Artificial intelligence is reshaping industries worldwide. According to recent research from the International Data Corporation (IDC), AI may contribute USD $19.9 Trillion to the global economy through 2030 and could help drive 3.5% of global GDP. Beyond the more obvious economic benefits of AI, the technology is often cited as having the potential to revolutionise administrative processes around the world, such as supply chain management, healthcare administration, regulatory compliance, and all aspects of taxation, from compliance to advisory and dispute resolution.
This article discusses how AI could reshape not just tax compliance but tax administration as a whole, by automating audits, assisting with record-keeping, detecting fraud, flagging risks, streamlining controversies, and improving revenue collection. With the ability to analyse vast datasets in real-time, AI may be able to identify discrepancies, flag potential tax risks, and even provide predictive insights to governments. If implemented effectively, these advancements could make tax systems more efficient and help reduce the administrative burden on taxpayers while improving compliance. However, the extent to which AI will deliver on these promises is yet to be showcased by any government.
The implementation of AI in taxation requires careful governance. Ensuring transparency, mitigating bias, and maintaining taxpayer privacy are critical challenges that tax authorities must address. Further, while AI is frequently promoted as a tool for efficiency, questions remain as to whether AI-driven decision-making will lead to fairer outcomes or simply amplify existing systemic issues.
The Australian National Audit Office (ANAO) recently released its report assessing the Australian Taxation Office’s (ATO) capabilities to adopt AI. The audit (yes - that's right - an audit. Taste of the ATO's own medicine?) evaluated whether the ATO has effective governance, design, deployment, and monitoring arrangements for AI systems. While AI may present significant opportunities for tax compliance and administration, the audit highlighted areas where the ATO strategy and systems currently lack the necessary foundations for smooth AI integration.
Key findings of the ANAO's audit of the ATO
The ANAO’s report concluded that the ATO has only partly effective arrangements in place to support AI adoption. While the agency has made progress in integrating AI into its operations, it lacks comprehensive governance structures and risk management frameworks specific to AI systems.
Governance gaps identified in the audit include the absence of clearly defined enterprise-wide roles and responsibilities, insufficient central oversight of AI use across the organisation, and the need for a dedicated AI risk management framework. The ATO has established governance bodies to oversee AI adoption, including a Data and Analytics Governance Committee formed in September 2024 and the appointment of a Chief Data Officer as the accountable official for AI use in November 2024. However, these governance structures are still in their infancy and it is unclear whether they will be sufficient to manage the complexities of AI adoption.
The report additionally found that the ATO has not sufficiently integrated ethical and legal considerations into AI design, raising concerns about the potential for bias, privacy breaches, lack of transparency, and accountability issues.
Monitoring and evaluation were also found to be lacking. The ATO had no structured framework for regularly assessing AI models in production, with 74 per cent of AI models missing required data ethics assessments. While the agency has since developed a monthly report on AI strategy implementation, it has not yet established a structured evaluation approach. Without robust monitoring and governance systems in place, there is a risk that AI-driven tax administration could create more problems than it solves.
In response to the above, the ANAO recommended that the ATO:
develop and implement AI-specific governance and risk management frameworks;
clearly define roles and responsibilities related to AI oversight;
establish formal policies and procedures for AI design, development, and deployment;
integrate ethical and legal considerations into AI systems more comprehensively; and
implement monitoring and reporting mechanisms to assess AI effectiveness over time.
AI and tax: global perspectives
AI is increasingly being explored as a core tool in tax administration worldwide. A recent International Monetary Fund (IMF) opinion piece highlighted how AI has the potential to improve tax compliance, fraud detection, and revenue collection. Governments globally are investing in AI-driven solutions with the aim of enhancing efficiency and improve enforcement mechanisms. However, the success of these initiatives varies significantly, and long-term effects remain uncertain.
AI systems can analyse large volumes of data to detect patterns indicative of tax evasion, fraudulent claims, or underreported income. Machine learning models can assist in identifying high-risk taxpayers, allowing tax authorities to allocate resources more efficiently. Where AI is deployed to analyse blockchain transactions, it could offer a powerful (even draconian) tool to analyse transaction information, automate collection, and audit tax compliance.
The IMF article contends that AI can assist tax authorities by enabling real-time risk assessments, providing early warnings for potentially non-compliant behaviour. AI's predictive capabilities could also assist tax authorities to forecast revenue more accurately. That said, predictive analysis carries risk, including the potential for over-reliance on models that may not be the most accurate or unbiased.
A recent article by Ernst & Young gave an example of how generative AI (Gen-AI) could assist accounting and audit functions:
AI IN PRACTICE: It can take multinational companies weeks, multiple team members and laborious manual processes to map general ledger (GL) accounts for tax purposes. The steps generally include extracting the data from multiple countries and populating spreadsheets, manually mapping the GL accounts into a database, identifying inconsistencies among countries and issues to report back to countries, correcting errors, repopulating the database manually and then finally having a team leader review the results. GenAI-assisted automation can streamline this process into just a few steps... ...[i]n this way, a labor-intensive task can be transformed into an automated, two-day process supplemented by human review.
In advance of tax controversy matters, generative AI can help tax professionals to identify, react to and report on potential areas of controversy risk...
Another application of AI in tax administration is automated decision-making. Many tax agencies currently use AI-powered chatbots and virtual assistants to provide taxpayer support, reducing the burden on human staff. However, this again raises risks, particularly if taxpayers lack the ability to challenge AI-driven determinations (can you imagine?). In Australia, taxation systems rely on the Commissioner of Taxation to exercise certain discretions to alleviate unjust or unique situations where the tax provisions were never intended to apply. It is unclear how and if AI can be adopted in such circumstances.
In essence, the use cases of AI in professional tax services (be it tax compliance, controversy, or advisory) are numerous. However, AI systems must be transparent and explainable, as bias in AI models can lead to disproportionate scrutiny of certain taxpayers or industries (for example, unfairly negative connotations and sentiment surrounding certain investment classes) which may lead to overly conservative or overly risky positions. Additionally, the integration of AI into tax systems must comply with data protection regulations to prevent breaches of taxpayer privacy.
The future of AI in tax
Despite these challenges, AI is expected to play an increasingly important role in tax administration and professional taxation services. As AI technology continues to develop, its potential to streamline processes and enhance enforcement remains a key area of focus. However, whether AI will ultimately lead to fairer or more effective tax administration depends on the safeguards and governance measures implemented. Businesses and governments (including Australia’s) will need to ensure that AI is deployed responsibly, with careful attention to fairness, transparency, and accountability, rather than simply embracing AI for efficiency’s sake.
Australia has seen the adoption of automated processes to streamline certain compliance and recovery measures in the past, including the infamous Robo-debt affair that took several years and a Royal Commission to reveal the human damage caused.
At a recent presentation at the University of Melbourne, the new Commissioner of Taxation, Rob Heferen, expressed his desire to adopt AI within the ATO’s processes and programs. The Commissioner confirmed that the adoption will ensure there is a level of standardisation amongst decisions. In a system that hinges upon taxpayer’s intentions, here's hoping that AI is able to recognise the difference between those that are acting in good faith and those that are not.
The ANAO’s audit highlights the need for a structured approach to AI governance in tax administration. The ATO’s ongoing efforts to address these gaps will be critical in determining whether it can responsibly harness AI's potential while mitigating its risks. AI-driven tax compliance is becoming more prevalent, but its long-term success will depend on regulatory oversight and a commitment to addressing ethical, legal, and operational risks.
Written by Steven Pettigrove, Luke Higgins, and Christian Febbraro
Comments