EY's innovative use of AI in detecting audit frauds sparks debate

EY has implemented artificial intelligence (AI) in its audit processes, leading to the detection of fraudulent activities. Out of the first 10 companies evaluated by this new system, suspicious activities were identified in two, which were later confirmed as frauds by the clients.

Kath Barrow, EY’s UK and Ireland assurance managing partner, commented on the effectiveness of the system, indicating that these early results demonstrate its potential utility in auditing.

While EY declined to reveal the details of its software or the nature of the frauds it had discovered, Barrow said the results suggested the technology had “legs” for auditing.

EY began experimenting with AI in audit in 2018.

Naoto Ichihara, an assurance partner for Ernst & Young ShinNihon LLC in the Tokyo office, has always had a passion for programming. His expertise in developing models and systems for audit led him to explore the application of machine learning in accounting data analysis.

Naoto’s extensive research of existing academic papers and algorithms sparked his realization that there was a better way to detect anomalies through machine learning.

Driven by his vision, Naoto coded an AI solution that could sense anomalous entries in large databases. This technology became the first-of-its-kind in the auditing field and was subsequently patented.

Recognising the need for collaboration, Naoto built a team of auditors and developers to test and improve the detection method of the solution, which was later named EY Helix GL Anomaly Detector or Helix GLAD.

Overcoming scepticism and gaining trust

The AI system’s ability to analyse vast amounts of data quickly and efficiently could provide a powerful tool for auditors, alerting them to signs of wrongdoing and other issues. However, the industry remains divided on the reliability of this technology, with some firms expressing scepticism about AI’s ability to detect the myriad forms of potential fraud.

Auditors play a crucial role in evaluating flagged entries and recommending appropriate actions. To gain their trust, the team conducted rigorous testing of the solution against a dataset where fraudulent journal entries were pre-determined.

As the assurance team observed the algorithm accurately uncovering the fraudulent entries, they began to believe in the potential of Helix GLAD to improve the accuracy of auditing processes. However, a critical aspect was missing – auditors had no insight into why the algorithm detected specific anomalies. This knowledge was vital in evaluating the impact and validity of the flagged entries.

To address this, the team devised a solution that leveraged data analytics to create visual maps of the flagged entries and the reasons behind their detection, providing auditors with a transparent view of the algorithm’s detection method.

The tool’s ability to analyse vast amounts of data and flag suspicious transactions has transformed the auditing landscape. It has not only enhanced the accuracy and efficiency of audits but has also significantly reduced the risk of financial irregularities going undetected.

Benefits and challenges

The integration of AI within audit fraud detection processes has brought poses numerous benefits for large accounting firms. AI algorithms can analyse large volumes of data in a fraction of the time it would take a human auditor.

This efficiency allows auditors to focus on analysing and interpreting the results rather than spending countless hours manually reviewing data. These models are also not subject to human bias or fatigue. They consistently apply predefined rules and criteria to identify anomalies, ensuring a more objective and reliable approach to fraud detection. This reduces the risk of overlooking suspicious transactions due to human error or oversight.

However, the implementation of AI in audit fraud detection is not without its challenges. One significant challenge is the integration of AI technology into existing auditing systems and processes.

Auditing firms must ensure that the AI algorithms are compatible with their existing infrastructure and can seamlessly integrate into their workflows. This requires careful planning, training, and collaboration between auditors and AI specialists.

Another challenge is the need for continuous monitoring and updating of AI algorithms. As fraudsters evolve their techniques, AI algorithms must also adapt to detect new patterns and anomalies. Auditors and developers need to work together to refine and update the algorithms to stay ahead of emerging threats.

Fraud in the future?

Regulators are likely to have the final say on whether accountants can use AI to detect fraud during the audit process – key to this will be whether accountants have the skills to critique AI systems.

According to a report in the Financial Times, Jason Bradley, head of assurance technology for the UK’s Financial Reporting Council, believes that AI presents opportunities to improve audit quality and efficiency if used appropriately.

Then there is the issue of data ownership. A company might regard its detailed financial data as proprietary information, making it difficult to use that private information to train a system that subsequently audited another company.