AI in Finance in Aotearoa New Zealand: A Living Whitepaper
Snapshot date: April 2, 2026
Introduction
AI adoption in New Zealand finance is still best described as targeted, supervised, and unevenly scaled. Banks, insurers, payment providers, wealth firms, and advisers are using AI most visibly in fraud detection, customer communications, workflow automation, analytics, and decision support. The sector is moving forward, but regulators are still framing most adoption as early-stage and control-dependent rather than fully autonomous transformation. (fma.govt.nz)
What has changed since the March 9, 2026 edition is not a sudden jump to fully automated finance. Instead, the biggest shift is that the public evidence base is getting sharper: the FMA has now published a substantial March 25, 2026 review of access to financial advice, open banking regulation is moving from framework to implementation, and the regulator’s sandbox is being widened to better support innovative financial firms. Together, these developments show a sector moving from experimentation toward selective operational scale. (fma.govt.nz)
Executive Summary
- The core pattern has not changed: AI in New Zealand finance is growing, but mostly in narrow, governed use cases rather than high-autonomy decisions. The FMA’s 2024 finance-sector research found all 13 participants either had generative AI in use or planned to adopt it soon, while the RBNZ said in November 2025 that sector use remained exploratory and limited in scope, though accelerating. (fma.govt.nz)
- Financial advice is now the clearest new scaling frontier. The FMA’s March 25, 2026 advice review found 40 Financial Advice Providers had a digital advice facility, with about 165,000 retail clients receiving digital advice in the past year that resulted in acquiring a financial product—up 90% on the prior period. (fma.govt.nz)
- The strongest current production use case remains fraud and scam prevention. ANZ’s phishing-disruption partnership, NZBA scam protections, Kiwibank’s fraud controls, and RBNZ’s broader cyber-resilience framing all point to defensive AI outpacing offensive or autonomous AI. (anz.com.au)
- Open banking is becoming a more concrete AI enabler. MBIE’s open banking regulations have been in force since December 1, 2025; Kiwibank is designated for payments from June 1, 2026 and account information from December 1, 2026; and the Commerce Commission expects the bulk of suboptimal access to transition to bank APIs by the end of 2026. (mbie.govt.nz)
- Governance is hardening into operating doctrine. RBNZ states that material AI-enabled decisions must be reviewable and subject to effective controls, and the FMA’s new advice review says the predominant use of AI is to enable advisers rather than replace them. (rbnz.govt.nz)
- Innovation policy is becoming more supportive. On March 12, 2026, the FMA said it wants to expand its fintech sandbox to include an on-ramp or restricted licence, signalling a more explicit regulatory pathway for innovative firms. (fma.govt.nz)
What’s New Since March 9, 2026
1) The FMA has published the strongest new evidence on AI in financial advice
The most important new public development is the FMA’s Access to financial advice in New Zealand review, published on March 25, 2026. It shows digital advice is no longer marginal: 40 Financial Advice Providers reported having a digital advice facility, and approximately 165,000 retail clients received digital advice in the previous year, a 90% increase from the prior period. The review also says the predominant use of AI is to extend adviser capacity rather than replace advisers, and that the FMA is planning a thematic review of AI use in financial advice. (fma.govt.nz)
This materially strengthens the earlier whitepaper’s conclusion that advice would become a major near-term expansion zone. As of April 2, 2026, advice is now one of the best-documented places where AI is moving from internal experimentation into real client-facing workflows. (fma.govt.nz)
2) The FMA is broadening its innovation posture
On March 12, 2026, the FMA said it intends to expand its sandbox pilot by introducing an on-ramp or restricted licence for innovative firms. It also said six firms entered the sandbox pilot and four identified a pathway to market for products or services that might otherwise have been delayed by regulatory uncertainty. (fma.govt.nz)
For AI in finance, this matters because it suggests the regulator is shifting from observation alone toward creating more practical pathways for supervised innovation—especially relevant for fintechs building AI-enabled payments, advice, crypto, tokenisation, or workflow products. (fma.govt.nz)
3) Open banking has become more operationally specific
MBIE’s regulated open banking regime has been live since December 1, 2025, with ANZ, ASB, BNZ, and Westpac required to be live from that date. Kiwibank is required to support payment services from June 1, 2026 and account information from December 1, 2026. MBIE also notes that only accredited data requestors can access data under the regime. (mbie.govt.nz)
The Commerce Commission’s March 5, 2026 open banking update letter adds an important competitive detail: it expects large banks to begin turning off suboptimal access for designated use cases from June 2026, expects the bulk of suboptimal access to transition to bank APIs by end-2026, and says Kiwibank’s API delivery should bring coverage to about 90% of consumer bank accounts in New Zealand. (comcom.govt.nz)
Current State of AI Adoption in New Zealand Finance
Sector-wide position
The best sector baseline remains the FMA’s September 2024 research. It covered 13 firms across banking, insurance, asset management, and advice, and found universal intent to adopt generative AI, with current use concentrated in fraud detection, risk management, decision support, product development, and product management. The FMA also recorded strong future demand for customer service use cases. (fma.govt.nz)
