For millions of South Africans, tax season still feels like an annual ritual of gathering documents, checking deductions, and submitting returns. Inside the South African Revenue Service (SARS), the country’s revenue authority, the process looks very different.

Long before many taxpayers log into eFiling, artificial intelligence (AI), machine learning, and advanced data analytics have already assessed risk, matched third-party information, and helped determine which returns deserve closer scrutiny. What was once a labour-intensive process driven largely by manual audits is becoming a technology-powered operation driven by data.

The shift is one of Africa’s most notable examples of AI being deployed at scale in government. Beyond improving tax collection, it offers a view into how algorithmic decision-making is reshaping public institutions and the relationship between citizens and the state.

SARS has quietly become one of the continent’s most sophisticated users of AI, embedding data science across nearly every stage of tax administration, from auto-assessments and fraud detection to compliance verification and audit selection. The approach reflects a broader trend among African governments using AI not only to improve efficiency, but also to close revenue gaps in digital economies.

“SARS uses data science, machine learning and AI as part of its broader modernisation programme to continuously innovate and improve tax compliance processes,” Siphithi Sibeko, Head of Communication and Media at SARS, told TechCabal in an interview on Monday. “The SARS strategy focuses on the customer experience and applying these capabilities to ensure that ‘tax just happens’.”

The scale of the technology’s impact is already visible. According to Sibeko, the SARS compliance programme contributed R304 billion ($18.2 billion) during the 2024/25 financial year. AI-assisted fraud detection and verification prevented more than R417 billion ($25 billion) in impermissible refund outflows over the past five years.

Sibeko also said that 100% of verification cases and 88.41% of complex audit cases are now selected using automated risk-assessment functionality, showing how algorithms have become central to identifying compliance risks.

Rather than relying only on information submitted through tax returns, SARS has built a broader digital picture of taxpayers by integrating data received under its statutory mandate from employers, financial institutions, medical schemes, retirement funds, insurers, investment managers, and other reporting entities. It also receives information from domestic government registers, foreign tax authorities, and cryptocurrency reporting frameworks.

In a July 1 statement, SARS said the upgrades are intended to make tax compliance “simpler, faster and more secure” for millions of taxpayers. It added that as of 1 July 2026, more than 1.9 million taxpayers had been auto-assessed, with about R8 billion ($479 million) in refunds paid out within 72 hours.

The expanding data ecosystem has become especially important as South Africa’s e-commerce market grows. Online retail sales are projected to reach R130 billion ($7.8 billion), representing nearly 10% of total retail sales, reflecting the increasing volume of digital transactions that generate taxable income and data trails.

Income from freelancing platforms, remote work, e-commerce businesses, and crypto assets often leaves digital traces that traditional tax systems struggle to follow. AI now enables SARS to analyse these complex datasets at a scale that would be impossible through manual investigation.

“The focus is not on any single technology, but on a platform approach that integrates data, analytics, AI and modern compliance capabilities to make compliance easier for honest taxpayers and harder to evade for those who choose not to comply,” Sibeko told TechCabal. “We are seeing increased participation in the digital economy, which reinforces the importance of ensuring that all taxable income is declared, regardless of how or where income is earned.”

SARS says technology is meant to support, not replace, human judgement despite its growing reliance on AI. Risk indicators generated by its systems are reviewed through governance processes and human oversight, and the models are continuously refined to improve accuracy and minimise false positives before any enforcement action is taken. That matters in a tax system where automated screening can shape who is reviewed first, even if final decisions still pass through officials.

Sibeko said that human-in-the-loop decision-making will remain important as governments around the world confront questions about AI accountability, transparency, and citizens’ rights when automated systems influence public decisions.

For South Africa, the implications extend beyond tax collection. SARS’ Modernisation 3.0 strategy aims to create a smart, digital, and data-driven revenue authority built around digital identities, unified taxpayer records, and AI-powered compliance systems.

As governments across Africa look for ways to improve revenue collection without raising tax rates, SARS is showing how AI is becoming embedded in the administrative machinery of the state, especially where digital commerce leaves more records than paper trails.