AI in Finance in Aotearoa New Zealand: A Living Whitepaper

Introduction

AI adoption across New Zealand’s financial sector is becoming more operational, measurable, and customer-outcome focused. The sector is still not characterised by widespread autonomous lending, underwriting, investment, or advice decisions. Instead, the strongest adoption continues to occur in fraud prevention, customer-service assistance, vulnerability detection, workflow automation, open banking, and employee productivity.

The most significant development since the July 13 edition is the emergence of further evidence from insurance and payments. Vero has deployed AI sentiment analysis across customer interactions, while the open-banking ecosystem has continued to mature. At the same time, industry discussion has shifted from whether finance should use AI to how it can be used transparently, safely, and with appropriate human accountability.

Executive Summary

• AI adoption is moving from pilots into bounded production use. Vero’s customer sentiment deployment is a new example of AI operating across live insurance interactions, with human staff required to validate and act on flagged outcomes. (vero.co.nz)
• Customer-service augmentation remains the leading visible banking use case. Westpac’s AI-enabled contact-centre rollout was scheduled for full deployment across its contact centres by August 2026, but no public post-deployment performance results had been released by the snapshot date. (westpac.co.nz)
• Open banking is becoming usable infrastructure rather than a policy objective. API Centre reporting shows the major banks ready with the Payment Initiation API, while Kiwibank is also ready with version 3.0 of both payment initiation and account-information standards. (apicentre.paymentsnz.co.nz)
• Agentic payments are moving into controlled ecosystem preparation. ANZ New Zealand, ASB, BNZ, and Kiwibank joined Visa’s Agentic Ready programme, following Mastercard and Westpac’s authenticated agentic payment demonstration. (visa.co.nz)
• Trust and transparency are now central industry concerns. The Financial Services Council’s August 2026 conference placed AI trust, open finance, responsible deployment, and customer guardrails at the centre of its programme. (fsc.org.nz)
• Regulatory attention is broadening beyond operational efficiency. MBIE’s July capital-markets consultation explicitly asks how New Zealand should respond to digital and AI-related innovation in financial products and markets. (mbie.govt.nz)
• Public evidence of autonomous consequential decision-making remains limited. The available New Zealand evidence still points to AI being used mainly as a copilot, detector, summariser, workflow engine, and service-quality tool. This is an assessment based on the publicly documented deployments reviewed for this edition. (vero.co.nz)

What Has Changed Since July 13, 2026

1. Vero has launched AI-powered customer sentiment monitoring

On July 16, 2026, Vero announced the rollout of Salesforce AI Sentiment Analysis across its sales and service operations. The system analyses customer phone and email interactions and identifies positive, neutral, or negative sentiment, as well as possible complaints and indicators of customer vulnerability. Where negative sentiment is detected, staff are prompted to review the interaction and take action where appropriate. (vero.co.nz)

Vero reported that:

• The system was piloted across approximately 45,000 customer interactions between November 2025 and March 2026.
• More than 65,000 sentiment outcomes had been generated after launch.
• Approximately 10% of customer email and voice interactions were being analysed in real time.
• Managers could gain visibility across approximately 58 interactions per consultant per month, nearly 30 times more than under the previous manual quality-assurance approach. (vero.co.nz)

This is important because the deployment is not simply an efficiency tool. It is being used to identify dissatisfaction, complaints, and vulnerability earlier in the customer journey. The case demonstrates how AI is being applied to support conduct, customer-care, and fair-treatment objectives while retaining human review.

2. Open banking infrastructure has reached a more mature stage

The API Centre’s current implementation reporting shows:

• ANZ, ASB, BNZ, and Westpac marked READY for Payment Initiation version 2.3.
• Those four banks also marked READY for Account Information version 2.3.
• Kiwibank marked READY for both Payment Initiation and Account Information version 3.0 on May 28, 2026.
• Kiwibank’s implementation includes retail and eligible business accounts and is backwards compatible with version 2.3. (apicentre.paymentsnz.co.nz)

This substantially strengthens the data and payments foundation for AI-enabled financial services. It allows fintechs and financial institutions to build services around consented data sharing, payment initiation, budgeting, cash-flow analysis, lending applications, and embedded finance.

