The universe of data continues to grow – in volume, velocity, and variety. For forensic accountants, that means the scope of potentially relevant evidence is growing too. Moreover, forensic accountants must increasingly contend with evidence that is unstructured – meaning the information format doesn’t fit with conventional databases. Between the complexity of emerging data types and the sheer volume, manual methods of data management and analysis can’t keep up.
by Akriti Anand Jun 1, 2022, 09:50 AM
The universe of data continues to grow – in volume, velocity, and variety. For forensic accountants, that means the scope of potentially relevant evidence is growing too. Moreover, forensic accountants must increasingly contend with evidence that is unstructured – meaning the information format doesn’t fit with conventional databases. Between the complexity of emerging data types and the sheer volume, manual methods of data management and analysis can’t keep up.
That’s where artificial intelligence (AI) comes in. AI is the umbrella term for software that performs one or more different forms of “intelligent” action. Modern AI is an extension of advanced analytics, powered by computer algorithms that process data in a way that mimics human behavior and thought processes. AI algorithms can be used to find and classify relationships, identify patterns or anomalies, analyze probabilities, and make predictions.
Machine learning is a subset of AI in which algorithms are “trained” to learn on their own. There is an element of independent adaptation, which operates by building on previous computations. Machine learning applications self-learn from past mistakes to produce more accurate insights after each iteration. The product is reliable, repeatable results.
AI and machine learning can be used in any stage of analysis – from collecting data (where certain rules and learned behavior can help collect only relevant data) to preparing and cleaning the data. Machine learning can train the data-using techniques through supervised learning (where systems are taught behaviors through examples) or unsupervised learning (where systems analyze data without human guidance to identify patterns). These techniques are particularly helpful when data comes from several different sources and may be incorrect, incomplete, or inconsistent.
While the goal of AI is to simulate human abilities – and perform tasks more efficiently – the intention is to complement humans rather than replace them. AI solutions can be used to expedite review, but the conclusions drawn are only as good as the people behind them. Success is contingent on including the right data and asking the right questions. Outputs must be contextualized with human insights and instincts based on real-world experiences.
A study published in 2018 by the Association of Certified Fraud Examiners looked at 2,690 cases of occupational fraud and identified a total loss of $7 billion to victim organizations over a period of one and a half years.1 The study also found that a typical organization, on average, loses 5 percent of its revenue to fraud. Forensic accountants are often ill-equipped to find the paper trail using traditional methods. For example, if the fraud involves falsified employee expenses, it is extremely tedious for forensic accountants to go through every receipt.
Employee expense fraud is one of the many areas where AI software can assist. Custom-developed tools and commercial tools, such as those from MindBridge Analytics Inc. and AppZen, offer sophisticated algorithms based on AI, machine learning, predictive analytics, deep learning, and natural language processing to detect fraud more quickly and powerfully than traditional methods. These tools improve efficiencies for forensic accountants and yield deeper insights. These insights can help companies better pinpoint the source and cause of the issues, which in turn helps identify where controls can be improved to avoid similar fraud in the future.
AI also provides forensic accountants with the computing power to process the large volumes of data they encounter in every engagement. Twenty years ago, CPAs were able to conduct virtually all analysis using spreadsheets – and this was still true for many projects just five or 10 years ago. This has changed drastically with the proliferation of data-generating devices and software packages, and every year it is reaching new proportions. With AI, CPAs no longer must manually comb through thousands of emails and other documents to find that one detail for which they are looking. With just the click of a button, CPAs can find specific words or patterns that require closer scrutiny.
A typical AI project for forensic accountants involves several possible steps. AI can be used to extract, organize, and structure data; then it can be used to link disparate sets of information, examine anomalies, and identify high-risk transactions. The forensic accountant directs the software to perform the tasks and can provide inputs at various stages of the process. In addition, the forensic accountant must be able to verify the results and explain how the results were reached – an important reason why AI will not be replacing forensic accountants anytime soon.
AI is useful for numerous areas of forensic accounting, commonly where there is a large volume of data, there are complex relationships between the data, and there are patterns within the data that need to be identified quickly. In addition to employee fraud, the following are key areas where AI is currently being applied within forensic accounting:
AI systems cannot run fully independent of humans because they do not perform all of the analysis experienced CPAs can. While AI can provide efficient and accurate data, forensic accountants are still needed to interpret and verify results, as well as to perform more complex and subjective evaluations. Changes in how organizations conduct business will provide additional opportunities for forensic accountants to add value in conjunction with AI systems, since these systems are designed for what is currently happening and trained on prior data. These systems do not fully understand how business environments change. Similarly, a company might decide to change directions in terms of services or products based on personal preference, which will require an AI system to be retrained based on these changes.
Soon, CPAs will no longer have to deal with the tedious tasks of collecting, organizing, and reviewing data manually. It will all be automated. Forensic accounting will move away from spreadsheet-based analysis and toward dashboards where the forensic accountant instructs AI software on what to analyze. The software will automatically identify red flags and other issues that require the CPA’s attention. Forensic accountants will spend the bulk of their time on interpreting results and performing higher-level analysis. The technology also may become similar to digital voice assistants, like Google Home or Amazon Alexa, where forensic accountants would be able to train it to extract, transform, and load data using speech commands.
AI and machine learning will become more pervasive in the coming years, and it will affect every business function in an organization. Algorithms will get smarter, faster, and more reliable. While they will provide more efficiency and make the influx of information more manageable, organizations will still depend on CPAs to interpret and validate analysis, add technical expertise, and inform strategic decision-making.
1 https://s3-us-west-2.amazonaws.com/acfepublic/2018-report-to-the-nations.pdf
Akriti Anand is an associate, data analytics and automation, for BDO USA LLP in the Houston area. She can be reached at aanand@bdo.com.