Revenue REAP — Ireland's AI risk engine for human-supervised tax-compliance targeting

Ireland's Office of Revenue Commissioners (ORC) built one of the earliest dedicated tax-administration AI units by 2011, and the OECD credits it as the first EU tax administration to deploy social network analysis at scale. Its Risk Evaluation, Analysis and Profiling (REAP) system combines social network analysis, outlier detection on income-consumption patterns, a real-time VAT Risk Transaction Framework, and web-scraping to identify unreported gig-economy and online-marketplace income.

REAP is designed to target tax-compliance interventions more precisely while preserving human decision authority: Revenue's published policy explicitly prohibits AI from making tax-related decisions, with all REAP outputs treated as advisory.

Since 2019 a Virtual Digital Agent has handled routine tax-clearance calls, and since 2024 an internal RevAssist AI assistant has supported staff navigating Tax and Duty Manuals. A published Code of Practice for Revenue Compliance Interventions gives taxpayers procedural rights and appeal mechanisms.

In 2023, Revenue conducted approximately 290,000 audit and compliance interventions yielding €787m, which the 2023 Annual Report attributes to a more focused, data-analytics-driven approach to selecting taxpayers. This compares with 428,000 interventions yielding €813m in 2022, indicating higher yield per intervention as targeting improved.

Sustained for over 15 years with statutory backing and OECD recognition, Revenue's human-supervised AI posture is described as one of the most mature and transparent frameworks for public-sector AI in tax administration, though published figures do not isolate the AI-specific contribution to compliance yield.