The relationship between open banking and AI is strategic rather than automatic: open banking does not itself constitute AI adoption, but it supplies the structured, consented data and transaction capabilities that intelligent financial services require. This is an inference from the current infrastructure and product direction. (apicentre.paymentsnz.co.nz)

3. AI has moved higher on the capital-markets policy agenda

MBIE’s Capital Markets Reform Phase Two consultation, published in July 2026, asks for views on how New Zealand should respond to digital and AI-related innovation in financial products and markets. The consultation is not an AI-specific regulatory framework, but it signals that AI is now being considered as part of the future competitiveness and design of the capital-markets regime. Submissions close on August 25, 2026. (mbie.govt.nz)

The policy questions include:

• How financial products and markets are changing through digital innovation.
• Whether existing rules remain appropriate for new forms of financial activity.
• How New Zealand can remain competitive while protecting investors and maintaining market integrity.
• How regulation can support innovation without creating unacceptable consumer or systemic risk. (mbie.govt.nz)

4. Industry dialogue has shifted toward trust and responsible AI

The Financial Services Council’s August 12–13, 2026 conference included dedicated sessions on AI and financial services, open finance, human–AI collaboration, interpretable AI, AI governance, and digital work.

The Council stated that new research presented at the conference examined how New Zealanders feel about AI in financial services, with particular emphasis on:

• Transparency.
• Clear guardrails.
• Meaningful use cases.
• Consumer confidence.
• Human-centred deployment. (fsc.org.nz)

The underlying research was not publicly available in full by the snapshot date, so it should be treated as an emerging evidence source rather than a quantified national benchmark.

Current State of AI Adoption

Adoption profile by use case

Use case	Current position in New Zealand	Assessment
Fraud and scam prevention	Widely prioritised by banks and regulators	Advanced and defensive
Contact-centre assistance	Moving into live deployment at banks and insurers	Scaling
Complaint and vulnerability detection	Emerging, with Vero providing a clear example	Early production
Adviser and employee productivity	Established in selected firms and workflows	Growing
Open banking and data-driven services	Core technical infrastructure increasingly available	Infrastructure maturing
Agentic payments	Demonstrations and ecosystem programmes underway	Experimental
Automated lending and underwriting	Limited public evidence of autonomous consequential decisions	Early and highly supervised
Autonomous investment or trading	No strong public evidence of scaled local deployment	Limited disclosure

This pattern is consistent with the broader New Zealand evidence base. The FMA’s 2024 research found that all 13 participating financial-services organisations either already used generative AI or expected to adopt it soon, with motivations including operational efficiency, customer outcomes, and fraud detection. (fma.govt.nz)

KPMG’s New Zealand financial-services research similarly reports early wins in fraud detection and data management, while identifying strategy, culture, skills, transparency, and ethical frameworks as the conditions required for wider value creation. (kpmg.com)

Current News and Developments

Vero: AI for sentiment, complaints, and vulnerability

Deployment: Salesforce AI Sentiment Analysis across sales and service teams.

Functionality:

• Reviews customer emails and calls.
• Classifies sentiment.
• Flags possible complaints.
• Identifies indicators of customer vulnerability.
• Creates follow-up tasks for human validation.

