LEVERAGING EXPLAINABLE AI IN ACCOUNTING SYSTEMS TO IMPROVE TRANSPARENCY, AUDIT TRAILS, AND STAKEHOLDER TRUST IN FINANCIAL DECISIONS

The deployment of artificial intelligence (AI) in accounting information systems has significantly improved
automation in areas such as transaction classification, fraud detection, and financial forecasting. However, the
increasing reliance on complex machine learning models particularly deep learning and ensemble methods has
introduced opacity that undermines auditability, regulatory compliance, and stakeholder confidence. This study
examines how Explainable Artificial Intelligence (XAI) can be systematically embedded within accounting workflows
to enhance transparency, traceability, and trust. At a systems level, XAI techniques such as SHAP (Shapley Additive
Explanations), LIME (Local Interpretable Model-Agnostic Explanations), and rule-based surrogate models are applied
to generate interpretable outputs for automated ledger postings, anomaly detection flags, and predictive risk scores.
These methods enable auditors and financial managers to reconstruct decision pathways, validate model behavior
against accounting standards (e.g., IFRS and GAAP), and maintain verifiable audit trails. Furthermore, integrating
XAI dashboards within enterprise resource planning (ERP) systems supports real-time justification of financial
decisions and facilitates regulatory inspections. Empirical evaluation demonstrates that XAI-enhanced models reduce
false positives in fraud detection and improve audit efficiency by enabling targeted investigations. The study concludes
that embedding explainability into AI-driven accounting systems is essential for aligning automation with governance,
enhancing accountability, and strengthening stakeholder trust in data-driven financial decision-making environments.