AI in audit: what has really changed, and what hasn't

From samples to full populations

For decades, auditors tested samples of transactions and drew conclusions from them, because examining every entry by hand was impractical. Today, data analytics and AI tools can analyse an entire population in a short time. Common uses include:

• Journal entry analysis: flagging unusual entries, such as those posted at odd times, in round amounts, or by users who rarely post.
• Reading documents and contracts: extracting key terms from lease, financing and revenue contracts and comparing them with the accounting treatment.
• Matching at scale: matching invoices to purchase orders, goods received notes and payments.
• Predictive analytics: comparing actual balances with expectations built from historical and industry data.

What do standard setters say?

No standard setter has issued a dedicated AI standard, but two clear directions are emerging:

• The International Auditing and Assurance Standards Board (IAASB) launched an initiative in June 2025 to examine how emerging technologies, particularly AI, affect audit and assurance engagements. Stakeholders agreed that the quality management standards provide a strong foundation for managing technology-related risks, while calling for more practical guidance. The Board is also updating the evidence-related standards, including ISA 500, ISA 330 and ISA 520, for technology-enabled audits.
• The US Public Company Accounting Oversight Board (PCAOB) amended AS 1105 and AS 2301 to address audit procedures that use technology-assisted analysis of electronic information. The amendments apply to audits of financial statements for fiscal years beginning on or after 15 December 2025. Their aim is to reduce the risk of an auditor issuing an opinion without sufficient evidence when relying on such tools.

The common message is that tools do not change evidence requirements. Auditors must still assess the reliability of the data feeding the tool, the soundness of its logic, and how exceptions are followed up.

What hasn't changed

However advanced the tools become, four things remain at the heart of the profession:

1. Professional judgement: a tool may flag a hundred unusual entries, but deciding which represent a material risk remains a professional decision.
2. Professional scepticism: AI output is not evidence in itself. These tools can be confidently wrong, presenting incorrect information in persuasive language.
3. Responsibility for the opinion: the auditor alone is responsible for the opinion they sign, whatever tools they used.
4. Confidentiality: entering client data into unsecured public tools can breach confidentiality duties and data protection requirements.

The picture in Egypt

Egypt has launched the second edition of its National Artificial Intelligence Strategy (2025–2030), overseen by the National Council for Artificial Intelligence and the Ministry of Communications and Information Technology. It rests on six pillars: governance, technology, data, infrastructure, the ecosystem and talent. In accounting and tax, the expansion of the e-invoice and e-receipt systems has made a large volume of transaction data available in structured electronic form. This opens the door to full-population analysis, both in audits and in preparing for tax inspections. Practical challenges remain:

• Data quality: many small and medium-sized entities use non-integrated systems or manual spreadsheets, and analysis is only as good as its inputs.
• Multiple systems: data may sit across the accounting system, the e-invoice platform and sales or inventory systems, with no link ensuring consistency.
• Data protection: handling personal and financial data requires compliance with Egyptian data protection legislation, which should guide the choice of any tool.
• Skills: finance and audit teams need training that combines accounting knowledge with analytical skills.

What finance teams should do now

• Start with data, not tools: standardise charts of accounts, item and customer codes, and reconcile the accounting system with the e-invoice platform regularly.
• Adopt a written policy on AI tools: define what may be entered into them, who reviews their output and how this is documented.
• Keep a clear audit trail: any figure or estimate produced by an automated tool should be traceable and explainable.
• Ask your auditor: which tools do they use, how do they test data reliability, and how do they protect your information?

Conclusion

AI does not replace the auditor; it shortens the distance between expertise and execution. It enables broader testing in less time and directs effort to the highest-risk areas. But the value of an audit was, and remains, the independent professional judgement that gives users reasonable assurance about the financial statements.

Disclaimer: This article is for general information and does not constitute professional advice on any specific situation.

Sources

• IAASB — Technology focus area
• PCAOB — Amendments on technology-assisted analysis (AS 1105, AS 2301)
• OECD.AI — Egypt National AI Strategy

This article is for general information only and does not constitute professional advice. Thresholds, rates and procedures change — please contact us or refer to the latest official sources before acting.