The integration of artificial intelligence (AI) into tax and customs systems is no longer a question of if but when. As technology evolves, so too do the expectations of AI in tax systems and possibilities for its practical application in area of tax governance. The International Monetary Fund’s (IMF) Technical Guidance Note, Understanding Artificial Intelligence in Tax and Customs Administration, offers a timely and robust framework for governments moving through this complex yet transformative terrain. By addressing the potential and the pitfalls of AI, the note lays a foundation for countries to thoughtfully consider how AI can enhance efficiency, reduce fraud, and improve public service delivery.
As The RegTech currently joins the African Tax Administration Forum Annual Meetings in Kigali, Rwanda (December 2nd – 6th, 2024), we are raising an important topic of AI implementation in the digital efforts of tax authorities across the continent. While the IMF technical note provides a valuable starting point, the journey to AI adoption demands a nuanced and proactive approach. Success will require understanding the institutional, ethical, and technological landscapes surrounding AI. Obviously, it is up to each country to chart its course, balancing ambition with practicality. By building a strong foundation of knowledge and readiness, tax and customs administrations can utilize AI to achieve their strategic objectives.
5 Key Takeaways
- AI’s Transformative Role in Governance: Artificial intelligence is reshaping tax and customs systems, with advanced capabilities like generative reasoning and pattern recognition offering opportunities to streamline operations, enhance compliance, and detect fraud. Its integration represents a critical step for governments aiming to modernize and address inefficiencies in public administration.
- Tailored Strategies for Diverse Maturity Levels: The success of AI implementation depends on aligning strategies with institutional readiness. While advanced economies focus on optimization and system integration, developing nations must first establish foundational infrastructure and governance to support sustainable digital transformation.
- Leadership and Accountability Are Crucial: Strategic leadership drives AI adoption. Whether through a modernization director, CIO, or cross-functional teams, clear accountability is essential. Effective management of AI systems, including partnerships with external providers, ensures alignment with organizational goals and mitigates risks such as data bias or flawed decision-making.
- Ethical and Legal Considerations in AI Adoption: AI’s adoption in tax and customs raises significant ethical questions around bias, privacy, and accountability. Governments must address these challenges by fostering transparency, adhering to emerging regulatory frameworks like the EU AI Act, and implementing rigorous oversight to maintain fairness and public trust.
- A Future-Focused Approach to Workforce and Technology: As AI evolves, tax and customs agencies must invest in workforce development to cultivate collaboration between humans and AI systems. Modular and scalable technological solutions will help administrations adapt to changing demands, unlocking efficiencies and positioning them for long-term success.
The Case for AI in Tax and Customs
AI is not a new concept in governance. Since the 1970s, systems like expert rules and machine learning have enhanced enforcement, compliance, and fraud detection. What sets today’s AI apart is its capacity for generative reasoning, natural language processing, and pattern recognition on a scale previously unimaginable. For tax and customs administrations, this means opportunities to streamline risk assessments, analyze taxpayer behavior, and combat complex fraud schemes.
Globally, the stakes for digital transformation in governance are high. Advanced economies are already reaping the benefits of AI-driven solutions, from automating tax filings to improving compliance rates. Meanwhile, many developing nations struggle with basic infrastructure and access, creating a growing disparity. AI offers the potential to bridge this divide, but only if its implementation is carefully tailored to each country’s unique challenges and needs.
The IMF’s guidance is particularly relevant for nations at varying levels of digital maturity. While it provides a broad framework, it emphasizes the importance of grounding AI strategies in existing institutional capabilities. Without this alignment, even the most advanced AI systems risk becoming underutilized or misapplied.
Building the Knowledge Base
Before adopting AI, tax and customs administrations must develop a deep understanding of the technology and its implications. This begins with demystifying AI’s capabilities and limitations. AI is often misunderstood as a one-size-fits-all solution, but its efficacy depends on specific use cases, data quality, and institutional readiness.
One of the most critical early steps is assessing where AI can deliver the most value. Risk management, for instance, is a well-established domain for AI in tax systems. Machine learning models can analyze historical data to identify high-risk taxpayers or importers, enabling targeted audits and investigations. Similarly, AI can support operational efficiency by automating routine tasks, freeing up human resources for higher-value activities.
