New Zealand lenders adopt AI amid data readiness gap

Experian research shows 76% of New Zealand financial institutions are using agentic AI to support underwriters, highlighting a gap between AI adoption and data readiness across the sector.

Only 2% of institutions surveyed said their data is fully ready for AI-driven decisioning, while 69% said it is either not ready or only partially ready. The findings suggest lenders are introducing AI into credit and fraud workflows even as many continue to struggle with data quality, integration and governance.

Almost two thirds of New Zealand respondents, or 65%, described their organisations as emerging or early in their use of AI across fraud and credit risk underwriting. This suggests AI use is spreading, but many firms have yet to embed it deeply in core decision-making processes.

Operational barriers featured strongly in the survey. Among New Zealand respondents, 49% cited lack of trust in AI outputs as a key challenge, 41% pointed to poor data quality, and 37% identified fragmented data systems.

The findings come as banks and other lenders face pressure to strengthen fraud controls while meeting stricter expectations around responsible lending and the use of customer data. In that context, the quality of internal data and the ability to explain automated recommendations are becoming more important.

More than half of respondents, or 55%, said data quality and governance are among the reasons AI deployments fail. At the same time, 88% said transparency of analytics and insights is highly valuable in improving decisions, underscoring the importance of explainability as AI takes on a larger role.

Regulatory scrutiny is also shaping how far firms are prepared to go. The survey found that 63% of New Zealand financial institutions reported increased scrutiny around data governance, consent and permissible use as AI-driven analytics expand.

Decision limits

Caution is evident in attitudes toward autonomous decision-making. Nearly half of respondents, or 47%, said they were comfortable allowing AI to make decisions without human review only in low-risk cases.

Just 2% said they were comfortable with fully autonomous decisioning at scale across most use cases. This suggests most firms still want human oversight in higher-stakes credit and fraud judgments, even where AI tools are already being used to support staff.

The survey also examined what institutions see as the main benefit of applying AI, data and software to underwriting operations. A total of 53% identified faster or real-time decision cycles as a core gain.

That focus reflects the commercial pressure on lenders to make quicker decisions in loan origination and fraud prevention. Faster decisions can improve customer experience and reduce operational delays, but they also place greater demands on the systems and data used to produce them.

Data problem

The report suggests the main obstacle is no longer whether institutions want to use AI, but whether their data environments are robust enough to support it. Fragmented systems, inconsistent data quality and governance weaknesses can undermine confidence in outputs and make audit trails harder to maintain.

New Zealand institutions also signalled interest in more consolidated technology arrangements. Nine in 10 respondents said they would pilot, test or adopt a vendor that could meet their data, software and AI needs for fraud and credit risk underwriting.

That result points to demand for fewer handoffs between separate data, analytics and decisioning tools. It also suggests some institutions are looking to simplify model oversight, governance controls and workflow integration.

The New Zealand findings form part of a wider study of senior decision-makers and expert interviews across multiple countries. In New Zealand, the sample covered 51 respondents, while 102 respondents were surveyed in Australia.

Mathew Demetriou, Managing Director, Software Solutions, Experian Australia and New Zealand, commented on the results: "What we're seeing with New Zealand lenders is the appetite for AI is clear and our research shows 76% are using agentic AI for decision support. However, they are grappling with trust. How do you rely on AI outputs and prove your decisions stand up to scrutiny? Getting the data foundations right is what makes that possible."

He said the next phase of adoption would depend on whether firms can improve those foundations. "Data quality, integration, trust and governance may determine whether AI can move from contained use cases into core decisioning. Without trusted and well-governed data, organisations may struggle to operationalise AI at scale. With appetite for integrated AI-enabled decisioning strong, the industry's next phase may be defined by how quickly firms can close the data readiness gap by bringing data, AI and governance together into a single, trusted decisioning environment. It's what we call connected intelligence," said Demetriou.