In 2023, tax administration leaders convened at the FTA Plenary in Singapore, underscoring a collective ambition: harness AI responsibly within tax systems. This ambition formed the core of Project D, launched in December 2023, aimed at fostering a “trustworthy” AI integration in tax administration. As digital transformation accelerates, taxpayers demand not only efficiency but also transparency and customization akin to private-sector experiences. Tax administrations (TAs), striving to meet these expectations, recognize AI as an enabler of efficient, transparent, and responsive tax systems. Project D, drawing from over 40 experts across 16 countries, has identified strategic guidelines for adopting trustworthy AI through a comprehensive Trustworthy AI Framework.
Transforming Tax Administration with AI
Tax administrations have historically employed machine learning to optimize compliance and routine tasks. However, the advent of more powerful AI—encompassing deep learning and generative AI—has broadened the spectrum of possibilities. AI, in essence, empowers machines to assist in decision-making that typically requires human intervention. Recent advancements promise to enhance traditional compliance processes, as well as pioneer applications in customer service and internal management.
AI Use Cases in Tax Administration
Project D’s inventory of 36 AI use cases underscores this shift. Traditional AI applications predominantly enhance compliance, relying on historical data to streamline processes. Meanwhile, generative AI is explored for internal management and customer service, though most implementations remain experimental. These AI initiatives are distributed across four primary sectors:
- Internal Management and Operations: Automating routine HR tasks, facilitating training through interactive visualizations, and enhancing data categorization for operational efficiency.
- Customer Service: Language translation, setting up payment plans, and protecting taxpayer accounts from identity theft or abuse.
- Compliance and Treatment: Identifying non-compliance patterns, detecting statistical anomalies, and protecting taxpayers from fraud.
- External Ecosystem: Standardizing tax codes across jurisdictions, pre-checking compliance, and monitoring industry compliance tools.
Production Status of AI Use Cases
Across the 16 participating tax administrations, a survey revealed over 150 AI use cases in active deployment, with an additional 315 proposed. While traditional AI comprises most operational use cases, generative AI is largely experimental, suggesting a cautious but promising expansion. Notably, compliance-focused applications lead the way in active AI use, while customer service remains an emerging area, perhaps reflecting a need for additional safeguards before engaging taxpayers directly with AI solutions.
Strategic Frameworks and Governance for AI in Tax Administration
Countries adopting AI in tax administration vary in approach, grouped into three segments: Advancing, Emerging, and Cautious. The “Advancing” group demonstrates the highest volume of active AI cases, while “Emerging” and “Cautious” countries are gradually expanding their AI footprint. These segments highlight the diversity in AI integration speed and governance.
Centralized Oversight and Governance
To standardize and monitor AI initiatives, some administrations have established dedicated senior roles and centralized hubs. This infrastructure supports governance and ensures compliance with ethical standards. Survey data indicates that advanced administrations are more likely to apply formalized guidelines, engage external stakeholders, and implement ethical protocols. For sustainable integration, Project D advocates for frameworks that align with national regulations and ethical considerations.
AI Assurance Framework
A cornerstone of Project D’s output is the AI Assurance Framework, designed to incorporate comprehensive assurance checks at each phase of the AI lifecycle. This framework emphasizes accountability, transparency, fairness, and data security, providing actionable checkpoints for all AI projects. Key elements of this framework include:
- Legal and Ethical Compliance: Ensuring AI applications align with national policies and ethical standards.
- Outcome Evaluation: Establishing measurable criteria to assess AI-driven results, verifying that AI investments achieve desired outcomes.
- Human Oversight: Maintaining a “human-in-the-loop” approach, ensuring humans can oversee, intervene, and understand AI-driven processes.
- Stakeholder Communication: Informing users about AI interactions, enabling them to provide feedback and report vulnerabilities.
- Bias and Privacy Protections: Implementing bias assessments and privacy protocols to protect taxpayer data and ensure fair treatment.
Future Directions for AI in Tax Administration
As Project D moves forward, continued engagement with stakeholders is key to refining the AI Assurance Framework. The project team emphasizes a commitment to ethical AI that aligns with international standards and addresses each phase of AI implementation—from foundational scoping to operational maintenance. As AI continues to evolve, tax administrations will adapt, leveraging trustworthy AI to meet the expectations of a digital-first world.
To keep updated on news, visit our Global News Page.
Don’t miss our most recent updates and articles; follow us on LinkedIn.
