Artificial intelligence cannot replace auditors.

At an international conference on improving auditing capacity on October 13, Deputy State Auditor General Bui Quoc Dung said that the development of AI allows auditing agencies to scan and process entire data sets, reducing subjective factors in assessment.

According to him, machine learning and deep learning algorithms help convert large volumes of text and records into accessible and collated information. This helps auditors detect violations and can predict trends, errors… to intervene early.

However, the Deputy State Auditor General believes that AI cannot replace humans because experience, intuition and professional ethics are still the factors that determine the quality of auditors. He cited a tax settlement report of an enterprise prepared with the support of specialized software that can be very “standard and beautiful” so AI is difficult to detect errors. However, auditors can detect suspicious signs thanks to experience and acumen.

“AI is a support tool, helping to increase efficiency, but cannot completely replace auditors,” said Mr. Dung. According to him, this requires auditors to have the qualifications and capacity to use technology to handle tasks that artificial intelligence cannot handle.

In fact, many countries apply artificial intelligence in the field of auditing. For example, in the US, the Government Audit Office (GAO) uses AI to monitor federal finances, health care and banking risks.

In the UK, the UK National Audit Office (NAO) applies AI in social welfare, health care, public contracts and builds a “playbook” using safe artificial intelligence. In Pakistan, AI has been used to detect 128,000 cases of “ghost pensioners” in pension payments.

Mr. Bui Quoc Dung said that the State Audit is building a big data platform, connecting with ministries and implementing AI projects to serve public auditing. They have created a database of more than 100 million records, connecting with the Ministry of Finance, Vietnam Social Security and the State Bank.

The application of new technology in the context of the volume and complexity of public data exceeds the coverage capacity of traditional sampling-based auditing methods. For example, Vietnam Social Security processes 17 million mandatory participations per month, issues 96 million health insurance cards per year and processes more than 200 million medical examinations and treatments. By the end of last year, the Tax Department had more than 950,000 businesses declaring electronically with nearly 16 million records and 150 million declarations.

“These are huge data pictures, continuously updated in real time. If we continue to process them using traditional manual approaches, we risk missing systemic risks and reducing the reliability of audit conclusions,” he said.

However, Mr. Pham Huy Thong, Deputy Director of the Department of Information Technology (State Audit) pointed out three major challenges when deploying artificial intelligence (AI) in public audits. That is, the technology infrastructure is not yet synchronized, making its application limited. Sharing data between ministries and branches is still difficult despite coordination regulations, and the quality of IT human resources and auditors is still lacking and uneven.

In this aspect, Mr. Phan Ngoc Anh, Deputy General Director of Deloitte Vietnam Audit Services, warned of challenges in data security and algorithm bias.

Meanwhile, a representative of the banking sector, Ms. Nguyen Ngoc Lan Anh, Director of Technology and Banking at Standard Chartered Vietnam, said that AI is not the solution to all problems. According to her, the most important thing is to put risk control in from the beginning, like equipping “safety brakes for a high-speed vehicle”.