually due to the volume of big data sets. Deep learning allows artificial intelligence to mine and extract meaningful patterns from these big data sets which creates great value for audit decision-making and risk assessment. Auditors can now take larger samples and analyze larger sets of data, all while cutting down on tedious and repetitive audit processes and enhancing audit effectiveness and efficiency. Checking inventories, processing paperwork, reviewing contracts, and drafting audit reports are only some of the tasks that have become automated through deep learning that were traditionally done manually during an audit.
iii. Applications to the Audit
No matter what industry you are in, cognitive technologies fall into three main categories: product, process, or insight. AI systems that fall under the product category use the technology to increase the value of a product or service by making them more effective, convenient, safer, faster, etc. The process category consists of systems that have been automated that used to be done manually, these are typically internally focused and benefit the organization rather than the customers. Lastly, the insight category uses artificial intelligence to learn from information, draw conclusions, and generate insights for companies that can help reduce costs, improve efficiency, or enhance customer service (Schatsky, 2015). When applied to an audit, most of the technology being used falls under the process or insight category.
One of the most common applications of artificial intelligence to audit procedures is text analysis. Through a company’s operations, a high amount of written data/reports are generated which can be analyzed through text analysis. Transcripts of conference calls, press releases, MD&As, earnings announcements, business contracts, and social media messages are only some of the input textual data that artificial intelligence can analyze and create meaningful output for auditors to base decisions off. Output features of text analysis can be a sentiment, emotion, entity, topics, concepts, keywords, and more. The real value of text analysis comes from when it is applied to improvised content such as conference call transcripts, status updates, and Q&A’s rather than prepared content (press releases, presentations). This is because improvised content consists of more linguistic clues that reflect the cognitive process of the speaker which the deep learning technology can pick up on and point out areas of potential risk in the output layer. Textual analysis can be applied to audit procedures such as inspection of documents, confirmations with third parties, and analytical procedures.
Similar to text analysis is speech recognition, however, this application can actually analyze audio files of speech which may provide more insight than transcripts or text files. During an audit, managers, employees and other professionals who have relationships with the client (bankers, prior auditors, shareholders, etc.) are interviewed in order to gain information and identify areas of risk in the company. Artificial intelligence can analyze the recordings of these interviews and how the individuals answer the questions asked to them. This is similar to why text analysis is better with improvised textual data, because terms used while speaking can indicate dishonesty or uncertainty. Speech recognition does not only transcribe audio files, it can also translate one language to another, which is useful if the client being audited has operations overseas such as a foreign subsidiary. The auditor would not need to hire a translator or find someone who could speak the language, making the audit more efficient. The input data for speech analysis deep learning could be recordings of interviews (as mentioned before), phone calls, meetings, and presentations. The output features of speech recognition is similar to those of test analysis, however, it includes areas of possible deception. This application is relevant to audit procedures such as inquiry of personnel and management, as well as inspection of (audio) files.
Another function of deep learning that artificial intelligence can apply to an audit is image and video analysis. Deep learning systems have the ability to extract a series of predefined numerical attributes describing the content of the image or video, attach searchable tags accordingly, and save both the attributes and the images to the auditor's output data (Sun, 2017). For example, this function can be useful to count and check the condition of inventory, as well as identify human faces and identify activity such as theft. The input data for the image and video analysis function includes inventory counting and other control activity, interviews, and videos taken in the office, warehouse, store, etc. Output features include objects, human faces (identities), concepts, scenes, and activities. This function is applicable to audit procedures such as observation, inquiry, and inspection of documentation.
iv. The Big Four’s Implementation of Artificial Intelligence
It is no surprise that artificial intelligence technology is very costly to implement and operate, which is part of the reason why the Big Four have been the first to have access to it. The Big Four accounting firms (Deloitte, PwC, EY, KPMG) are the largest accounting firms in the world, thus they have the resources to implement innovative and expensive technology to improve the quality of their work. The Big Four serve some of the largest companies in the world, so the data sets that they receive for audit, tax, or consulting work are often way too large for a group of people, yet along a single person, to comprehend without the help of technology. The Big Four have all introduced artificial intelligence to their work recently, embracing the new era of automation in accounting. However, not all the technology that has been implemented at these firms are the same, they have invested large amounts of money in order to come up with their own proprietary technology that is unique to their firm.