Strategic significance:

• Extends quality monitoring beyond small manual samples.
• Supports earlier intervention in stressful insurance interactions.
• Creates a stronger feedback loop between customer experience, conduct risk, and operational coaching.
• Demonstrates a regulated use case in which AI supports, rather than replaces, human judgement. (vero.co.nz)

Westpac: AI-supported customer conversations

Westpac began rolling out Microsoft Dynamics 365 Contact Centre as a Service in April 2026. The AI component provides customer-service staff with relevant customer and product information during live conversations. Westpac said it expected all contact centres to be deployed by August 2026. (westpac.co.nz)

The bank’s April research found:

• 65% of respondents were comfortable or neutral about AI supporting contact-centre employees.
• 70% were comfortable or neutral about AI being used to detect fraud and scams.
• 66% said they already used AI in some aspect of their personal, work, or other activities. (westpac.co.nz)

The deployment illustrates a recurring New Zealand pattern: public acceptance is higher when AI is framed as a tool that helps employees deliver faster, more informed service, rather than as a replacement for human contact.

Westpac has also created a dedicated Chief Data, Digital and AI Officer role. Russell Jones was appointed to the position in June 2026, with responsibility for Westpac New Zealand’s data, digital, and AI strategy. (westpac.co.nz)

Agentic payments: from demonstration to ecosystem preparation

New Zealand’s most visible agentic-finance activity remains concentrated in payments.

In February 2026, Mastercard and Westpac completed authenticated agentic transactions in New Zealand using a Westpac-issued debit card to purchase cinema tickets. Mastercard’s model treats the AI agent as a visible and governed participant in the payment flow. (westpac.co.nz)

In April, Visa announced that ANZ New Zealand, ASB, BNZ, and Kiwibank had joined its Agentic Ready programme. The programme is designed around tokens, identity, risk controls, and consumer oversight as AI agents begin to interact with merchants and payment networks. (visa.co.nz)

These developments are not evidence that agentic banking is already mainstream. They indicate that payment networks and banks are preparing the authentication, consent, liability, and fraud-control mechanisms required before agent-initiated transactions can scale.

ASB: AI capability-building beyond the bank

ASB’s Pathway to Productivity programme aims to support more than 4,100 New Zealand businesses in its first year. One component is an AI bootcamp co-developed with Xero and delivered by academyEX. Another places emerging AI and data-science talent with ASB business customers. (asb.co.nz)

This represents a broader role for banks as AI ecosystem enablers. Rather than using AI only inside its own operations, ASB is positioning itself to help business customers develop AI capability, improve productivity, and access relevant expertise.

Research and Policy Overview

Financial Markets Authority

The FMA’s latest broad public research remains its September 2024 study of 13 representatives from banking, insurance, asset management, and financial advice. All participants either used generative AI or expected to adopt it soon. The principal motivations were:

• Better customer outcomes.
• Operational efficiency.
• Fraud detection.
• More personalised services.
• Improved internal productivity. (fma.govt.nz)

The previous edition also identified the FMA’s more recent work on AI in financial advice as an important indicator that supervisory attention is moving from general awareness to use-case-specific oversight.

Reserve Bank of New Zealand

RBNZ’s May 2026 Financial Stability Report describes AI as both a source of potential efficiency and a possible amplifier of financial-system risk.

The principal risks identified include:

• Dependence on a small number of third-party AI providers.
• Biased, misleading, or fraudulent model outputs.
• Greater cyber risk.
• Operational-resilience weaknesses.
• Potential credit effects if AI contributes to job losses in particular sectors.
• Amplification of social-media-driven loss of confidence in banks. (rbnz.govt.nz)

The RBNZ’s 2026 smaller-bank stress-test scenario includes a bank “name crisis” in which social media and AI contribute to deposit withdrawals. This demonstrates that AI is being considered not only as a productivity tool, but also as part of the threat environment surrounding liquidity and confidence. (rbnz.govt.nz)

The Reserve Bank’s July 2026 Monetary Policy Review also noted that AI-related investment was supporting global growth while warning that a correction in AI-related asset prices could affect financial conditions. (rbnz.govt.nz)

KPMG New Zealand

KPMG’s latest New Zealand financial-services report describes AI as moving beyond narrow automation and becoming a strategic driver of innovation and customer experience.

