Accounting from spreadsheets to AI

Unlock Editor’s Digest Lock for Free

FT editor Roula Khalaf will select your favorite stories in this weekly newsletter.

Martin Mulyadi is a professor of accounting and Unita Anwar is an assistant professor of accounting at the Shenandoah University School of Business in Winchester, Virginia

The series of audit failures highlights that traditional audit approaches may no longer fit into today’s financial world. However, research suggests that artificial intelligence can significantly improve the efficiency and accuracy of audit practices, but without a robust governance framework, proper training, ethical AI practices, and human monitoring, this technology poses a great risk to auditors.

AI has the potential to transform audit practices in a way similar to the impact of the birth of digital spreadsheets. By automating calculations, development in 1979 allowed accountants to increase their time in decision making and change their role in the business world.

AI can have similarly transformative effects. Daniel Davis, Managing Director of Consultant Frontline Analyst, explains the future in which AI solves dense data in its annual report and transforms static, often vague documents into dynamic, interactive tools. AI can sort huge amounts of financial data, he says, to enable you to find patterns and inconsistencies.

Based on the productivity outlook, accounting firms have invested billions in developing AI tools to autonomously handle everyday tasks, making them essential to their business. It also partners with high-tech companies such as Nvidia, Microsoft, Google, Oracle, and Salesforce to integrate AI into core services.

Accounting companies are beginning to see profits. For example, an AI fraud detection system tried by EY along with 10 UK audit clients who flagged suspicious activity in two companies. In both cases, clients later see that fraud occurred, suggesting that AI can dramatically improve the quality of their audits by catching irregularities that have overlooked traditional methods.

As Accenture and Grant Thornton discovered, AI also offers increased efficiency. Thomson Reuters Beta Testers reportedly cut sample sizes and testing times for certain procedures in half, but Deloitte believes AI can free financial agents from thousands of hours of work each year, potentially reducing costs by up to 25%. By automating the boring process of sifting through mountains of data, auditors can focus on higher risk areas, make complex decisions, and allow accountants to work as strategic advisors to use human knowledge to support AI analytics.

Rather than relying on historical sampling, AI can analyze the complete dataset, making it unusually prone to zero for auditors to zero. You can also streamline the pitching process for new businesses by utilizing a database of past work. This could improve efficiency and profitability.

The nature of AI’s “black box” – lack of visibility of training data and methods – makes it difficult to understand how and why AI technology reaches conclusions

However, systems with AI also create dilemmas. One academic study has identified issues such as bias built into AI algorithms and warns auditors to ensure that the decisions they make are fair, accountable and transparent. For example, as empirical studies suggest, the application of a large-scale language model to mortgage underwriting resulted in higher rejection rates and interest rates for black borrowers compared to black borrowers.

Another concern is the nature of AI’s “black box.” This means there is no vision in training data and methods. As the Audit Quality Center pointed out, this makes it difficult to understand how and why AI technology can reach its conclusions. Furthermore, AI technology is probabilistic. Rather than retrieving fact data like a search engine, it predicts the response. This means that asking the same question multiple times gives you a different answer, which can produce “hastisation” or inaccuracy. AI can produce inconsistent results because data is misconfigured or processes are not standardized.

One study warns that they are overreliant on AI without proper human scrutiny, known as the “human loop.” The technology warns that it can undermine the auditor’s ability to identify subtle irregularities and fraud. Deloitte and KPMG have expressed similar concerns, arguing that if AI is trained in the case of past fraud, it may not detect new forms of fraud designed to circumvent existing security measures.

Companies need to balance automation and human judgment, and they need to equip employees to evaluate AI output and understand the limitations

As a result, businesses need to take a cautious step when incorporating AI into accounting and reporting. Robust governance, ethical monitoring and excellent data management are essential. On the other hand, businesses need to balance automation with human judgment. This means that employees are prepared to evaluate AI power and understand the limitations of technology. Implementing AI also requires an overall strategy driven by senior leaders.

This comes with high costs and uncertain returns on investment. However, companies that have slower adoption of AI in audits also face risks such as reduced efficiency and quality of audits, according to accounting training firm Mercia Group. Meanwhile, employees are increasingly fond of dealing with new technology, so businesses can encounter difficulties in hiring top talent.

While AI offers the potential to increase productivity and efficiency, there are many challenges when incorporating AI into accounting and auditing. Accounting and auditing experts should weigh the transformational power of AI-powered systems against the liability and risks associated with using them.