The insurance sector does not have a data problem. Rather, it has a decision problem. Underwriting, pricing, claims, fraud, customer engagement and capital all shape profitability, retention and regulatory outcomes, often using different systems, measures and definitions of the same customer or policy.
SAS and Risk.net research into insurance decision-making has found that many insurers still struggle to connect risk, revenue and cost in a way that supports fast, confident decisions. Four in 10 insurance leaders say they are not confident their organisation has a comprehensive, real-time view of risk, revenue and costs, and nearly 40% are unsure their structure, processes and technology were aligned to support strategy.
If the organisation cannot see the whole picture, it will keep optimising components in isolation but should not be surprised if the outcomes do not deliver the value expected.
The modern insurance pressure test
The same SAS research lists the pressures insurers name most often. These include economic uncertainty (55%), regulatory change (41%), technological innovation (35%) and cost pressures (34%). Among technology and data leaders specifically, 60% point to budget constraints as their biggest challenge.
In South Africa, those pressures show up as rising claims costs, persistent fraud attempts, more demanding customers, tighter scrutiny on conduct and fairness, and executive teams expecting faster answers with less tolerance for uncertainty. Even though the operational strain is real, so is the temptation to respond with fixes like adopting a new fraud approach, a different pricing model or a new customer platform running alongside the old one.
Slow decisions cost money. When it takes too long to move from quotation to policy issue or from claim submission to claims closure, insurers lose customers and costs increase.
Silos are not just about technology
The silo problem is often described as a legacy issue. In practice, it is usually a mix of structure, incentives and sequencing.
A SAS and Risk.net insurance research report describes gaps in decision-making across functions and the difficulty of consistently sharing data and analytics across the business. Significantly, only 28% of respondents said their organisation is “very confident” in its ability to use data and analytics effectively.
In a South African context, that gap often shows up in familiar moments: a customer receives a retention offer while a claims dispute is still unresolved; a broker gets one view of product eligibility while the call centre uses another; a legitimate claim triggers an overcautious fraud rule because the model cannot see context from service history; or a premium increase is technically justified, but poorly explained, and the customer churns.
When teams use different data definitions, errors and rework increase, leading to additional compliance checks and slowing core operations. Of course, technology alone will not fix this. AI can help, but if we do not redesign workflows and clarify who owns what, we just create silos faster and may continue working in silos.
What ‘agile insurance’ actually means for CIOs
Agility in insurance is often framed as speed. In reality, it is the ability to change direction without breaking governance, customer trust or operational stability.
That requires a shift from “analytics as projects” to “analytics as infrastructure”. The goal is not more dashboards. It is decision consistency, where marketing, underwriting, pricing, fraud and claims management can act on the same underlying truth, apply the same governance standards, and understand trade-offs between growth, risk and service.
This is where cloud and modern data architecture matter; they enable scaling analytics, reusing models and applying controls consistently, rather than rebuilding decision logic function by function.
AI raises the stakes on governance
Locally, SAS argues that insurers are entering a period in which AI is no longer optional experimentation. It is becoming embedded in operational decision flows, raising the bar for transparency, model management and responsible use.
This is the part CIOs cannot delegate. As AI becomes more present in pricing, triage, fraud detection and service automation, the organisation needs to be able to answer basic questions quickly and defensibly: what data was used? Which model made the recommendation? What constraints were applied? How is drift monitored? Who owns the outcome?
If the business cannot trace decisions, it cannot scale them. Worse, it cannot correct them quickly when conditions change.
From disjointed teams to an intelligent enterprise
Even though technology can help, fragmented technology investment is a poor substitute for a cohesive strategy. For South African insurers, the practical path forward is not transformation but disciplined integration.
This can take the form of unifying data definitions that drive customer and policy truth and standardising how models are deployed and monitored across functions. Furthermore, insurance should build governance that operates at operational speed and prioritise decision journeys where cross-functional conflict is most costly, such as onboarding, claims handling, fraud escalation and retention.
Done properly, breaking silos becomes profitable. It reduces duplicate work, limits conflicting customer interactions, improves model performance by providing better context, and gives executives a clearer line of sight into risk and return.
The research signals that insurers already understand the pressure. The organisations that will gain the most momentum will be the ones that treat data and analytics as the connective tissue of the operating model, not a set of disconnected initiatives.