ZATCA makes a breakthrough in post-clearance audit through the use of AI
17 November 2026
By the Zakat, Tax and Customs Authority (ZATCA)In 2024, the Zakat, Tax and Customs Authority (ZATCA) led a major step in Customs modernization by launching the Holistic Post-Clearance Audit (PCA) Engine. Developed to support the “Clearance within Two Hours” initiative, the engine enables faster and more accurate assessment of trader compliance levels.
The Holistic PCA Engine strengthens post-clearance enforcement by addressing three areas: understatement of Customs value, product misclassification, and misuse of Customs exemptions. The consolidation of multiple data streams into a single analytical engine allows ZATCA to detect risk patterns more effectively and deploy audit resources in a data-driven, prioritized manner.
Background and rationale
Conducting post-clearance audit (PCA) is a means to measure and improve trader compliance. The PCA process itself consists, after Customs has released the cargo, of a structured examination of the relevant commercial data, sales contracts, financial and non-financial records, physical stock and other trader assets. A well-conceived process, PCA is recognized as the best method to recover revenues lost through under-declaration, misclassification and misdeclaration of origin. Implementing a robust PCA process also enables a Customs administration to develop a trader segmentation programme to distinguish between traders on the basis of their respective compliance levels, thus contributing to effective risk management and reducing the need for Customs intervention at the border for compliant traders. In addition, it increases the transparency of the administration’s taxation process, which significantly contributes to increased trust in the Customs administration from the business sector.
Before the introduction of the Holistic Post-Clearance Audit Engine, the ZATCA PCA process relied heavily on traditional, rule-based risk systems. These systems focused primarily on analysing Customs declarations data using expert-defined rules, limiting the ability to capture evolving trade behaviours and complex patterns of non-compliance.
Challenges faced were fragmented risk models operating independently for different risk areas; overreliance on declaration-level data; low adaptability to emerging fraud trends; and inefficient use of audit capacity. To overcome these limitations, the Zakat, Tax and Customs Authority envisioned a unified, machine-learning-driven PCA engine capable of learning from millions of historical declarations and diverse data sources.
The Holistic PCA Engine
The Holistic PCA Engine represents a transformation from segmented analysis to an integrated, 360-degree view of import transactions. It combines all risk dimensions (valuation, classification, and exemptions) into a single scoring framework where each declaration is assigned a composite risk score, with indicators highlighting the specific risk areas. This provides auditors with clarity on both where and why a transaction is considered high-risk.
The engine draws on a broad ecosystem of data sources, spanning operational records, taxpayer profiles and external trade information. By integrating these diverse data streams, the engine uncovers non-compliance patterns that would be difficult to detect through any single source on its own. The development of the PCA Engine represents a major step in ZATCA’s use of data analytics, achieved through the integration of machine learning (ML) and natural language processing (NLP). A range of ML models were tested to achieve optimal accuracy and robustness, serving four key purposes: to detect patterns and understatements relating to peer importers; to recognize misclassification patterns; to estimate potential revenue uplift; and to continuously refine its predictive accuracy through feedback loops. In parallel, NLP algorithms analyse declaration descriptions to detect potential HS code misclassifications, thereby reducing manual inspection efforts and improving overall compliance.
The outcome is a dynamic, self-improving model that aligns with international best practices, while remaining tailored to Saudi Arabia’s Customs environment.
Implementation and impact
Since its deployment in 2024, the Holistic PCA Engine has delivered strong operational and strategic results. Detection accuracy has improved, with a higher hit rate of flagged cases compared to older models. Audit efficiency has increased as auditors now prioritize cases based on expected revenue gains and risk indicators. The engine has also helped prevent multi-million SAR in potential revenue leakage, strengthening revenue protection. Compliance behaviour has improved through transparent, data-driven audit criteria that have raised trader awareness. Finally, process efficiency has advanced by enabling faster pre-clearance decisions without compromising control. Overall, the engine strengthens compliance assurance, while supporting ZATCA’s trade facilitation goals.
Future developments
Building on its success, the Zakat, Tax and Customs Authority is advancing several next-phase enhancements. These include real-time data integration with ports, logistics providers, and international databases; predictive compliance scoring to encourage voluntary compliance; broader use of AI to detect complex fraud patterns; and continuous learning loops to align engine performance with audit outcomes. Together, these upgrades will strengthen ZATCA’s ability to protect revenue, facilitate trade, and maintain a fair and transparent Customs environment.
More information
Contact the author