Data analytics and AI raise audit quality and speed — but only when clients and auditors are technologically ready. Small firms and emerging markets lag behind due to skill gaps, resource limits and cybersecurity and regulatory hurdles.

Main Finding

Technology and data analytics (including AI/ML and federated learning) materially improve audit quality and efficiency, but benefits are context-dependent. Adoption is moderated by client and firm technological readiness, cybersecurity, regulatory/infrastructural constraints, and auditor skill gaps—especially in small and medium practices and emerging markets. Collaborative privacy-preserving AI approaches (federated and federated-continual learning) show promise for cross-firm anomaly detection without data sharing.

Key Points

• Evidence base: systematic review (PRISMA) of 260 records from IEEE Xplore, Web of Science, ScienceDirect, and Scopus; 10 empirical studies (2020–2025) included after screening.
• Positive impacts:
  • Big data & data analytics (BD&A/BDA/ADA) improve audit processes, auditor competence, audit quality, and audit review continuity.
  • Technology-Based Audit Techniques (TBATs) enable more audits, better risk identification, more recommendations, and fewer audit days.
  • Federated learning and federated continual learning frameworks can detect accounting anomalies across decentralized clients while preserving confidentiality; in non-i.i.d. settings some novel approaches outperform FedAvg.
• Moderators and complements:
  • Client technological readiness strengthens remote-audit effectiveness; auditor technology readiness can sometimes weaken it (mixed effects).
  • Cybersecurity moderates and strengthens the positive effects of ADA on audit quality and continuity.
  • Organizational factors (board composition) can moderate disclosure → audit quality links (e.g., Oman study: family board membership positive, royal membership negative).
• Limits and costs:
  • Skill gaps and hiring costs increase for effective TBAT implementation; small/medium audit practices need capacity building.
  • Some studies (Egypt) found BD&A had insignificant effects on audit fees.
  • Adoption uneven across sectors and countries; emerging markets face regulatory and infrastructure barriers.
• Research gaps: small number of comparable studies, methodological heterogeneity, limited longitudinal and cross-country evidence; authors call for longitudinal, comparative, and causal studies.

Data & Methods

• Review approach:
  • PRISMA-guided systematic review; databases searched: IEEE Xplore (72), Web of Science (48), ScienceDirect (47), Scopus (93); 260 initial records → 96 after deduplication → 65 full-text assessed → 10 included.
  • Risk-of-bias assessment: ROBINS-I tool.
  • Narrative synthesis used due to heterogeneity of methods and outcomes.
• Methods in included studies (examples):
  • Quantitative surveys with PLS-SEM / SmartPLS (Egypt, Indonesia, Indonesia public sector, Thailand) assessing BD&A/BDA/ADA adoption, acceptance, and impacts.
  • Content analysis with OLS and panel regressions on annual reports (Oman).
  • Mixed-methods surveys + interviews (Germany) examining TBATs.
  • Experimental/empirical evaluations on synthetic and real journal-entry datasets for federated learning approaches (Japan; US federated continual learning work).
  • Qualitative interviews and grounded-theory analysis for SME internal audit practices (South Africa).
• Geographical scope of included studies: Oman, Egypt, Germany, Thailand, Japan, USA, South Africa, Indonesia. Many studies from emerging markets or single-country contexts.

Implications for AI Economics

• Productivity and labor:
  • AI and analytics increase auditor productivity (more audits, fewer days) but shift required skill mixes — raising demand for data-literate auditors and likely increasing wage premia for those skills. Expect transitional dislocations and upskilling costs.
• Market structure and concentration:
  • High fixed costs of AI tools, data infrastructure, and specialized hiring can generate economies of scale, favoring larger firms and potentially increasing market concentration unless SMEs receive support or access to shared models.
  • Federated and collaborative model approaches reduce data-sharing frictions and could lower barriers for smaller firms to access high-quality anomaly-detection models, mitigating concentration pressures.
• Pricing and fees:
  • Early evidence suggests BD&A may not uniformly raise audit fees (insignificant effect in one Egyptian study), implying productivity gains could offset cost increases; rigorous cost–benefit and price-competition analyses are needed.
• Complementary investments and policy:
  • Cybersecurity and client-side digital readiness are complements to analytics — underinvestment in these areas limits returns to AI adoption. Public policy (standards, subsidies, training programs) can accelerate socially desirable adoption and reduce uneven distribution of benefits.
• Privacy, data externalities, and public-good models:
  • Federated/federated-continual learning frameworks create a mechanism to internalize positive externalities from shared model improvements while preserving privacy — relevant for designing industry-level infrastructure and regulation.
• Research & policy priorities for AI economics:
  • Quantify aggregate productivity gains, fee and wage effects, and welfare implications of audit AI adoption.
  • Model adoption diffusion across firm sizes and countries, including complementarities (cybersecurity, client digitalization) and regulatory constraints.
  • Conduct longitudinal and causal studies to assess dynamic effects on employment, firm entry/exit, competition, and audit market quality.
  • Evaluate policy interventions (training subsidies, shared model platforms, disclosure requirements) for efficiency and distributional impacts.

Suggested next steps for researchers/policymakers: build cross-country longitudinal datasets linking firm-level audit outcomes, technology adoption indicators, auditor labor market outcomes, and regulatory variables; pilot federated model infrastructures for SMEs and evaluate competitive and welfare effects.