Deloitte has a number of artificial intelligence-enabled processes that have been introduced over the past decade. For example, their document-review platform has automated the process of reviewing and extracting all the relevant information from contracts, which reduces exhaustive and difficult human efforts. The application of this technology has reduced the time spent retrieving crucial information of documents such as contracts, invoices, financial statements, and meeting minutes by fifty percent or more (Faggella, 2019). However, Deloitte is not only using artificial intelligence for audit procedures. They recently created a system for employees called “Vitals” which collects information from internal systems to see when workers may need a break. “Employees can see a complete picture of how many hours they are working… how much of that time they’re spending away from home, how many flights they have taken in the past week, and when they last took PTO. It also allows employees to share energy levels with their coach.” (Kohll, 2019). This tool helps identify employees that may be at risk of burnout before they crash, which is just another benefit of artificial intelligence in the workplace.
Another Big Four firm, EY, has created its own proprietary Robotic Process Automation (RPA) system to help with audits. This uses robots that mimic human actions and automate repetitive tasks across multiple business applications without altering existing infrastructure and systems. In an audit, this technology can be used to account audit requests, perform data analytics, and assess internal controls (Robotic Process Automation, 2016).
EY has also launched artificial intelligence that uses computer vision to enable airborne drones to monitor inventory during the auditing process (Faggella, 2019). The use of drones will allow more data to be captured and analyzed during the audit, focusing the auditor’s attention more on risk areas rather than having to manually count inventory. The drones can observe and physically examine evidence while the auditors can apply their minds to issues that are more strategy or judgement-oriented. This initiative has not been set in practice yet, it is still in research and development, however it is a promising technology that gives us insight to the future of auditing.
PwC has some unique audit technology as well. In October 2017, the International Accounting Bulletin named PwC’s “GL.ai” ‘Audit Innovation of the Year’. This technology examines every uploaded transaction, user, amount, and account to find unusual transactions in the general ledger, which could indicate error or fraud, without bias or variability (PricewaterhouseCoopers, 2019). The interesting thing about this technology is that it uses PwC’s experience within its algorithms, so it actually gets smarter the more they use it. GL.ai increases the efficiency of the audit and provides the auditors with comfort that they are focuses on areas of true risk.
Lastly, KPMG is no short comer when it comes to artificial intelligence, they have built their own portfolio of AI tools which they refer to as KPMG Ignite. Some examples of what lies in KPMG’s AI portfolio are a call center analytics engine, AI anomalous event predicting tool, and document compliance assessment engine. The call center analytics engine uses artificial intelligence to create a transcript of customer calls, and can identify keywords, sentiment, and predict future trends. The AI anomalous event predicting tool also uses an AI model that can predict future business events. And the document compliance assessment engine uses artificial intelligence to read the documentation and extract appropriate information, similar to technology found at some of the other Big Four firms (Faggella, 2019).
v. Risks of Artificial Intelligence
There are obvious benefits to artificial intelligence, as stated above, however with new technology often comes new risks. The main goal of artificial intelligence is to mitigate risk, however, does it bring any new risks with its implementation? Yes, there have been risks identified that stem from artificial intelligence, however, if you can identify the risks beforehand then they can be managed. One risk associated with machine learning is algorithmic bias. The algorithms used in artificial intelligence technology identify patterns in data, and if those patterns or data reflect an existing bias, the bias is likely to be amplified and the results will support the existing patterns of discrimination. Artificial intelligence systems are very complex so they are prone to programmatic errors, and if these errors are not identified then the results can be misleading which can have serious consequences if relied on too heavily. Also, AI systems are commonly the target of cyber-attacks, so they must be well secured or they run the risk of allowing hackers access to personal data or confidential information. One last major risk of artificial intelligence is that it is an increasingly new concept, although the idea has been around since the 1950s, recent advancements are extremely remarkable and unforeseen. That being said, there is little legislation governing this technology, but there will be soon, and systems that are currently able to analyze these large sets of data may not comply with future regulations (Boillet, 2018).
vi. Conclusion
Artificial intelligence may seem like the wave of the future, however, for many industries, it is already here. The accounting industry is not the first to implement machine learning and it will not be the last. Although we are mostly seeing deep learning technology being implemented at large accounting firms who have the resources to do so, there is no doubt that it will make its way into firms of all sizes relatively soon.
Artificial intelligence is becoming exponentially more advanced and will be a major part of anybody’s day-to-day whether they work in accounting or any other industry. The audit process will be streamlined due to this technology, allowing for more efficient and reliable audits, which will not only benefit the accounting firms, but also businesses and society as a whole.
Stuck on your essay?
