CFOs struggle to explain AI decisions to accountants

Recent research shows that many CFOs in the Netherlands and Europe have difficulty explaining algorithmic decisions, particularly in machine learning and generative AI, to their accountants. This issue is especially relevant within ERP and accounting software, creating tension between year-end audits and internal audits. The increasing pressure from the European AI regulation and the General Data Protection Regulation (GDPR) makes this issue even more pertinent.

Accountants are bound by strict standards and need to understand how financial data is generated to conduct reliable audits. However, the use of AI makes it more challenging to obtain this information transparently, resulting in a lack of audit information. This exacerbates the risks that arise when transparency in financial processes is lacking. The NBA and the AFM have long emphasized that organizations using AI, for example for journal entries and risk analyses, must be able to clarify the processes and decisions made by the AI.

The European AI regulation imposes specific requirements on transparency and risk management, which also affects applications within the financial sector. Meanwhile, the GDPR remains in effect, forcing organizations to minimize data usage and to carefully handle personal data. Many finance departments lack an adequate model register and a documented audit trail, making it more difficult to retrieve data and validate models. This has far-reaching consequences, not only for internal control but also for the publication of financial statements, which may be delayed.

Regarding the implementation of explainable AI, companies are advised to formulate clear policies regarding the use of AI in finance, and to establish a model register that facilitates oversight and evaluations. It is crucial that organizations demand that their software providers make AI technical documentation available and keep it up to date. By investing in documentation and explainability, audit processes can run more smoothly and risks can be reduced. In a time where transparency is becoming increasingly important, this will not only strengthen the trust of executives but also help ensure strong governance.

Read the full article from AI Insider.