RBNZ’s May 2025 AI special topic and November 2025 Financial Stability Report still frame the sector as being in an early but accelerating phase. RBNZ says AI can improve model accuracy, risk assessment, cyber resilience, productivity, and customer offerings, but warns of systemic risks from concentration, vendor dependency, cyber misuse, and weak controls. In November 2025 it said current AI use remained exploratory and limited in scope, with adoption accelerating. (rbnz.govt.nz)
Where adoption is now strongest
The clearest areas of real-world traction are:
- Fraud and scam prevention, where ROI is immediate and measurable. (anz.com.au)
- Customer communications and workflow support, especially in financial advice and service operations. (fma.govt.nz)
- Operational productivity, including CRM, note-taking, summarisation, and internal analysis. ANZ, for example, announced on February 5, 2026 that it had deployed Salesforce’s Agentforce within a new CRM for business banking, describing it as an at-scale rollout intended to automate routine tasks and improve banker productivity. (anz.com.au)
- Payments and API-enabled experiences, where open banking is improving the data and action layer required for AI-enabled services. (mbie.govt.nz)
Workforce implications
A useful broader signal comes from RBNZ’s February 3, 2026 analytical note, which found AI exposure is highest in professional, managerial, and administrative occupations. That is not finance-specific, but it strongly suggests many finance roles in New Zealand sit in the parts of the labour market most exposed to AI-assisted task redesign. This is an inference from the occupational mix, not a direct sector estimate. (rbnz.govt.nz)
Recent News and Market Developments
Advice and wealth: AI is becoming infrastructure, not just experiment
The FMA’s March 2026 advice review is the single most important new indicator of adoption maturity. It says the sector is cautious, but it also records a growing set of live tools: AI agents that provide advice, compliance-checking tools, record-keeping tools, dashboards that summarise client situations, and personal finance tools that proactively engage customers. At the same time, the FMA says the predominant model is still AI with adviser oversight, not adviser replacement. (fma.govt.nz)
Consumer trust remains mixed. In the same review, net trust in AI-provided advice ranged from 28% to 41% across product categories, and only 3% of respondents said AI would be their preferred way to interact with a financial adviser. Simpler products attracted more trust than complex advice areas. (fma.govt.nz)
Banking and payments: agentic systems are emerging, but carefully
Westpac NZ announced on February 17, 2026 that Mastercard had completed New Zealand’s first fully authenticated agentic transactions on its network with Westpac. The framing is important: the announcement emphasised secure, transparent, authenticated participation of AI agents in the payment flow. (westpac.co.nz)
This is still early-stage, but it marks a shift in the market narrative. New Zealand finance is beginning to test AI not only as an analytical assistant, but as a system that can participate in authenticated commercial action under a controlled framework. (westpac.co.nz)
Fraud and scams: ecosystem coordination is still the clearest success story
ANZ said on January 15, 2026 that its pilot with 2degrees had expanded to include Kiwibank, TSB, One NZ, Spark, and Trade Me, with the collaboration disrupting more than 5,000 phishing domains in two months and reducing ANZ customer card-phishing cases by 39% over the same period. ANZ explicitly linked the effort to increasingly complex and AI-generated attacks. (anz.com.au)
At the industry level, NZBA said banks’ scam protections include pre-transaction warnings, identification of high-risk transactions, 24/7 reporting channels, and cross-bank sharing of scammer account information, while Kiwibank’s February 2026 result highlighted confirmation of payee, real-time fraud blocking, high-risk transaction monitoring, and in-the-moment scam education. (nzba.org.nz)
Research Overview
Official regulator and policy research
The most relevant current public research stack is now:
- FMA, September 10, 2024: sector baseline on AI in financial services. (fma.govt.nz)
- RBNZ, May 5, 2025: systemic benefits and risks of AI in financial stability. (rbnz.govt.nz)
- RBNZ, November 2025: AI still exploratory, but accelerating; reviewability, controls, cyber, and third-party oversight are central. (rbnz.govt.nz)
- FMA, March 25, 2026: digital advice, AI-enabled advice tools, consumer trust, and a planned thematic review of AI in advice. (fma.govt.nz)
- MBIE, from December 1, 2025 onward: open banking regulations and standards that create a regulated data-sharing and payment-initiation layer for future AI-enabled services. (mbie.govt.nz)
What this body of evidence now says
Taken together, the research says New Zealand finance is not standing still, but it is scaling AI in a distinctly conservative way:
- Start with lower-regret use cases such as fraud, servicing, summarisation, and workflow support. (fma.govt.nz)
- Keep humans accountable for material decisions. (rbnz.govt.nz)
- Build around regulated data access and secure payments rails rather than around unconstrained model autonomy. (mbie.govt.nz)
- Treat vendor concentration, cyber risk, and transformation risk as first-order issues, not implementation footnotes. (rbnz.govt.nz)
Case Studies
Case Study 1: Financial advice is shifting from manual delivery to hybrid AI-enabled service
