In Indonesia’s Coretax rollout, firms reporting AI and cloud accounting use also report higher tax compliance, and external tax consultants amplify those reported gains; evidence comes from a cross-sectional survey of 150 PMA manufacturers and cannot establish causality.

Main Finding

In a survey of 150 fiscal leaders at foreign-invested (PMA) manufacturing firms in the Bekasi industrial cluster (Indonesia), AI adoption (β = 0.312, p = 0.001) and cloud accounting adoption (β = 0.285, p = 0.003) each significantly increase corporate tax compliance. External tax consultants significantly strengthen those effects (AI×Consultant β = 0.241, p = 0.011; Cloud×Consultant β = 0.198, p = 0.032). The full model explains ~68.4% of variance in tax compliance (R²adj = 0.684; Q² = 0.442).

Key Points

• Theoretical framing: integrates Technology Acceptance Model (TAM) with Contingency Theory and Stakeholder Theory; introduces the “Complexity Gap” — technology yields technical outputs that may be legally ambiguous without professional judgement.
• Direct effects:
  • AI adoption → higher tax compliance (β = 0.312, f² = 0.185).
  • Cloud accounting adoption → higher tax compliance (β = 0.285, f² = 0.210).
• Moderation effects:
  • External tax consultants amplify the AI → compliance link (interaction β = 0.241, f² = 0.095).
  • External tax consultants amplify the cloud → compliance link (interaction β = 0.198, f² = 0.112).
• Control variable (firm size, log assets) was not significant.
• Measurement: constructs measured by 3 indicator items each on 5‑point Likert scales (AI, cloud, consultants, tax compliance). Reliability and validity checks (loadings > .70, CR .85–.87, AVE .66–.69, HTMT < .85) support the measurement model.
• Common method bias was assessed; VIFs between 1.412–2.150 (below conservative thresholds).

Data & Methods

• Population & sample: PMA manufacturing firms in Bekasi (MM2100, Jababeka, EJIP). Purposive sampling required firms to have both AI and cloud accounting adoption and to retain external tax consultants. Final N = 150 (response rate 81.1% of 185 distributed).
• Data collection: structured electronic questionnaires (Feb–May 2026).
• Analysis: PLS-SEM (SmartPLS 4) with 5,000 bootstrap resamples. Evaluated measurement model (convergent & discriminant validity) and structural model (path coefficients, R², Q², f²).
• Key model statistics: R²adj (tax compliance) = 0.684; Q² = 0.442; PLSpredict showed better RMSE than naive benchmark.
• Limitations noted by authors:
  • Purposive inclusion criterion (firms had to have both technologies and external consultants) restricts generalizability, especially to firms without such resources.
  • Cross-sectional, self‑reported survey data (single source) — although procedural/statistical CMB checks were applied.
  • Single-country, single-industry cluster context (Bekasi PMA manufacturing).

Implications for AI Economics

• Complementarity of capital and expert labor: Results show AI and cloud tech raise compliance, but their effectiveness is meaningfully augmented by specialist human advisors. This evidences a key economic complementarity — investments in automation increase demand for high‑skill advisory services rather than fully substituting them.
• Market for expert services: As tax authorities (e.g., Indonesia’s Coretax) automate audit and compliance monitoring, firms upgrading accounting tech are more likely to hire (or deepen ties with) external tax consultants. Expect upward pressure on demand, fees, and market concentration for high‑quality tax advisory services—especially for cross‑border/fiscal complexity.
• Redistribution of tasks and labor dynamics: Routine data processing and anomaly detection move toward AI/cloud; interpretive, judgmental, and regulatory liaison tasks remain human-intensive. This supports models where AI augments rather than replaces specialized professional work in regulated domains.
• Compliance costs and firm heterogeneity: Digitalized tax administration raises fixed compliance/technology costs. Large or PMA firms can internalize those costs and capture compliance gains; smaller firms or those lacking advisory access may face disproportionate burdens — potential policy concern for distributional effects and market entry.
• Enforcement and tax revenue economics: Better-aligned firm tech + advisory ecosystems should reduce reporting errors and detection lags, potentially narrowing tax gaps. However, strategic behavior could shift (e.g., more sophisticated tax planning) and tax administrations may need to adapt enforcement models.
• Research agenda for AI economics:
  • Quantify welfare/tradeoffs: cost of technology + advisory vs. reduction in compliance errors and enforcement costs.
  • Measure labor reallocations: demand for advisory labor, wage effects, and skill premiums in tax/accounting labor markets.
  • Causal and longitudinal studies: track firms before/after Coretax integration and tech adoption to estimate dynamic impacts and persistence.
  • Heterogeneity analysis: how effects differ by firm size, industry, cross‑border exposure, and consultant quality.

Practical takeaway: In digitalized tax regimes, investing in AI and cloud accounting raises compliance outcomes, but firms capture the full governance benefit only when technical systems are paired with professional tax expertise — policymakers should consider support mechanisms (standards, subsidized advisory for SMEs, interoperability rules) to avoid uneven compliance burdens.