Paradoxes of scaling AI in South African financial services

Paradoxes of scaling AI in South African Financial Services

May 2026

AI in financial servicesContents

Thank you for reading

Survey of South African banks and insurers reveals four paradoxes shaping how AI is actually being scaled in South African financial services.

South African banks and insurers have moved beyond AI pilots, but most are still struggling to scale AI into an enterprise capability.
AI investment is focused on efficiency and productivity, with limited impact yet on how firms compete or differentiate.
Governance is strengthening, but unclear ownership and weak data foundations continue to slow progress.

South African banking and insurance is entering a new phase of competition where speed to adapt, cost‑to‑serve, and personalised service matter as much as balance sheet strength and brand trust. The next decade will not be decided by who runs the best pilots, but by who can redesign how decisions are made, how work is done and how customers are served. AI is the accelerant. The real story is reinvention.

To understand how the market is responding, we surveyed 12 South African financial institutions, using a framework based on EY’s earlier banking and insurance surveys in the United States. The survey gives us two useful perspectives: a view of how leading South African institutions are approaching AI today, and an international reference point for how those patterns compare with a more scaled market. Taken together, the results tell a more interesting story than simple optimism or simple lag. South African institutions have clearly moved beyond curiosity. But most have not yet fully crossed into scale.

The difference shows up quickly in practice. Launching pilots, standing up governance forums, and pointing to early use cases is one thing. Turning those efforts into an enterprise AI capability that meaningfully improves performance, sharpens competitiveness, and stands up to the realities of regulation is something else entirely. That is the harder task now facing boards and executive teams.

Some tensions are visible in the South African data alone. Others become clear when set against the US reference point. Either way, the conclusion is consistent: AI progress is real, but it is uneven, cautious in places, and not yet translating consistently into strategic advantage.

Four paradoxes stand out

1. AI is live, but not built to scale

South African financial institutions are past the “what is AI?” phase. Money is committed, teams are mobilised, and deployments are starting to show up in the wild. Every surveyed bank had already rolled out or soft‑launched GenAI. Similarly, most surveyed insurers had moved beyond experimentation into pilot, beta or launch‑ready releases. That is meaningful momentum, but it is not yet scale.

Across banking and insurance, the same issues keep resurfacing: weak data foundations, limited access to usable data, difficult integration into legacy environments, and the challenge of embedding AI into day-to-day workflows. In banking, data quality and legacy integration emerged as the top two blockers. Insurance pointed in much the same direction. These are not minor technical frustrations. They are the structural reasons why promising use cases fail to become a repeatable capability.

That leaves many firms in an awkward middle ground. AI is visible enough to sound reassuring. Boards hear updates. Executives can point to pilots and early wins. But the underlying foundations are still too fragile to support consistent enterprise-wide scale. The US comparison matters here only as a reminder of where this goes next.

In a more scaled market, the firms that pull ahead are not the ones with the most pilots. They are the ones that fix data, integration and architecture early enough for value to compound. If South African institutions do not treat those foundations as a strategic priority, AI will remain active but uneven.

2. There is plenty of innovation, but almost no visible failure

There is no shortage of activity in the South African data. Institutions are experimenting, piloting, and expanding their use-case portfolios. On the surface, that looks encouraging.

What is harder to see is whether organisations are learning with the same discipline they use when launching. That is where the picture starts to look less mature. There is far less evidence of hard-edged value measures, explicit stop rules, or a strong institutional habit of shutting down weak ideas and reusing what has been learned. This is where the US comparison becomes useful. US institutions appear further along not only because they are larger and better funded, but because many seem more comfortable absorbing controlled failure as part of capability building. In South Africa, by contrast, the number of reported visible failures is strikingly low relative to the amount of activity underway.

In banking, respondents reported virtually no failures. Insurance showed the same broad pattern: plenty of activity, very little failure being surfaced. Some of this may be due to timing, and some to definitional issues. However, there are likely to be elements of avoidance, where financial institutions are more comfortable launching pilots than retiring them, or where their ambitions are set too low.

That is a problem. A healthy AI portfolio should produce some dead ends. What matters is whether the institution learns quickly enough to stop weak ideas, redirect investment and build on what works. Without that discipline, innovation generates noise rather than momentum. In practice, this often shows up as pilots that remain isolated within functions, with little reuse across the organisation.

3. Most firms are using AI to run better, not compete differently

Across both banking and insurance, the current AI agenda still leans heavily toward productivity, efficiency and operational use cases. That is sensible. Claims, underwriting, servicing, fraud, risk operations, and internal productivity are all rational starting points. They are measurable, easier to justify, and usually lower risk.

The survey data makes that bias clear. In banking, use cases clustered heavily in the back office, often around 40% to 50% of the portfolio, while front-office use cases tended to sit closer to 10% to 20%. In insurance, the centre of gravity was claims, underwriting, and servicing, with distribution notably less prominent. In other words, firms are mostly using AI to improve how the machine runs, not yet to rethink how the business competes. But that is still only a starting point.

The bigger strategic opportunity lies elsewhere: in changing how decisions are made, how customers are served, how products are designed, and how value is created. That is where AI stops being a tool for optimisation and starts becoming a force that can reshape competitive advantage.

