Governing with AI: The new face of tax authorities

The OECD has released its new report Governing with Artificial Intelligence (PDF, 5.3 MB), a landmark study on how governments can adopt artificial intelligence responsibly. While much of the public sector still lags behind private industry, the report highlights tax administration as one of the few areas already showing concrete results.

Around the world, AI is helping tax authorities both strengthen compliance, by detecting fraud and errors, and improve services by making filing simpler and more personalised. Yet, as the OECD cautions, the adoption of AI in such a sensitive field must be balanced with trust, transparency, and fairness.

Detecting tax fraud with algorithms

Tax administrations sit on enormous datasets: returns, invoices, payroll reports, and customs declarations among others. This large volume overwhelms traditional audit processes. AI is transforming this landscape. Machine-learning models can scan millions of records, identify unusual patterns, and assign risk scores that direct auditors’ attention where it matters most.

The results are striking. Australia’s Taxation Office used real-time analytics to flag anomalies across more than 600,000 taxpayer interactions, generated close to AUD 80 million in revenue in just one year. Austria’s predictive analytics centre uncovered irregularities worth €185 million across income, corporate, VAT, and customs taxes. 

Emerging economies are also testing AI: Madagascar’s customs authority plans to deploy generative AI on a decade of trade data to strengthen fraud detection. With the U.S. Government Accountability Office estimating that fraud and improper payments cost the federal government between $233 and $521 billion annually, the stakes could not be higher.

Personalised and proactive taxpayer services

If fraud detection is the “stick,” AI-enabled services are the “carrot.” Filing taxes is often viewed as opaque and intimidating. AI can change that by acting as a personalised guide.

Singapore’s multilingual “Ask Jamie” chatbot has already halved call-centre inquiries by providing instant, accurate responses. In South Korea, a virtual assistant leads taxpayers step-by-step through filing and payment, reducing errors and boosting on-time submissions.

France uses natural-language AI to triage taxpayer emails and draft suggested responses, so staff only need to validate the final version. And in the UK, the “GOV.UK Chat” generative AI pilot is giving thousands of citizens quick answers on tax and business rules.

Beyond chatbots, personalisation increasingly means pre-filled returns and tailored reminders. A taxpayer who fails to claim allowable deductions might receive a prompt to check eligibility, or income already reported to government systems can appear automatically in draft filings.

This proactive approach reduces mistakes, saves time, and signals that the system is working with citizens, not against them. The result: smoother experiences and higher voluntary compliance.

Streamlining operations and boosting efficiency

AI is also reshaping the back office. Automating repetitive tasks such as data entry, form checking, and invoice validation frees staff to focus on complex cases and policy.

The efficiencies are substantial. A 2022 study by the Alan Turing Institute found that more than 80% of routine public-sector transactions in the UK could be automated, freeing 1,200 staff-years annually. In Ireland, real-time payroll reporting combined with analytics has enabled refunds to be issued within days instead of weeks. In Australia, AI-driven checks correct discrepancies on the spot, improving both accuracy and speed.

These improvements do more than cut costs. Faster refunds mean happier citizens. Fewer errors mean fewer disputes. And more efficient internal workflows allow governments to “do more with less”, a critical advantage when public budgets are tight.

The trust factor: Transparency, fairness and oversight

Yet efficiency alone is not enough. Taxation is one of the most visible expressions of the social contract, and public confidence is fragile. If AI systems are perceived as unfair, biased, or opaque, they risk undermining trust not only in the tax system but in government as a whole.

The risks are real. An AI audit tool that disproportionately targets corporations within a certain industry, or a chatbot that delivers incorrect advice, could cause financial harm and erode legitimacy. That is why transparency and oversight must be central. Taxpayers should understand why an AI-driven decision was made and have the ability to contest it.

France requires human validation of AI-drafted communications. South Korea routes sensitive chatbot queries to human staff. In the UK, users are warned explicitly that generative AI answers may be imperfect and are directed to official links for verification.

Privacy is equally vital. Tax data is among the most sensitive information governments hold. Many authorities are therefore training models only on anonymised data, auditing systems for bias or leakage, and publishing information on how AI operates. The OECD’s AI Principles, fairness, transparency, accountability, and safety, set the standard. Without such guardrails, a single scandal could undo years of progress.

A new era, if done right

The trajectory is unmistakable: AI is moving from pilots to mainstream deployment in tax administration. One can imagine “digital-first” agencies where routine filings, registrations, and basic audits are automated, while humans focus on complex cases, policy, and oversight. Tax inspectors could become AI-augmented analysts; taxpayers might feel as though they have a knowledgeable adviser rather than a faceless bureaucracy.

But this future is not guaranteed. Many tax authorities still rely on outdated IT systems, skills need upgrading, and governance frameworks are evolving. Leaders must find the right balance: cautious enough to safeguard trust, but proactive enough to meet citizens’ rising expectations for digital service. Early wins in fraud detection and service delivery can build momentum, but they must be paired with clear communication and strong oversight.

AI offers tax administrations an extraordinary opportunity: to collect revenue more effectively, serve citizens more efficiently, and strengthen fairness. Done responsibly, it can enhance the social contract. The true measure of success will not be revenue recovered or refunds accelerated, but whether citizens feel the system is transparent, fair, and worthy of trust. Those tax authorities that innovate boldly while embedding accountability will set the benchmark for governing with AI. Those that do not risk losing the very confidence they depend on.

Expert guidance in tax

Discover how our tax services can support your business

Explore our tax services