AIBRIEF
NAB's technology chief says the bank is running a 12 month pre customer safety phase, with security as the lead concern, before deploying AI that can act on a customer's behalf.
National Australia Bank will spend the next year testing whether its AI agents are safe before putting them in front of customers, the bank's group executive for technology and enterprise operations told Bloomberg.
Patrick Wright said the bank is running a 12-month proof-of-guardrails phase, with agentic AI piloted inside the bank first and security named as the lead concern. "Security is the thing I think about," Wright said. "We've got to do it in a way that's safe."
NAB already uses agentic AI internally to help operations staff complete complex tasks more quickly. The customer-facing version, where AI agents would interact with bank staff and systems on behalf of customers, is the next step, and the bank wants a year of pre-customer testing before that happens.
That sequencing puts NAB behind Commonwealth Bank of Australia, which has rolled out new AI services this year. CBA chief executive Matt Comyn has said agentic systems are helping write fraud rules and strengthen security patching. On the Evident AI Index 2025, a benchmark that ranks global banks on AI maturity, NAB sits 34th, well behind CBA at 4th, per Bloomberg. The ranking is a single-snapshot measure, not live standings.
Wright did not name a customer launch date.