Understanding the implications of South Africa's R3.6bn pension loss and R57bn withdrawals

South Africa’s pension system has absorbed two shocks that speak to the pressures shaping household financial life today. The first is the reported R3.6 billion impairment reported by the Government Employees Pension Fund (GEPF) for 2024/25 after delays in the recovery of unlisted investments. The second is the R57 billion that members withdrew in the first ten months of the two-pot system, a release that was meant to be gradual but quickly became a measure of how stretched families really are. These numbers reveal more than financial strain. They expose a system built on expectations that no longer match the realities of how people work, earn and save.

For many South Africans, the path into stable work is not linear. People enter the formal labour market late, often around their late twenties or early thirties. They move between short-term contracts, gig work and periodsof interruption that weaken contribution histories. They support extended families out of the same income that is meant to fund retirement, navigate uneven earnings and face rising costs that consume a growing share of disposable pay. By the time stability arrives, the runway to retirement is shorter and the pressure to make upsavings is high.

The R57 billion withdrawn under the two-pot system is therefore not surprising. It reflects a deep sense of immediacy in households where pensions serve as both long-term security and short-term relief. Older members carry a different risk. Many rely on digital systems that they were never prepared for and are exposed to increasingly sophisticated fraud. Pension communication is often written in technical language that obscures rather than clarifies. The result is a system where young and old members face different challenges yet share the same anxiety: uncertainty about what their retirement will look like and how much control they truly have.This is the landscape into which artificial intelligence is emerging.

Not as a futuristic ambition but as a practical tool that, in other parts of the world, is already reshaping how people understand their money. In the leading pension systems, AI enhances transparency by giving members real-time visibility into where their contributions are invested, how fees accumulate and what long-term projections look like under different scenarios. It simplifies complexity by translating technical portfolio information into plain language that people can interpret without financial training. It strengthens fraud detection by monitoring unusual transactions and securing accounts for older members who are at higher risk. It personalises investment strategies by adjusting risk and contribution levels in line with age, income volatility, household responsibilities and behavioural patterns. In markets like Norway, Canada and the United Kingdom, AI is not replacing trustees. It is giving them a widerfield of vision and sharper tools to manage risk.

The question facing South Africa is how these benefits translate into a context with late labour market entry, high income volatility and low preservation. AI can help younger members see the real cost of delayed saving by modelling how each year of postponement affects their retirement horizon. It can build personalised catch- up paths for people who start saving late or earn irregular income. It can simulate the long-term impact of withdrawals at moments of financial stress. It can guide members towards diversified portfolios that include global exposure, which is especially important in an economy with fluctuating growth and concentrated domestic investment opportunities. For older members, it can strengthen digital safety by detecting fraud early and simplifying access through clearly designed dashboards.

Yet the extent to which AI improves outcomes depends less on the technology and more on the governance structures around it. Defined-benefit systems like the GEPF adopt innovations cautiously because the priorityis stability. Defined-contribution systems adopt more quickly because competition shapes behaviour and members expect responsiveness. Private retirement funds with access to global markets and flexible allocations adopt new tools faster because they must justify performance and cost efficiency.

The GEPF, with its narrower offshore exposure and broader public obligations, modernises at a different pace. These different starting points matter. They determine how soon AI can be integrated, how it is supervised and how members experience its benefits.Trustees will need the skills to oversee AI models, and regulators will have to set standards that ensure transparency, accountability and protection from unintended outcomes. Models will need to be tested for bias, scenario responses and stability under stress. Members will need clarity about how AI influences decisions that affect their savings. Without these safeguards, AI risks automating existing weaknesses rather than correcting them.

The governance question, therefore, sits at the heart of the debate. It is not whether AI will enter the system. It is whether the system will be ready to be used responsibly.The intergenerational aspect is equally important. A young person entering work at 29 with fragmented income, educational debt and family obligations starts from a structurally different position than someone who enteredwork at 21. AI can help narrow this gap if it guides early engagement and consistent saving. It can also widen the gap if the tools are only accessible to the financially literate or the digitally confident.

Older members need protection from digital exclusion and must be supported through transitions that rely on online interfaces. A pension system must work for every generation, not only those who modernise with ease. By 2035, these pressures will be more visible. The young adults entering work today will be in mid-career. The older workers of today will be retired. The consequences of delayed saving, fragmented contributions, inconsistent preservation and uneven access to technology will be fully evident. This is not a prediction but asimple reflection of demographic timing. The decisions made now will shape outcomes that cannot be reversed later. The challenge is to build a pension system that strengthens understanding, increases transparency and supports both early and late savers with the tools they need to navigate uncertainty.

AI is not the future of pensions. Trust is. AI is only the instrument that reveals whether the system is designed for people or for process. South Africa’s pension landscape is entering a moment of transition driven by economic pressure and behavioural change. The R3.6 billion impairment shows the cost of delayed decision-making in a complex investment environment. The R57 billion withdrawal shows how vulnerable households are and how urgently people need clarity about their financial future. Young people entering work need tools that help them build confidence and manage volatility. Older members need protection from digital risk and reassurance that their savings are secure.

The future of retirement in South Africa will not be determined by technology alone. It will be determined by how clearly the system communicates with its members, how responsibly it adopts innovations and how fairly it distributes risk across generations. AI offers the possibility of a pension system that people understand, trust and can navigate with confidence. It offers personalised guidance that reflects how people actually live and earn. It offers transparency that replaces uncertainty with clarity. But its value depends on the choices we makeabout oversight and accountability.

We stand at a turning point. AI will enter our pension system whether we prepare for it or not. The real questionis whether we use it to build clarity and confidence or whether we allow it to automate the uncertainty that people already experience. The outcome will determine what kind of future today’s young people will inherit and what dignity older members will retire into. South Africa’s pension system can become a foundation forstability and trust or a reflection of pressures that were never addressed. The difference will be the balance we strike between technology, governance and the lived realities of the people the system is meant to serve.

Nomvula Zeldah Mabuza is a Risk Governance and Compliance Specialist with extensive experience in strategic risk and industrial operations. She holds a Diploma in Business Management (Accounting) from Brunel University, UK, and is an MBA candidate at Henley Business School, South Africa.

*** The views expressed here do not necessarily represent those of Independent Media or IOL.

BUSINESS REPORT