Its key messages are:

• Early wins are visible in fraud detection and data management.
• The largest opportunity lies in aligning AI with long-term strategy and organisational culture.
• Transparency and skills investment are essential.
• Ethical frameworks are increasingly linked to customer trust and value creation.
• Embedded finance and fintech partnerships will be important complements to AI. (kpmg.com)

Financial Services Council

The FSC’s 2026 State of the Sector report provides broader context for the industry’s ability to invest in technology. It reports that financial services contributed NZ$16.1 billion to New Zealand GDP in the year to June 2025, while labour and capital inputs increased by 22% between 2020 and 2024, including investment in technology, digital platforms, cybersecurity, and regulatory capability. (blog.fsc.org.nz)

The FSC’s August conference indicates that the next phase of AI discussion is likely to focus less on generic adoption and more on:

• Consumer trust.
• Explainability.
• The relationship between AI and professional advice.
• Open finance.
• Human–AI collaboration.
• Governance of more autonomous systems. (fsc.org.nz)

Case Studies

Case Study 1: Vero AI Sentiment Analysis

Objective: Detect dissatisfaction, complaints, and vulnerability earlier.

Scale: Pilot of approximately 45,000 interactions; more than 65,000 sentiment outcomes after launch.

Human oversight: Consultants validate flagged outcomes and decide what action is appropriate.

Assessment: One of the strongest new examples of AI being applied to customer outcomes and conduct monitoring in New Zealand insurance. It provides a practical model for using AI to expand oversight without delegating final judgement to a model. (vero.co.nz)

Case Study 2: Westpac AI-enabled contact centre

Objective: Give customer-service employees relevant customer and product information during live conversations.

Technology: Microsoft Dynamics 365 Contact Centre as a Service with built-in AI.

Deployment position: Full contact-centre deployment was expected by August 2026; public performance results were not available by the snapshot date.

Assessment: A significant banking use case because it places AI directly inside regulated customer interactions while preserving a human interface. (westpac.co.nz)

Case Study 3: Kiwibank open-banking readiness

Objective: Provide standardised payment-initiation and account-information APIs across relevant personal and business banking channels.

Status: Kiwibank is marked READY for version 3.0 of both standards, in addition to being backwards compatible with version 2.3.

Assessment: Kiwibank’s readiness broadens coverage of the open-banking ecosystem and reduces the competitive asymmetry that previously existed between the largest banks and smaller providers. It also creates a stronger platform for future AI-enabled budgeting, lending, payments, and financial-management tools. (apicentre.paymentsnz.co.nz)

Case Study 4: Mastercard and Westpac agentic payments

Objective: Demonstrate authenticated payments initiated by an AI agent.

Result: A New Zealand Agent Pay transaction used a Westpac-issued debit card to purchase cinema tickets.

Assessment: This remains a controlled demonstration rather than mainstream adoption. Its importance lies in testing the governance layer around agent identity, authentication, transaction visibility, fraud controls, and dispute handling. (westpac.co.nz)

Case Study 5: Tower’s AI-enabled contact centre

The previous edition documented Tower’s AI-enabled contact-centre deployment, including reported reductions in customer interaction time, average handling time, and total time spent by customers across the service operation.

Tower remains an important benchmark because it demonstrates that AI adoption in New Zealand insurance is extending beyond pilots into measurable operational change. The case also reinforces the prevailing model of AI-assisted service rather than autonomous claims adjudication.

Key Trends

Trend 1: Assistive AI remains the dominant model

The strongest public evidence continues to involve AI assisting employees with:

• Summarisation.
• Customer-service preparation.
• Product and policy retrieval.
• Sentiment analysis.
• Fraud detection.
• Complaint identification.
• Workflow prioritisation.

There is still limited public evidence of AI independently making final decisions on credit, insurance claims, investment allocation, or financial advice at scale.