The FMA’s March 2026 review shows a live market for digital and AI-supported advice: 40 FAPs with digital advice facilities, roughly 165,000 retail clients receiving digital advice in the last year, and use cases spanning AI agents, compliance checking, record creation, summarisation, and financial analysis dashboards. The regulator’s reading is that AI is mainly augmenting advisers and increasing capacity. (fma.govt.nz)
Why it matters
- It is the clearest evidence that AI in New Zealand finance is moving into customer-facing workflows at scale. (fma.govt.nz)
- It suggests the next big growth area is not fully automated advice, but hybrid advice models that combine digital journeys, AI support, and human accountability. (fma.govt.nz)
Case Study 2: Anti-phishing collaboration shows AI works best when intelligence is shared
The ANZ-led collaboration with telcos, platforms, and other banks is a strong example of AI deployment beyond a single institution. The partnership disrupted more than 5,000 phishing domains in two months and ANZ reported a 39% drop in card-phishing cases over the same period. (anz.com.au)
Why it matters
- It shows the most mature AI use cases in finance are often ecosystem use cases, not only internal automation. (anz.com.au)
- It reinforces that in New Zealand finance, fraud prevention is still the area with the strongest operational proof. (anz.com.au)
Case Study 3: Westpac and Mastercard point to the next phase—AI that can act
Westpac’s February 2026 announcement with Mastercard is early, but strategically important because it places AI agents inside an authenticated payment framework rather than leaving them outside the transaction perimeter. (westpac.co.nz)
Why it matters
- It is a preview of how action-taking AI may enter finance in New Zealand: first in controlled, consented, standards-based payment contexts. (westpac.co.nz)
- It also aligns with the regulatory direction that material decisions and actions need clear controls, traceability, and reviewability. (rbnz.govt.nz)
Key Trends
Trend 1: Defensive AI still leads
Fraud detection, scam prevention, cyber monitoring, and anomaly detection continue to scale faster than high-autonomy lending, pricing, or advice. This remains the clearest pattern across bank announcements, regulator commentary, and industry safeguards. (anz.com.au)
Trend 2: Advice is becoming a serious scaling channel
The March 2026 FMA review makes this clearer than before. Digital advice usage is rising quickly, and the regulator is now openly discussing AI-supported advice models and preparing a thematic review. (fma.govt.nz)
Trend 3: Open banking is moving from promise to operating layer
The combination of MBIE regulation, accreditation, standards, and the Commerce Commission’s transition expectations means open banking is now less of a future concept and more of a live market structure. That matters because AI value in finance often depends on access to permissioned data and secure payment initiation. (mbie.govt.nz)
Trend 4: Governance is becoming product architecture
RBNZ’s position is now explicit: data quality, human oversight, and reviewable decisions are core safeguards. In practice, that means governance is no longer separate from product design; it is part of how finance AI is built and approved. (rbnz.govt.nz)
Trend 5: Third-party concentration and transformation risk are rising, not fading
RBNZ’s November 2025 report says 26% of surveyed entities were only partially meeting baseline expectations for third-party documentation, and 22% were only partially meeting termination expectations. It also warns that concentration in cloud, identity, and network services can create correlated disruption across firms. (rbnz.govt.nz)
Constraints and Risks
Trust and consumer risk
AI in finance remains constrained by trust. The FMA’s consumer research shows meaningful but limited trust in AI advice, especially for more complex products, and a strong continued preference for human channels. This implies customer-facing AI can grow, but likely through hybrid models rather than pure automation. (fma.govt.nz)
Conduct and decision risk
RBNZ says material AI-enabled decisions must be reviewable and subject to effective controls, while the FMA says AI in advice should be deployed with oversight and design that ensures conduct obligations are met. That sets a high bar for any move into underwriting, suitability, or other consequential decisions. (rbnz.govt.nz)
Vendor and infrastructure risk
RBNZ continues to warn that third-party dependency, shared service providers, cloud concentration, and transformation programmes can amplify operational and systemic risk. For a relatively small market like New Zealand, where many technology layers are imported, this is likely to remain one of the defining structural constraints on finance AI. (rbnz.govt.nz)
Overall Assessment
As of April 2, 2026, AI in New Zealand finance is best understood as operationally real, strategically important, and still tightly bounded by governance, trust, and infrastructure constraints. The sector has clearly advanced beyond simple experimentation: financial advice has measurable digital scale, banks are using coordinated intelligence to fight scams, open banking is under active regulation and phased rollout, and early agentic payment activity has appeared in-market. (fma.govt.nz)
But the center of gravity has not shifted to autonomous finance. The dominant New Zealand model is still AI as copilot, detector, summariser, recommender, and workflow engine, with humans retaining responsibility for consequential decisions. That is consistent across FMA, RBNZ, MBIE, and current bank deployments. (fma.govt.nz)
In short: New Zealand finance is moving deeper into the implementation era of AI, and the March 2026 evidence shows real acceleration in advice, payments, and fraud defence. The near-term winners are still likely to be firms that combine AI with secure data access, strong third-party controls, reviewable decisions, and customer-trust design—rather than firms chasing autonomy for its own sake. (fma.govt.nz)