However, understanding AI’s limitations is equally important. For example, AI models can perpetuate bias if trained on skewed data, leading to unfair outcomes. Similarly, a lack of explainability in AI systems can undermine public trust, particularly when decisions have financial or legal consequences. Addressing these concerns requires not only technical expertise but also a commitment to transparency and accountability.
Strategic Leadership and Accountability
The integration of AI into tax and customs systems must be driven by strategic leadership. Deciding who will spearhead AI initiatives is a critical decision. Options include appointing a director for modernization, leveraging the expertise of a Chief Information Officer (CIO) or Chief Data Officer (CDO), or forming cross-functional working groups.
Leadership roles must balance technological, operational, and ethical considerations. For example, a director for modernization may focus on aligning AI initiatives with broader digital transformation goals, while a CDO might emphasize data governance and quality. Regardless of the approach, accountability should be clear and aligned with the organization’s strategic priorities.
Governments must also navigate the complexities of AI service supply chains. Many AI systems rely on external providers, creating risks related to data quality, intellectual property, and liability. For instance, if a third-party vendor supplies biased training data, this could result in flawed enforcement decisions. Careful contracting and oversight are essential to mitigate these risks and establish trust in AI systems.
The Ethical and Legal Dimensions
AI adoption in tax and customs administration raises significant ethical and legal questions. Bias, privacy, and accountability are at the forefront of these concerns. Bias in AI systems can lead to discriminatory outcomes, undermining the principles of fairness and equity. Mitigating bias requires rigorous testing, diverse datasets, and continuous monitoring.
Privacy and data governance are equally critical. Tax and customs administrations handle sensitive information, making strong safeguards essential. Clear policies on data use, storage, and sharing can help balance operational efficiency with individual rights.
Emerging regulatory frameworks provide valuable guidance. The EU AI Act, for example, emphasizes risk-based approaches and transparency, while the Council of Europe’s Draft Framework Convention on AI promotes accountability and human oversight. By aligning AI strategies with these frameworks, governments can improve levels of trust and compliance.
AI in Tax: Promoting Responsible Use
As regulatory frameworks for artificial intelligence (AI) continue to develop, senior leaders in tax and customs administrations can take proactive measures to ensure the responsible use of AI. By implementing a set of practical management practices, these institutions can promote certainty, transparency, and accountability across their AI initiatives. Below are ten key actions that can help achieve these objectives:
1. Implementing a Formal Policy on AI
Developing a comprehensive AI policy that outlines the institution’s approach to AI adoption and use. This policy should include leadership expectations, ethical principles, compliance requirements, transparency standards, governance structures, and training provisions. The policy provides a foundation for ensuring alignment with institutional values and external regulations.
2. Sensitizing Staff to AI and Individual Duty of Care
Conducting targeted training to educate staff, especially senior leadership, on the benefits, limitations, and responsibilities associated with AI. Staff should understand how AI systems function, potential biases, the importance of human oversight, and the ethical implications of AI in decision-making processes.
3. Building from Strong Fundamentals with an AI Strategy
Creating a dedicated AI strategy that aligns with broader institutional reform goals while prioritizing core operational strengths. This strategy should balance advanced AI development with the ongoing need to address weaknesses in fundamental tax and customs operations.
4. Establishing a Formal Inventory of AI Use Cases
Documenting all existing and planned AI applications in a structured inventory. Each use case should detail the type of AI technology used (e.g., machine learning, natural language processing) and its intended purpose. This inventory serves as a foundation for oversight, risk assessment, and strategic planning.
5. Subjecting Use Cases to Legal and Ethical Review
Evaluating AI use cases for compliance with legal standards and ethical principles. A systematic risk assessment methodology should be employed to identify high-risk applications, which require ongoing scrutiny and mitigation strategies.
