The pitch versus the balance sheet
Westpac NZ managing director Sarah Hearn told RNZ that AI has the potential to bring a bit more humanity back into banking by freeing bankers from data drudgery so they can have real conversations with customers. It is a warm message. It is also an incomplete one.
Westpac NZ’s latest disclosure statement shows operating expenses of $1,493 million, up 9.4% on the prior year. When costs are climbing that fast on a $128 billion asset base, the motivation for automation is not philosophical. It is financial.
Westpac’s own industry economist Paul Clark has been blunt about the competitive stakes. Firms that do not adopt AI will lose competitiveness, he wrote, warning that without a unique value proposition it will become “increasingly difficult for firms who remain tied to the past to match the productivity gains of those that adopt AI.” That is not a customer service argument. That is a cost-of-survival argument.
Every bank is moving, but nobody wants to say where
A Financial Markets Authority survey of 13 finance sector representatives found all had either integrated generative AI or planned to. The language across the sector follows a careful script: Kiwibank says AI will help staff “work smarter, not harder.” ANZ is taking “a considered approach.” ASB insists “it’s our people that set us apart” while running an AI chatbot called Josie since 2018.
The clearest preview of where this ends sits across the Tasman. ASB’s parent Commonwealth Bank is trialling a generative AI chatbot processing 50,000 calls daily, converting conversations to transcripts in 1.2 seconds and assessing customer sentiment in real time. CBA has 2,400 local call centre staff. What CBA does at scale, ASB’s New Zealand operation will face pressure to replicate.
The technology is no longer experimental. Enterprise deployments of agentic AI, systems that handle multi-step decisions across full customer journeys, are achieving containment rates above 80% for routine queries with customer satisfaction scores improving up to 20%. NiCE Chief AI Officer Philipp Heltewig put it plainly: “Agentic AI is not a chatbot upgrade. It is a new operating model for customer experience.”
Customers are already voting with their patience
ServiceNow research found 72% of New Zealanders say they want to use self-service first. New Zealanders spent 22 million hours on hold last year, down two million from the year before. Banks and retailers already achieve the fastest resolution times at 2.4 days on average, compared with 6.5 days for government and manufacturing.
The demand side is not the problem. The risk is on the supply side. AI Forum NZ founding executive director Ben Reid has cautioned that “the reliability and accuracy of these AI voice solutions is still not quite good enough for prime-time.” Victoria University AI lecturer Andrew Lensen pointed to the Air Canada chatbot case, where an AI invented a bereavement policy and the airline lost in court, as evidence that hallucination risk in customer-facing AI remains very real.
The junior workforce gap nobody is planning for
Here is the part of the story the banks are not talking about. Research from people2people Recruitment found 45% of employers anticipate hiring fewer junior roles within three to five years. Yet 69% of employees say their employer is not preparing them for an AI-driven future.
Catherine Kennedy, NSW Managing Director at people2people, warned: “If junior work is increasingly automated, organisations must redesign pathways, not eliminate them, or we’ll create a long-term talent bottleneck.”
The productivity evidence supports the shift. University of Bath Professor Vivek Soundararajan documented that AI-assisted customer service agents showed a 15% increase in issues resolved per hour, while management consultants completed work 25% faster with 40% higher quality. But as Soundararajan observed, the distributional problem is real: “AI helps those already inside the door while quietly narrowing the door for those trying to get in.”
NZ Treasury’s own analysis reinforces this concern. Its analytical note flagged that AI’s disproportionate impact on higher-skilled tasks means advanced economies like New Zealand may be more exposed, while the country’s traditionally slow technology diffusion could mean we absorb the disruption before capturing the productivity gains.
What this means for every business with a service desk
Banks are moving first because they have the capital and the call volume. Inland Revenue is already further ahead than most realise, with AI-driven debt collection recovering over $51 million in just 10 weeks. The regulatory framework is deliberately light, with Technology Minister Judith Collins arguing new rules would stifle innovation.
The pattern the banks are setting, automate the routine, thin the junior pipeline, frame it as service improvement, will be replicated across professional services, insurance, and mid-sized business operations within a few years. Business owners should be thinking less about whether to trust Westpac’s chatbot and more about what their own service team looks like in 2028.
Sources
- RNZ: Using AI to ‘bring a bit more humanity back into banking’ (2025)
- Westpac NZ Disclosure Statement – Annual Report 2025 (2025)
- Scoop: Westpac Industry Insight – Artificial Intelligence (2023-11)
- NZ Herald: AI could take thousands of call centre jobs at ASB’s Australian parent
- NiCE report shows agentic AI transforming contact centres
- Newstalk ZB: Kiwis willing to embrace AI as phone hold times fall
- IT Brief: AI plans threaten junior hiring as staff left unready
- Interest.co.nz: AI and work – What’s yet to come?
- NZ Treasury: The Impact of Artificial Intelligence – Analytical Note 24/06 (2024-07)
- Inland Revenue: Update on Artificial Intelligence – IR2025/365 (2025-10)
- Businessdesk: Light rules, high stakes – Agencies push AI into frontline work