en.Wedoany.com Reported - The South African Revenue Service (Sars) has used artificial intelligence to prevent over R100 million in non-compliant capital outflows and plans to further expand AI applications to handle the majority of taxpayer interactions.
At a Sars event in Pretoria, Commissioner Johnstone Makhubu outlined a dual-track AI strategy: one already yielding results in compliance and fraud prevention, and another aimed at transforming the taxpayer experience with the agency. The 2026 tax season begins with auto-assessments on July 1, with individual filing from July 13 to October 23. During the event, Makhubu showcased technology equipment prepared for the season, including 4×4 mobile service vehicles covering rural areas, upgraded self-service terminals, and digital channels such as WhatsApp, mobile apps, USSD, and the "Ask Lwazi" chatbot. He stated that Sars is transitioning from a regulator to a service provider, from complexity to simplicity.
Over the past decade, Sars has been embedding machine learning into compliance operations, continuously enhancing capabilities in risk detection, case prioritization, and operational efficiency. This is part of the "Modernization 3.0" reforms announced earlier this year, which integrate AI-driven compliance with taxpayer biometric identification and a planned real-time payment system. Makhubu noted that Sars, using algorithms and large language models to read multiple data sources, has prevented over R100 million in irregular capital outflows. The system simultaneously cross-checks bank statements, VAT returns, and data from the Companies and Intellectual Property Commission, using matching algorithms and analytics to flag anomalies and erroneous declarations. Sars emphasized that AI does not make compliance decisions autonomously; risk signals are passed to a case selection unit, which weighs other inputs and applies established business rules, all within a data governance framework.
The customer-facing strategy is a more ambitious goal, relying on updated generative AI. Sars has deployed the AI assistant "Ask Lwazi" on its website and plans to achieve hyper-personalization. Makhubu, benchmarking against Chinese tax authorities, stated that Sars could use AI to handle 80% of incoming call inquiries. He described a near-term goal of using optical character recognition technology to process submitted documents via AI agents without human intervention. Deputy Commissioner Carl Scholtz said AI is being used to match and clean third-party data Sars receives from employers, banks, medical aid schemes, and other institutions. AI-assisted verification is in the third phase of internal testing, with some features expected to enter production within the next 12 months. Enterprise Data Management Head Marius Papenfus introduced several probabilistic models: one prioritizes cases based on the likelihood and feasibility of collection, another predicts whether individuals should be registered for tax based on economic activity, and a third estimates the potential magnitude of underreported income using corporate ownership data and government tender records.
Sars insists on retaining human final decision-making in compliance matters, which will be central to its ambition of large-scale automation while maintaining a human oversight baseline.
This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com
