The role of artificial intelligence in auditing: enhancing accuracy and reducing fraud in the post-pandemic era

Abstract

The COVID-19 pandemic accelerated digital transformation throughout business and financial ecosystems; auditing was no exception. With remote working, distributed data sources, and rapidly changing risk landscapes, auditors confronted both logistical constraints and novel fraud risks. Artificial intelligence (AI) — including machine learning (ML), natural language processing (NLP), anomaly detection, and robotic process automation (RPA) — has emerged as a pivotal technology to help auditors enhance evidence-gathering, automate routine procedures, and detect sophisticated fraud patterns that are difficult to find with traditional sampling techniques. This paper analyses how AI has changed audit methodologies in the post-pandemic era, assesses its capacity to improve accuracy and fraud detection, identifies practical and ethical challenges (such as model bias, explainability, and governance), and proposes a framework for safe, effective, and regulated AI adoption in audit practice. The study uses a mixed-methods approach: a systematic review of literature up to 2024; policy and standards analysis (including pandemic audit-guidance); and case-based illustrations from Big Four and large-firm implementations. Key findings indicate that AI tools can increase anomaly detection rates, expand population testing beyond statistical samples, and shorten time to detection for unusual transactions — but these gains depend on data quality, model selection, interpretability, and robust governance. Independent reviews and regulatory bodies have observed that firms often lack metrics to evaluate AI’s impact on audit quality and rarely maintain formal monitoring of algorithmic performance and risk. Ethical concerns—algorithmic bias, over-reliance on automation, model drift, and opaque decision-making—are material and require both technical and procedural safeguards. Effective adoption requires harmonizing technology with auditing standards, enhancing auditor capabilities (data science literacy), instituting continuous model validation and performance metrics, and creating a layered governance model that ties AI outputs to professional skepticism and human oversight. Policy recommendations include: (1) standardized AI-audit performance metrics and reporting; (2) mandatory documentation and explainability requirements for AI tools used in substantive procedures; (3) third-party or regulator-led audits of AI systems (AI-audit of audit tools); (4) auditor upskilling programs and interdisciplinary teams; and (5) alignment between professional standards and technology risk frameworks. The paper concludes that while AI has strong potential to both enhance audit accuracy and reduce fraud in the post-pandemic environment, the benefits will materialize sustainably only if auditing firms, standard setters, and regulators collaborate to ensure transparent, accountable and validated use of AI.