iscover 
According to credit reporting bureau Experian’s latest Connected Intelligence Report, based on responses from 102 senior decision-makers in Australia’s lending market, almost three-quarters of respondents (72 per cent) are using agentic AI to assist underwriters with recommendations or decision support.
However, just 3 per cent of respondents said their data was fully AI-ready, while 67 per cent said their data was either not ready or only partially ready to support AI-driven decisioning.
While the adoption of the technology is snowballing, the report found that many lenders were still in the early phases of operationalisation.
More than seven in 10 Australian respondents (72 per cent) described their organisation as either emerging or early in its use of AI across fraud and credit risk underwriting, while just 11 per cent said AI was widely implemented across underwriting processes.
Experian also found the biggest barriers to scaling AI were largely operational, with fragmented data systems unable to provide a unified customer view identified by 45 per cent of respondents, followed by poor data quality (42 per cent) and a lack of trust in AI outputs (31 per cent).
Mathew Demetriou, managing director of software solutions at Experian Australia and New Zealand, said that while many players in the industry were now pushing AI into underwriting workflows, the challenge was now ensuring the technology could be trusted and governed at scale.
“What we’re seeing with Australian lenders is that AI is already in underwriting workflows with the research showing 72 per cent are using agentic AI for decision support, but the harder question being asked is how to trust and govern it at scale, especially as regulators sharpen their focus on how data is used,” Demetriou said.
Majors develop capability
While Experian’s report highlighted the inconsistencies for AI adoption in the lending market, some banks have surged ahead.
Speaking at the Amazon Web Services event in Sydney and reported in The Australian Financial Review, Westpac’s chief digital, data and AI officer Andrew McMullan said that thousands of AI agents had helped the major bank save 12,500 hours in their first month of operations.
McMullan said that the agents were processing upwards of 1.5 million transactions and more than 32,000 payslips every week.
“For customers, a payslip is just one document in an application process. For our people, verifying it can be really complex, time-consuming, and full of policy judgements,” McMullan said.
“Now our specialist agents work together around the banker. They can classify the payslip, they read it, they can extract the data, run the calculations in the back, and verify it against policy before giving our banker the output to review and continue the application with speed, confidence, and trust.”
Commonwealth Bank of Australia (CBA), which outpaces its major rivals on the Evident Insights AI adoption ratings of global banks, said that it is investing $2.4 billion annually in technology and capability, $500 million more than its closest rival.
“We are not simply buying generic intelligence. We are strengthening advantages that are hard to replicate: customer relationships at scale, deep knowledge of the Australian economy, secure use of data and the everyday banking relationships we earn with customers,” Matt Comyn, CEO of CBA, said.
‘The foundations underneath AI still need to catch up’
While strides have been made among some lenders, Experian’s report found that there remained shortfalls in implementation.
More than two-thirds of respondents (69 per cent) agreed data quality and governance were among the reasons AI implementations fail, while 84 per cent said transparency of analytics and insights was highly valuable to improving decisions.
Experian said that implementation had become increasingly important as AI moves further into core credit and fraud risk decisions, with changes to the Privacy Act, Scams Prevention Framework, and Consumer Data Right reshaping how financial institutions collect, share, and use customer data.
That caution was also reflected in how lenders view AI decision making. More than half of Australian respondents (51 per cent) said they were comfortable allowing AI to make decisions without human review only for low-risk decisions, while just 2 per cent were comfortable with fully autonomous decisioning at scale across most use cases.
Despite the reservations, the appetite for AI-enabled decisioning remains strong, with 92 per cent of Australian respondents saying they would pilot, test, or adopt a vendor that could meet their data, software, and AI needs for fraud and credit risk underwriting.
Demetriou said the next stage of adoption would depend on lenders closing the gap between AI capability and the quality of the data supporting it.
“The foundations underneath AI still need to catch up,” he said.
“Data quality, integration, trust and governance may determine whether AI can move from contained use cases into core decisioning.”
[Related: Is AI coming for brokers? Industry weighs impact]
Want to see more stories from trusted news sources?Make Broker Daily a preferred news source on Google.