Blog
/
Discover how CBA, ANZ, NAB and Westpac are embracing Generative AI and agentic AI to transform banking in Australia.
For over two centuries, Australia’s banking sector has thrived by adapting to cycles of change — from the introduction of ATMs in the 1960s, to internet banking in the late 1990s, to the emergence of mobile-first banking apps in the 2010s. Each wave of innovation forced institutions to rewire their operations, retrain their workforces, and rethink how they interact with customers.
Now, we stand at the threshold of what may be the most consequential transformation of all: the arrival of generative AI (GenAI) and agentic intelligence. Unlike past technologies, which digitized or automated specific workflows, GenAI reshapes the very interface between humans and banking systems. Instead of logging into a portal, navigating menus, or filling forms, customers and bankers alike will increasingly converse with intelligent agents that not only answer questions but also take action — moving money, generating reports, reconciling accounts, or designing personalized financial plans.
This is not theoretical. Australia’s largest banks — particularly Commonwealth Bank of Australia (CBA) and ANZ Bank — are moving decisively, with live partnerships, workforce programs, and job postings that illustrate both the speed and seriousness of their intentions. For CIOs, CTOs, and business leaders, the implications are profound: a reimagining of infrastructure, governance, workforce models, and even customer trust.
It is worth dwelling on why this moment feels different from previous “AI” waves. Australian banks have experimented with machine learning for decades — credit risk modeling, fraud analytics, robo-advisors. But those tools were invisible to customers and narrow in scope.
Generative AI differs in two decisive ways:
The Interface Shift:
For the first time, AI is front-of-house. ChatGPT, Claude, and Gemini have trained millions of Australians to type or speak questions and expect articulate, context-aware responses. Banks must now decide whether to meet customers in that new mode or risk irrelevance.
The Agentic Turn:
The next step goes beyond answering. Agentic AI refers to systems that can plan, decide, and act within defined parameters. A banker might ask an agent to “summarize all SME loans with exposure to energy prices and draft risk mitigation scenarios.” The agent doesn’t just retrieve data — it executes across multiple systems, compiles insights, and produces a decision-ready output.
These two shifts — conversational interfaces and agentic execution — redefine what it means to “use” a bank. In the same way that the browser replaced desktop software, and the mobile app displaced much of the browser, the AI agent may become the new default interface for banking.
CBA, as the nation’s largest retail and business bank, has a history of being at the forefront of digital banking. It was among the first to launch internet banking, one of the earliest to adopt mobile apps, and consistently ranks highest in digital engagement among Australian majors. It is no surprise, then, that CBA is pushing hardest into generative and agentic AI.
In August 2025, CBA announced a strategic partnership with OpenAI, cementing itself as OpenAI’s key banking partner in Australia. This is not a generic software license; it is a co-development arrangement aimed squarely at two mission-critical areas: fraud and scam detection and personalized customer services.
CBA’s CEO, Matt Comyn, put it starkly: “To be globally competitive, Australia must embrace this new era of rapid technological change.” By aligning directly with the creator of ChatGPT, CBA is signaling to shareholders, regulators, and employees that it sees GenAI not as an experiment but as infrastructure.
One of the most significant elements of the deal is the progressive rollout of ChatGPT Enterprise to staff. This ensures bankers and internal teams can use AI safely within a secure environment, with enterprise-level controls, audit trails, and data protections. The bank is not waiting for consumer demand to force its hand; it is proactively upskilling its workforce so that by the time customer-facing agents are mature, staff will already be fluent in how to use and supervise them.
Only months earlier, in February 2025, CBA announced a five-year strategic collaboration with AWS to deepen its AI adoption. The agreement makes AWS the bank’s “preferred cloud provider,” and explicitly ties cloud infrastructure to AI innovation.
The payoff is already visible. One of the flagship deployments is a business-customer messaging agent that provides “ChatGPT-style responses” to queries about payments, transactions, and account management. Tens of thousands of SMEs have already interacted with it. For many business owners, this is their first encounter with a generative AI assistant embedded into day-to-day banking.
By pairing OpenAI’s frontier models with AWS’s scalable infrastructure, CBA is essentially building an AI substrate for the bank. Every future deployment — from customer-facing assistants to internal developer tools — will be faster, safer, and more auditable because the plumbing has been laid.
Beyond partnerships, perhaps the clearest signal of CBA’s intent lies in its job postings. Two recent roles illustrate the bank’s direction:
A Senior GenAI Software Engineer role requires expertise in large language models (LLMs), integration of AI into production systems, and fluency with tools such as v0.dev, an AI-native coding assistant. The listing makes clear this is not about experimentation — it is about shipping enterprise-scale AI services.
A Staff Software Engineer — GenAI role similarly emphasizes system architecture, API design, and the ability to embed GenAI into critical financial products.
These postings reflect a recognition that talent is strategy. To build and supervise agentic systems, banks need engineers who understand both financial services complexity and modern AI workflows. The skillset is rare — and CBA is making a public bet that it will secure that talent first.
No transformation is without missteps. In August 2025, CBA made headlines for all the wrong reasons when it attempted to replace 45 call-centre roles with an AI-powered voice bot. Within weeks, the program faltered. Customers complained of poor service, unions mobilized, and the bank ultimately reversed course — apologizing and re-hiring or redeploying affected staff.