Trend 2: AI is increasingly connected to customer-protection objectives

The Vero deployment is particularly significant because it links AI to:

• Vulnerable-customer identification.
• Complaint prevention.
• Earlier intervention.
• Conduct monitoring.
• Service-quality improvement.

This suggests that firms may find it easier to justify AI adoption when it produces demonstrable improvements in customer protection rather than only internal cost reduction. (vero.co.nz)

Trend 3: Open banking is becoming the substrate for intelligent finance

The combination of standardised account information, payment initiation, consent mechanisms, and real-time payment capability creates the conditions for:

• Automated cash-flow management.
• Personalised financial guidance.
• Embedded lending.
• Merchant payment optimisation.
• Real-time affordability assessment.
• Agent-assisted payments.

The commercial value of these services will depend on customer adoption, data quality, consent design, cyber resilience, and the ability to explain AI-generated recommendations.

Trend 4: Agentic finance is being developed through payment controls first

New Zealand’s agentic-finance activity is currently concentrated in payments rather than lending or investment decisions. This is logical: payment networks can define transaction limits, authentication rules, tokenisation, identity controls, and dispute processes more clearly than they can govern open-ended autonomous financial advice.

Trend 5: Governance is becoming part of operating architecture

Governance signals now include:

• Dedicated executive ownership of AI at Westpac.
• Greater regulator attention to third-party AI dependence and cyber risk.
• MBIE consideration of AI-related capital-markets innovation.
• Industry focus on trust, transparency, and guardrails.
• Increased emphasis on human review in customer-facing deployments.

Trend 6: Public disclosure remains uneven

Large banks and insurers are increasingly announcing AI initiatives, but public information about:

• Model performance.
• Error rates.
• Bias testing.
• Customer outcomes.
• Complaints involving AI.
• Operating costs.
• Workforce impacts.

remains limited. This makes it difficult to compare deployments or assess sector-wide productivity gains independently.

Outlook: Late 2026 to 2027

The next phase of New Zealand’s finance-AI development is likely to be shaped by five developments:

1. Open-banking scale-up
   The next question is no longer whether APIs will be available, but whether consumers and businesses will use them frequently enough to support viable AI-enabled products.

2. More customer-outcome monitoring
   Vero’s use of sentiment and vulnerability signals may encourage other insurers and banks to deploy AI for complaints, conduct, accessibility, and financial hardship support.

3. Controlled agentic-payment pilots
   Visa, Mastercard, banks, fintechs, and merchants are likely to continue testing agent identity, authentication, delegated consent, transaction limits, and liability.

4. More explicit supervisory expectations
   RBNZ and FMA attention is likely to expand from general AI awareness toward model governance, outsourcing, operational resilience, cyber risk, and use in consequential decisions. RBNZ has scheduled its next Financial Stability Report for November 11, 2026, while results from 2026 bank stress tests are expected later in the year. (rbnz.govt.nz)

5. Greater pressure to demonstrate value
   As AI budgets grow, financial institutions will need to show measurable improvements in customer experience, fraud losses, processing time, adviser capacity, claims handling, or risk detection—not simply announce experimentation.

Overall Assessment

As of August 18, 2026, AI in New Zealand finance is operationally established in selected domains, expanding through insurance and payments, and increasingly governed as a strategic and prudential issue.

The sector’s current model can be summarised as:

AI as copilot, detector, summariser, workflow engine, and customer-protection tool—rather than autonomous decision-maker.

The clearest new evidence is Vero’s production deployment of AI sentiment monitoring, which extends AI into customer vulnerability and complaint prevention. Open banking has also reached a more mature infrastructure stage, improving the prospects for data-driven and embedded financial services. Meanwhile, agentic payments are being developed through controlled demonstrations and trust-layer programmes rather than unrestricted autonomy.

The strategic challenge for New Zealand financial institutions is now less about proving that AI can produce efficiency gains. It is about demonstrating that AI can deliver those gains while preserving fair treatment, privacy, explainability, human accountability, operational resilience, and public trust.