AI and the Future of Accounting: Opportunities and Risks

Accountants in the Netherlands are currently investigating the impact of artificial intelligence (AI) on their professional practice. They are testing various systems such as Microsoft Copilot, OpenAI’s GPT-4o, and Google Gemini for daily tasks, in the context of the upcoming European AI regulation. The objectives are aimed at improving the speed and accuracy of their work while ensuring that control and professional ethics remain intact. It remains important that accountants maintain ultimate responsibility, despite the support that AI provides.

Generative AI, such as GPT-4o and Gemini 1.5 Pro, can generate texts and analyze data, but it is not infallible and can 'hallucinate', leading to unreliable outcomes. This requires accountants to thoroughly check and document every generated output, including the prompts and source data used, so that oversight and control can take place. This professionalization is crucial, especially in light of regulations such as the IESBA Code of Ethics and NV COS, which emphasize the accountant's own judgment.

Moreover, AI provides tools for productivity improvement, particularly in data analysis and reporting. Intelligent algorithms can signal anomalies in large datasets, but detailed checks remain essential, especially with exceptional or complex data. Despite the benefits, some AI applications may struggle with multilingual documents or poorly structured data, necessitating extra caution.

The European AI regulation imposes strict rules on the use of AI in accounting. This legislation emphasizes that users such as accounting firms must assess risks and ensure human supervision, while GDPR continues to apply to the processing of client data. Additionally, choosing the right AI system is important. Many organizations are considering hosting sensitive data locally or using open-source models hosted within the EU. Training and internal policies are evolving to integrate new skills into the work of accountants, paying attention to privacy, risks, and quality assurance.

It is advisable to start the implementation of AI with small, defined tasks and carefully measure the results before scaling up. By making clear agreements about what should be automated and what should be done manually, compliance can be promoted, and transparency can be created for clients regarding the systems used and their implications. This helps to build trust in the results produced by AI.

Read the full article from AI Insider.