从理论到实践：税务管理的战略性人工智能集成模型（英）2025

FROM THEORY TO PRACTICE:A STRATEGIC AI INTEGRATIONMODEL FOR REVENUEADMINISTRATIONSRaúl Junquera, Ivan Krsul, Vladimir Calderón,Joey Ghaleb, and Cristian Lucas © 2025 International Bank for Reconstruction and Development / The World Bank1818 H Street NWWashington DC 20433Telephone: 202-473-1000Internet: www.worldbank.orgThis work is a product of the staff of The World Bank with external contributions. The findings,interpretations, and conclusions expressed in this work do not necessarily reflect the views of TheWorld Bank, its Board of Executive Directors, or the governments they represent.The World Bank does not guarantee the accuracy, completeness, or currency of the data includedin this work and does not assume responsibility for any errors, omissions, or discrepancies in theinformation, or liability with respect to the use of or failure to use the information, methods, processes,or conclusions set forth. The boundaries, colors, denominations, links/footnotes and other informationshown in this work do not imply any judgment on the part of The World Bank concerning the legal statusof any territory or the endorsement or acceptance of such boundaries. The citation of works authoredby others does not mean the World Bank endorses the views expressed by those authors or the contentof their works.Nothing herein shall constitute or be construed or considered to be a limitation upon or waiver of theprivileges and immunities of The World Bank, all of which are specifically reserved.Rights and PermissionsThe material in this work is subject to copyright. Because The World Bank encourages dissemination ofits knowledge, this work may be reproduced, in whole or in part, for noncommercial purposes as longas full attribution to this work is given.Any queries on rights and licenses, including subsidiary rights, should be addressed to World BankPublications, The World Bank Group, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-2625; e-mail:pubrights@worldbank.org. TABLEOF CONTENTSAbstract1. Introduction2. A Brief Overview of AI3. Use of AI in Revenue Administration3.1Objectives and Goals of AI Integration into a Revenue Administration3.2Potential Impact of AI on the Organizational Structure and Operations ofRevenue Administrations3.3The Importance of Human-AI Collaboration3.4The Evolutionary Nature of AI Development3.5Challenges in Implementing AI Without a Framework4. Data Governance and its Relation to Synthetic Data5. A Strategic Framework for AI Integration5.1Key Components of the Framework5.2Framework Genesis5.3Phases5.3.1Inception Phase5.3.2Consolidation Phase5.3.3Optimization Phase5.4Framework Roadmap vi14789111214151718192020222224 6. Use Cases6.1AI in Tax Administration6.2AI in Customs Administration7. ConclusionReferences 2729384345 ABSTRACTThis paper presents a comprehensive strategic framework for integrating ArtificialIntelligence (AI) into revenue administrations. The framework addresses thechallenges of implementing AI without a structured approach and emphasizes theimportance of human-AI collaboration. It proposes a three-phase implementationstrategy—inception, consolidation, and optimization—designed to incrementallybuild capacity, establish governance structures, and optimize AI systems overtime.By following this framework,revenue administrations can effectivelyharness the power of AI to enhance efficiency, improve taxpayer services, andstrengthen compliance efforts while maintaining public trust and transparency. INTRODUCTIONArtificialIntelligence(AI)has rapidly evolvedfrom a theoretical concept to a practical reality invarious sectors, including tax administration andcustoms. The latest Organisation for Economic Co-operation and Development (OECD) study on TaxAdministration 2024 reveals that over 50 percent1. of tax administrations are now using AI in somecapacity, particularly in areas such as taxpayerassistance, risk assessment, and fraud detection.Thiswidespread adoption signifies a pivotalshift in how tax authorities operate and interactwith taxpayers. Butthe implementation of AI in revenueadministrations is not without challenges. Unliketraditional software development, AI integrationrequires a nuanced approach that considers notonlytechnical aspects but also ethical,legal,operational, and behavioral implications. Revenueadministrations must navigate complex issues suchas data governance, model transparency, and thepotential for bias, all while maintaining public trustand ensuring fair treatment of taxpayers. Anothercritical consideration is the explainability of AImodelsused in tax administration.ExplainableAI (XAI) refers to AI systems that provide clear,understandablereasons for their decisions,which is essential for maintaining transparency,fairness,and trust in automated processes.Revenue administrations need to ensure that AI-drivendecisions,particularly those related tocomplianceand enforcement,are interpretablenot only by technical staff but also by the generalpublic and