This is one of the clearest places where the US benchmark sharpens the picture. In a more scaled market, the conversation has moved further toward broader commercial integration and more advanced deployment. In South Africa, many firms still appear to be in the safer phase: using AI to improve the current model rather than challenge it.

That may be the right sequence in the short term. It becomes a risk if it lasts too long. Efficiency creates breathing room. It does not, on its own, prepare an institution for a market in which the basis of competition may be shifting.

4. Governance is stronger but ownership is still too unclear

The survey data suggests that governance is getting more attention, and that is the right instinct. In a regulated sector, AI without guardrails is not innovation; it is unmanaged risk. But governance only helps if it accelerates clear decisions. When it becomes a collection of committees and reviewers, it can slow delivery, dilute accountability and turn AI into a compliance exercise rather than a performance lever.

This is especially visible in banking. Half of the respondents said the primary AI decision-maker was a dedicated governance committee, while only a quarter pointed to executive leadership. That structure may feel safe, but it often means that no single executive is accountable for value, trade-offs, and delivery pace. The practical outcome is predictable: priorities drift, teams build in parallel, approvals take longer, and scale stalls quietly.

The “so what” is simple: AI at scale needs both control and cadence. Keep the guardrails strong but make decision rights explicit, empower accountable executives, and measure governance by outcomes (speed, reuse, risk incidents avoided), not by the number of forums.

What stands out in banking: momentum outpacing readiness

Banking appears further along in practical GenAI rollout and is more comfortable discussing advanced patterns such as agentic AI. All banks in our survey had already rolled out or soft-launched GenAI. That sounds encouraging, and in many respects it is. But it also creates a specific risk: mistaking activity for maturity.

The US contrast sharpens the point. American banks appear further along in their scaled deployment and are more open about experimentation that does not work.

South African banks do not need to copy that posture exactly. But they do need to ask whether they are learning fast enough and simplifying enough to avoid expensive fragmentation.

What stands out in insurance: advancing, not accelerating

Insurance tells a different story. The tone is more cautious, spending levels appear lower, and the use cases are more clearly anchored in operational improvement. Claims, underwriting, and servicing are natural starting points, and in many ways sensible ones. What is striking, though, more than half of the insurers in our survey referenced the EU AI Act when describing the governance frameworks shaping their thinking, even though regulatory readiness still appeared to be one of the weaker areas of maturity. That is revealing. It suggests a market that is referencing global standards faster than it is building the capability to meet them.

The risk is that this caution becomes a trap. If insurers stay too focused on efficiency and too hesitant about strategic reinvention, they may miss where AI could reshape the economics of distribution, service, and underwriting. In other words, a sensible incremental agenda can start to look slow surprisingly quickly. That does not mean insurers should chase novelty for its own sake. It does mean they should ask whether the current level of ambition matches the scale of the potential shift.

_{The real risk is not AI, but how slowly the model is changing. All of this would matter less if the competitive landscape were stable. It is not.}

Marius Van Den Berg, EY Africa Financial Services Leader

In our recent client conversations, a new concern has begun to surface more explicitly: the threat of AI-native challengers. Not just firms with better tools, but firms built around fundamentally different cost models, service models and product economics. Businesses designed from the ground up with different assumptions about how much work people should do, how fast offerings can evolve, and what customers should expect.

That should concern incumbents more than the next pilot. Established financial institutions still have major strengths: trust, capital, customer bases, brands and regulatory credibility. But incumbency becomes a liability when it hardens into caution at exactly the moment the market starts to reward speed, adaptability, and business model re-imagination. That is why this moment matters. The issue is not whether South African financial institutions are doing enough to look active on AI. Many are. The issue is whether they are making the harder choices that turn activity into advantage before the basis of competition shifts.

Many institutions are clearly active on AI. What is less clear is whether they are making the harder choices that turn that activity into advantage before the basis of competition shifts.

Three questions for boards and executive teams

Are we building AI as an enterprise capability with clear business ownership, or are we still funding a collection of promising initiatives run by the technology function?
Do our governance structures sharpen decisions and accelerate learning, or are they mainly making us feel safer while slowing us down?
Are we using AI to improve today’s business model, or to prepare for the models that could challenge it tomorrow?

In summary

South African banks and insurers have moved decisively beyond AI experimentation, but most have yet to translate activity into enterprise‑wide scale.

A survey of 12 institutions shows strong momentum in pilots, governance forums and early GenAI deployments, particularly in banking. However, weak data foundations, legacy integration challenges and unclear executive ownership continue to limit impact.

AI investment is largely focused on productivity and operational efficiency rather than reshaping competitive advantage. Compared with US peers, visible failure and learning discipline remain limited. The central risk is not insufficient AI activity, but the slow pace of underlying business and operating model change.

Three immediate actions

For leaders who want to move from reflection to action, three immediate priorities stand out:

Treat AI as an enterprise capability, not a technology initiative.
Establish clear business ownership with authority across business, technology, risk and operations.
Make governance a performance enabler, not just a control layer.
Clarify decision rights and judge governance by pace, reuse and business outcomes, not process alone.
Back a small number of domains where AI could reshape economics, not just improve efficiency.
Focus on areas such as servicing, underwriting, claims, fraud or distribution where AI could change how the business competes.