6. Where Appropriate, Keeping a “Human-in-the-Loop” (HITL)
Guaranteeing human involvement in AI-driven processes where decisions may have significant consequences. HITL ensures direct human oversight, while “Human-on-the-Loop” (HOTL) provides supervisory control. Both approaches safeguard against fully automating sensitive processes.
7. Risk-Assessing New Use Cases Prior to Introduction
Conducting a thorough evaluation of new AI applications before implementation. This includes assessing financial costs, potential risks, and the broader impact on compliance and operational efficiency.
8. Regularly Publishing the Use Case Inventory
Promoting transparency by publishing a filtered version of the AI use case inventory. Exclude sensitive applications related to national security or those with negligible impact. This practice fosters trust and invites feedback from stakeholders.
9. Prominently Disclosing the Use of AI in Operations
Clearly informing stakeholders when AI systems are in use. For instance, virtual assistants should disclose their non-human nature, and AI-generated content should include clear labeling to distinguish it from human-produced outputs.
10. Evaluating Use Cases for Performance and Intent
Continuously monitoring AI use cases to ensure they meet their intended goals and do not produce unintended consequences. Regular reviews should assess both the effectiveness and the alignment of AI outcomes with institutional objectives.
By implementing these actions, tax and customs administrations can build robust frameworks for responsible AI use, ensuring alignment with ethical principles, operational excellence, and public trust.
AI in Tax: Integration Through Structured Change
The introduction of AI in tax and customs systems benefits from applying established digital service management concepts. Leveraging a structured lifecycle approach ensures that AI projects are seamlessly integrated into existing IT operations while addressing unique complexities associated with artificial intelligence.
A tailored framework for managing AI, as depicted in the structured lifecycle model, emphasizes the importance of “control gates.” These gates act as checkpoints where significant changes in AI use cases are evaluated and approved by governance structures, typically in the form of committees like Change Advisory Boards (CABs). This approach balances project management, technical development, and IT operations to minimize risks and maximize alignment with institutional goals.
Life Cycle Approach
At the outset, AI use cases are presented to a CAB for an initial review. Upon endorsement, the project formulation stage incorporates appropriate change controls that match the complexity and scale of the AI initiative. This is followed by a rigorous assessment of legal and ethical risks, ensuring that the AI use case adheres to the institution’s operational and regulatory standards.
Throughout the lifecycle, critical stages such as the System Requirements Review (SRR), Test Readiness Review (TRR), and User Readiness Review (URR) ensure that AI systems meet design expectations and are ready for deployment. Post-implementation evaluations, conducted through Post-Implementation Reviews (PIRs) and Project Review Boards (PRBs), assess the performance and outcomes of AI initiatives, determining their long-term value or the need for deactivation.
This lifecycle approach is particularly well-suited for mid- to large-sized administrations. By integrating AI into the portfolio of IT services through established methodologies, administrations can ensure that AI projects are systematically developed, responsibly deployed, and continuously evaluated, safeguarding both operational efficiency and public trust.
Planning for the Future
The pace of AI development shows no signs of slowing. Over the next 5, 10, or 15 years, AI is expected to become integral to tax and customs operations. Agencies that fail to adopt AI risk falling behind, while those that do can unlock new efficiencies and capabilities.
Preparing for this future involves more than just adopting the latest technologies. Agencies must invest in workforce development, equipping staff with the skills to manage and interpret AI systems. This includes not only technical training but also taking care of a culture of collaboration between human and machine intelligence.
Flexibility is another key consideration. As AI evolves, so too will its applications in tax and customs. Agencies should adopt modular and scalable solutions that can adapt to changing needs and priorities. By doing so, they can stay ahead of technological trends while avoiding costly system overhauls.
Tailoring Strategies to Maturity
AI strategies must be tailored to each institution’s level of digital maturity. For agencies with advanced digital infrastructures, the focus might be on integrating AI into existing systems and optimizing processes. For those at earlier stages of digital transformation, foundational investments in data quality, infrastructure, and governance are essential.
Leapfrogging is an attractive concept for less mature institutions, but it comes with risks. Skipping foundational steps can lead to fragmented systems and missed opportunities for long-term growth. By building strong digital fundamentals, agencies can create a stable platform for AI adoption.
