Machine learning, NLP and robotic process automation are raising audit efficiency and fraud detection globally, but Indonesia remains behind the innovators and early adopters. High investment costs, data-privacy risks and limited auditor competencies are the chief obstacles to wider implementation.

Main Finding

AI technologies—especially machine learning (ML), natural language processing (NLP), and robotic process automation (RPA)—substantially improve audit efficiency, accuracy, and quality, and enable risk‑based and continuous auditing. However, widespread implementation faces non‑technical barriers (auditor competencies, high investment costs, ethical and data‑security concerns). Globally, adoption sits at innovators→early adopters; Indonesia is largely at the early‑majority stage. Successful implementation requires aligning technological readiness, auditor acceptance (TAM), and organizational diffusion (Rogers).

Key Points

• Scope: Systematic Literature Review (SLR) of research on AI in financial audit, period 2020–2025; final corpus = 15 peer‑reviewed articles from Scopus (Q1–Q4) and Sinta (S1–S2).
• Dominant AI types: Machine learning, NLP, RPA, predictive analytics, some generative AI and deep learning.
• Main benefits reported (examples & effect sizes cited in corpus):
  • Audit efficiency/productivity gains: +40% (Deloitte/Adolph cited; Abdullah & Almaqtari noted +40%); automation of ~60% routine audit processes (Zhou & Liu).
  • Fraud detection / accuracy: deep learning detection accuracy ~92% in experiments (Issa et al.); materiality/risk detection up to +35% faster (Bao et al.).
  • Administrative burden reduction: RPA can reduce repetitive tasks by ~70% (Wulandari).
  • Correlation evidence: positive correlations between AI adoption and audit quality (e.g., Othman reported r ≈ +0.72; Noordin et al. reported +67% improvement).
  • Auditor perceptions: high perceived usefulness increases adoption intent (TAM); Indonesian survey: preference for assisted AI (85%) over fully autonomous (45%) (Luthfi & Purwati).
• Key challenges:
  • Skill gaps and need for reskilling; auditor training a principal barrier.
  • High upfront investment and infrastructure costs for AI deployment.
  • Ethical concerns, data privacy/security, algorithmic bias, regulatory uncertainty.
  • Heterogeneous adoption: Big firms innovate faster (Big Four), smaller KAPs lag (approx. 16% laggards).
• Research gaps identified:
  • Limited cross‑country comparative empirical work on costs, privacy, and readiness in developing economies.
  • Sparse technical implementation studies within Indonesia; limited attention to regulatory and market‑structure effects.

Data & Methods

• Review approach: Systematic Literature Review following PRISMA 2020.
• Search sources: Scopus (Q1–Q4) and Sinta (S1–S2).
• Search syntax: combinations of ("artificial intelligence" OR "AI" OR "machine learning" OR "deep learning" OR "NLP") AND ("audit" OR "auditing" OR "financial statement audit").
• Inclusion criteria: 2020–2025; English or Indonesian; peer‑reviewed journals (Scopus Q1–Q4 or Sinta 1–2); full text accessible; focused on AI in audit.
• Screening flow: identification = 2,370 records → title/abstract screening → 1,565 → eligibility filtering eliminated 270 → final included articles = 15.
• Analytical method: Thematic analysis (Braun & Clarke) with PICO framework to structure focus:
  • Population: articles on AI in audit
  • Intervention: ML, NLP, RPA
  • Comparison: AI‑based vs conventional audit
  • Outcomes: efficiency, quality, fraud detection, implementation challenges
• Thematic findings grouped into 3 categories: (1) trends & AI technologies in audit, (2) benefits & implementation challenges, (3) global vs Indonesian research gaps and adoption stages.

Implications for AI Economics

• Labor demand and skill composition
  • Automation of routine audit tasks implies reduced demand for repetitive junior tasks but increased demand for higher‑skilled roles (modelers, data analysts, AI auditors). Expect skill‑biased technical change: upskilling/reskilling is critical; potential rise in skill premium in auditing labor markets.
• Productivity and firm performance
  • Reported productivity gains (e.g., ~40% improvements) suggest strong micro‑level returns to AI investment in auditing. These productivity boosts can cascade to faster, higher‑quality financial reporting, affecting firm valuation and reducing informational frictions in capital markets.
• Investment and adoption dynamics
  • High fixed costs imply economies of scale: large audit firms (Big Four) internalize costs more easily, accelerating concentration and widening productivity gaps between large and smaller firms. This leads to non‑uniform diffusion and potential market structure shifts.
• Diffusion, externalities, and inequality
  • Heterogeneous adoption across countries (innovators vs early majority) risks widening productivity and compliance gaps between advanced and developing markets. Cross‑firm and cross‑country externalities (data standards, shared tooling) can either mitigate or amplify such disparities.
• Regulation, trust, and market efficiency
  • Ethical/data privacy concerns and regulatory uncertainty are economic frictions that can delay adoption and generate compliance costs. Conversely, credible regulation and audit standards for AI can lower adoption uncertainty and reduce negative externalities (e.g., biased audit outcomes).
• Macroeconomic & financial stability considerations
  • Real‑time / continuous auditing enabled by AI could improve monitoring, reduce information asymmetry, and influence asset pricing and systemic risk assessment. But simultaneous adoption without robust safeguards could create correlated model‑risk across firms.
• Policy and research priorities (economic perspective)
  • Promote targeted training subsidies and certification to reduce transition costs and labor frictions.
  • Support modular/open infrastructure and shared platforms to lower entry costs for smaller audit firms and reduce market concentration.
  • Empirical research needed on aggregate welfare impacts: (i) quantifying productivity spillovers to clients and markets, (ii) modeling labor reallocation and wage effects, (iii) cost‑benefit analyses of AI adoption across firm sizes and countries, and (iv) regulatory scenarios assessing systemic model risk.
• Suggested economic modeling directions
  • Incorporate AI adoption as a diffusion process in firm‑level production functions to estimate TFP gains.
  • Use heterogeneous‑firm frameworks to study market concentration and equilibrium wages in the audit sector.
  • Develop macro models to simulate effects of continuous auditing on information flows and asset pricing.

Summary: The paper provides a concise, evidence‑based SLR showing AI materially reshapes auditing productivity and quality but raises distributional, regulatory, and adoption‑cost issues central to AI economics. Future work should quantify aggregate gains, labor market transitions, and policy levers to manage distributional and systemic risks.