For other CIOs, this episode is a cautionary tale: just because you can automate, doesn’t mean you should. AI without human-centred design, adequate piloting, and staged rollout risks damaging trust. The lesson for the sector is not to retreat from AI, but to ensure governance, user experience, and workforce transition plans are built in from the start.
If CBA’s strategy is anchored in partnerships and engineering, ANZ’s is characterized by foresight in infrastructure and people enablement.
At Sibos 2025, ANZ’s David Buckthought articulated a shift that should resonate with every CIO: APIs can no longer be designed only for developers. Increasingly, they must be discoverable and callable by non-human agents.
“It used to be driven by developers and humans… we’re now building our service catalogue so that agents can discover and use them. The real challenge is ensuring identity, entitlements, and that the agent acts on behalf of the consumer within the parameters set.”
This is an extraordinary statement. It suggests that ANZ is actively designing its digital rails for machine-to-machine banking, where agents, not humans, will query balances, initiate transactions, and manage liquidity. The complexity of entitlement, identity, and audit trails in such a world is enormous — but ANZ is signaling it wants to be ready.
Beyond APIs, ANZ is piloting a multi-agent assistant called “amie” for its institutional bankers. Rather than being a general-purpose chatbot, amie is a domain-specific agent trained to act as a “personalized markets analyst,” helping bankers retrieve intelligence, draft insights, and interact with internal systems more efficiently.
The move from horizontal productivity tools (like Microsoft Copilot) to vertical, role-specific agents is significant. It shows that ANZ understands where GenAI creates differentiated value — not in generic summaries, but in augmenting the highly skilled work of relationship managers, traders, and risk specialists.
ANZ’s AI Immersion Centre, launched in partnership with Microsoft, is another cornerstone. Located at its Melbourne headquarters, the Centre will put 3,000 leaders through structured training and deploy 3,000 Copilot licences to staff.
This is not an “innovation lab” tucked away in a corner; it is mainstream workforce transformation. The message is clear: ANZ does not want pockets of GenAI literacy. It wants organizational fluency, so that when agentic interfaces mature, the workforce can adopt them without fear or confusion.
Beyond official announcements, LinkedIn posts from ANZ employees show how deeply AI is seeping into the culture:
A data analyst celebrated ANZ’s presence at the Google Cloud Summit Sydney, showcasing how institutional data teams are embedding conversational AI into advanced analytics, enabling business customers to query data without technical skills.
Carina Parisella, ANZ’s Tribe Lead and Head of Workforce & AI Immersion, frequently shares insights about AI literacy, workforce adaptability, and cultural change — emphasizing that AI adoption is not just about models, but about people.
For CIOs, these signals matter. They suggest ANZ is not only retooling systems but also retooling its culture, ensuring that AI fluency becomes a shared competency rather than a siloed skill.
While CBA and ANZ are most visibly accelerating, their peers are hardly idle.
NAB has positioned itself as an early adopter of GenAI with “considered and balanced” governance frameworks, emphasizing risk overlays and consumer protection.
Westpac is experimenting with developer productivity gains using Azure OpenAI, AWS, and specialized models like Kasisto’s Kai-GPT, signaling that pluralism — using the right model for the right job — will dominate strategy.
Meanwhile, regulators are tightening oversight. APRA’s 2024–25 Corporate Plan emphasizes resilience and cyber safeguards, while ASIC’s Report 798 highlights governance gaps and the need for robust lifecycle controls.
The message is clear: Australian banks can lead, but they must do so with proportional governance and compliance readiness.
With banks investing billions, what should CIOs and technology leaders do now?
Audit API readiness: Are your systems discoverable by agents? Do you have metadata, entitlements, and identity safeguards in place?
Build governed RAG platforms: Centralize retrieval, prompt management, and redaction. Make compliance part of the pipeline.
Start with enterprise copilots, graduate to domain agents: Follow ANZ’s lead — use Copilot and ChatGPT Enterprise to build literacy, then develop banker- or risk-specific agents.
Form cross-functional AI Councils: Include risk, legal, ops, and frontline in every AI product decision.
Invest in telemetry and audit logs: Every prompt, response, and action must be traceable.
Pilot safe but valuable use cases: Fraud monitoring, SME banking queries, reconciliations.
Embed change management: Communicate openly with staff, avoid “automation surprise.”
Measure end-to-end impact: Don’t just count call deflection. Track re-contact rates, NPS, operational resilience.
Design for model plurality: Assume you’ll need multiple models. Avoid lock-in.
Engage regulators early: Demonstrate frameworks before scale-up.
Australian banks are racing ahead — CBA with deep OpenAI/AWS partnerships and active GenAI hiring; ANZ with agentic API design, banker assistants, and immersive workforce transformation. NAB and Westpac are experimenting carefully, while regulators set the guardrails.
The direction of travel is clear: the interface of the future is conversational and agentic. The winners will be those who marry bold moves with balanced governance and human-centered design.
For CIOs and leaders, the challenge is no longer whether to adopt GenAI — it is how to deploy it responsibly, scalably, and in ways that build trust. The decisions made in 2025 will set the trajectory for the next decade of Australian banking.