In many respects, South Africa has therefore already taken its first steps towards a world in which tax increasingly “just happens”. The more interesting question is: how much further could this go?
Azam Khan, an academic programme leader for the Higher Certificate in Accounting at Mancosa, asks: How far can automated tax go? (Image source: © 123rf
123rf)
For many South Africans, the idea of tax happening automatically is no longer futuristic.
Qualifying individual taxpayers are already familiar with SARS’s Auto Assessment process. Instead of completing an income tax return from scratch, SARS uses information received from employers, banks, medical schemes, retirement funds, insurers and other third parties to calculate an individual’s tax position. If the information is correct, the taxpayer may have little more to do than review the assessment.
The South African Revenue Service (SARS) is moving towards an increasingly data-driven tax administration system. Its Modernisation 3.0 programme seeks to use data, automation and artificial intelligence (AI) to build a smarter tax administration system. What taxpayers currently experience through Auto Assessments may therefore offer only a glimpse of a much broader transformation.
And this is not a conversation for some distant future. Public consultation on SARS’s VAT modernisation proposal remains open until 16 October 2026, making now an important moment to consider not simply whether technology can make tax administration more efficient, but what we want a modern tax system to look like.
From income tax to VAT
SARS’s proposed Digital VAT Model brings together e-invoicing, secure digital information exchange and e-reporting, allowing VAT transaction data to move closer to real time. Over time, this could support pre-filled VAT returns, improved risk detection and eventually VAT auto-assessment.
In simple terms, the distance between transaction and taxation is shrinking.
There is a strong case for this. Better data could reduce administrative burdens, identify errors earlier, improve compliance and make fraud more difficult. SARS itself anticipates lower administrative effort and costs for vendors, improved data quality and stronger capacity to narrow the VAT compliance gap.
But modernisation also creates an opportunity to think beyond simply digitising the system we already have.
This direction resonates with an idea I previously put forward that South Africa consider the feasibility of a real-time VAT payment mechanism in which the VAT component of a transaction is transferred directly to SARS when a purchase takes place, rather than being held by the vendor and subsequently paid through the traditional VAT cycle.
The thinking was relatively simple. If technology allows VAT to be identified at the point of transaction, could it also allow that revenue to reach the fiscus sooner, potentially reducing elements of the collection and administrative burden?
SARS’s current proposal does not go that far. Its focus is on the near-real-time flow of transaction data, rather than the real-time transfer of the VAT payment itself. Yet this is precisely the kind of question that modernisation should encourage us to explore.
Such a system would require careful consideration of business cash flow, input VAT credits and refunds, banking and payment infrastructure, implementation costs and safeguards for errors.
But modernisation should not simply make an old system digital; it should create space to ask whether technology allows us to design a more efficient one.
What happens when the computer gets it wrong?
Efficiency, however, cannot be the only measure of success.
Technology can process enormous amounts of information quickly, but incorrect data can still produce an incorrect outcome only faster.
Auto Assessments already demonstrate why taxpayer oversight remains important: automation does not remove responsibility; it changes its nature.
As systems become more sophisticated, there is a danger that taxpayers assume a computer-generated outcome must be correct. Convenience can easily become complacency.
The future taxpayer may therefore need a different kind of financial literacy. Knowing how to complete a tax return could become less important than knowing how to interrogate one that has effectively been prepared for you.
A digital system still needs people
There is also a South African reality that cannot be ignored. A large organisation with integrated accounting systems and specialist tax expertise will experience modernisation differently from a small business with basic systems and limited technical support.
Modernisation must therefore mean more than digitisation. It must include education, accessibility, realistic transition periods and practical support.
SARS’s proposed implementation recognises this through a phased approach extending from consultation and testing towards implementation from 2030, subject to readiness and approvals.
The same transformation will reshape the accountant’s role. If systems increasingly capture information and identify anomalies automatically, professional value shifts towards interpreting, questioning and validating what those systems produce.
The question may increasingly move from “Can you calculate the tax?” to “Can you explain why this calculation is correct?”
That has implications for accounting and taxation education. Future graduates will still need strong technical knowledge, but also digital literacy, ethical awareness, critical thinking and professional scepticism.
Ultimately, South Africa’s move towards intelligent tax administration is about more than AI. It is about trust.
Tax may increasingly “just happen”. The real test will be whether South Africans understand how it happened, trust that it happened correctly and know what to do when it did not.
That may ultimately be the real test of the AI taxman.
