Auditors say AI can sharply speed up and improve audits—particularly fraud detection—but uptake is slowed by high costs, a shortage of skilled staff, and data-security worries, making a human-plus-AI hybrid the expected outcome.

Main Finding

AI substantially improves audit efficiency and accuracy by automating repetitive tasks and analyzing large datasets, enabling better fraud detection, anomaly identification, and near real‑time monitoring. However, meaningful constraints—cost, integration complexity, scarce AI-skilled auditors, and data privacy/security risks—mean a hybrid model (AI augmenting human auditors, not replacing them) is the recommended path forward.

Key Points

• Scope and contribution
  • Comparative study of AI-based vs. traditional auditing on three dimensions: efficiency (time), accuracy/reliability, and scope (breadth of data analysis).
  • Adds empirical and qualitative evidence to debates about automation in auditing and recommended practice (hybrid model).
• Principal benefits of AI in auditing
  • Automates repetitive, rule-driven tasks (e.g., reconciliations, sampling), reducing time to complete audit tasks.
  • Processes large volumes of transactional and digital-source data, enabling more comprehensive audits than traditional sampling.
  • Improves detection of fraud and anomalies and allows for ongoing/real‑time monitoring.
  • Increases consistency and reduces human error when models and rules are properly validated.
• Key constraints and risks
  • High upfront and ongoing investment (software, infrastructure).
  • Shortage of auditors with AI/RPA/data-analytics expertise.
  • Integration challenges with legacy ERP and accounting systems.
  • Data confidentiality, security, and system inconsistency risks (potential data loss).
  • Risk of over‑reliance/misinterpretation of model outputs; need for human judgment on ambiguous or ethical issues.
• Recommendations from authors
  • Strengthen data security standards and governance.
  • Invest in auditor training and capacity building on AI and data analytics.
  • Continue research into ethical implications and create oversight frameworks.
  • Adopt a hybrid auditing paradigm: AI augments human auditors, with humans retaining responsibility for judgment, ethics, and final assurance.

Data & Methods

• Study design: mixed-methods approach combining qualitative thematic analysis and quantitative survey analysis.
• Primary data collection: online questionnaire surveys (used to gather practitioner views and perceptions).
• Qualitative approach: thematic analysis of interviews/responses to identify perceived benefits, limitations, and practical challenges.
• Quantitative approach: survey analysis comparing perceptions/metrics of efficiency, accuracy, and scope between AI-enabled and traditional methods.
• Scope and sample: focused on auditing industry practitioners and firms; exact sample sizes, sampling frame, and statistical details are not provided in the excerpt.
• Limitations (method-related): potential generalizability limits due to unspecified sample details; rapid evolution in AI tools may outdate specific technologies referenced.

Implications for AI Economics

• Labor and skills
  • Routine audit tasks are likely to be automated, reducing demand for low‑skill audit labor but raising demand for auditors with AI, data analytics, and IT-integrations skills.
  • Net effect: short‑term displacement and transition costs; medium‑term upskilling and reallocation of labor toward higher‑value judgments, advisory roles, and oversight.
• Capital and productivity
  • Adoption requires substantial capital investment (software, cloud compute, data pipelines). Early adopters may gain productivity and quality advantages, generating first‑mover benefits and potential market consolidation.
  • Improved audit quality and faster reporting can reduce information asymmetries in capital markets, potentially lowering cost of capital for firms with higher-quality audited financials.
• Market structure and competition
  • AI capabilities may become a competitive differentiator for audit firms and fintech providers; platforms that combine data access + analytics may capture larger market shares.
  • Smaller firms may face barriers to entry/adoption unless lower-cost AI tools or shared services emerge.
• Risk, regulation, and governance
  • Increased reliance on AI elevates regulatory needs: model validation standards, explainability requirements, liability rules, and data protection enforcement.
  • Mis-specified or poorly governed AI could generate systemic risks (e.g., consistent misclassification across many audits), so policy intervention on standards and audits of models is economically justified.
• Welfare and externalities
  • Better fraud detection and anomaly identification reduce fraud-related losses and improve investor trust—positive social welfare effects.
  • Transition costs (training, restructuring) and potential privacy harms are negative externalities that may require public support (training subsidies, standardized infrastructure).
• Policy and industry actions worth considering
  • Public subsidies or tax credits for auditor retraining and AI adoption to reduce transition frictions.
  • Minimum standards for model validation, explainability, and data governance in audit tools.
  • Support for shared, secure data infrastructure or APIs to lower integration costs for smaller auditors.
  • Research funding and regulatory sandboxes to test audit-AI approaches while monitoring systemic effects.

Overall, the paper implies that adoption of AI in auditing will shift economic value within the audit ecosystem (from manual processing to analytics and judgment), raise productivity and market benefits if governed well, but requires targeted investments and regulation to manage distributional and systemic risks.