Australia’s big banks have spent the past two years talking up artificial intelligence, and now the market is asking them to show their working. Westpac, the country’s oldest bank and one of the four pillars of the domestic financial system, has become the latest to lean heavily on so-called AI agents as it tries to convince increasingly impatient investors that the money flowing into the technology is translating into genuine cost savings rather than expensive experiments.
The shift matters because the mood among shareholders has changed. During the early hype cycle, simply announcing an AI strategy was enough to reassure the market that a company was keeping pace. That grace period is closing. According to reporting from Kalkine, Westpac is turning to AI agents specifically as investors look for evidence that the outlay is showing up on the bottom line. The distinction is important. Investors are no longer content to hear that a bank is trialling chatbots or drafting internal policies. They want to see the productivity gains landing in reported results.
From assistants to agents
The language here is deliberate. For most of the past decade, banks deployed AI in the background, powering fraud detection, credit scoring and the recommendation engines that nudge customers towards a savings product. The newer wave, described as agentic AI, is different in kind rather than degree. Rather than answering a single query, an AI agent can be handed a task and left to work through the steps required to complete it, calling on data, systems and other tools as it goes. In a bank, that might mean pulling together the documentation for a loan assessment, reconciling records across departments or triaging the flood of customer service requests that arrive every day.
For an institution the size of Westpac, with tens of thousands of staff and a cost base measured in the billions, even modest efficiency gains across those repetitive processes add up quickly. That is precisely the promise the bank is making, and precisely the promise the market now wants tested against real figures. Westpac’s leadership, including chief executive Anthony Miller, has framed technology simplification as central to the bank’s future competitiveness, and agentic AI sits squarely inside that agenda.
Two ways to read the bet
There are two broad camps forming around this kind of investment, and both have a fair case. The optimists argue that banks are almost ideal candidates for agentic AI. Their work is heavily document-driven, rules-based and repetitive, which is exactly the terrain where agents perform best. Supporters point out that the alternative, leaving legacy processes untouched, carries its own mounting cost as competitors and nimble fintechs chip away at the edges. On this view, the banks that move early and prove out the savings will enjoy a structural advantage, while laggards will be left carrying bloated cost bases.
The sceptics counter that the sector has a long and expensive history of technology projects that promised transformation and delivered overruns. They note that agentic systems are still immature, that they can behave unpredictably, and that a bank cannot afford the reputational damage of an AI agent making an unsupervised decision that affects a customer’s money or credit. In a tightly regulated industry, every efficiency has to be weighed against the risk of getting it wrong at scale. That tension explains why banks are moving carefully, running agents inside guardrails and keeping humans firmly in the loop, even as they talk publicly about the upside.
That caution is not hypothetical for the sector. National Australia Bank has been openly testing safeguards around its own AI agents before letting them loose on live banking tasks, a sign that the major lenders understand the technology’s promise cannot be separated from its risks. The competitive pressure is real, but so is the shared awareness that a single high-profile failure could sour public trust and invite tougher regulatory scrutiny for everyone.
Why this matters for Australia
Westpac’s approach is worth watching well beyond its own shareholder register. The big four banks are among the largest technology employers and buyers in the country, and their choices ripple through the entire local ecosystem of software vendors, consultancies and startups. If agentic AI proves it can deliver measurable savings inside a bank of Westpac’s scale, it will accelerate adoption across insurers, superannuation funds and government agencies that face similar mountains of document-heavy, process-driven work. Australian technology firms building agent platforms and integration tools stand to benefit from that momentum, while offshore providers will compete hard for the same contracts.
There is also a workforce dimension that Australia cannot ignore. Banking has historically been a major source of stable, white-collar employment, particularly in operations and customer service roles that agentic AI is explicitly designed to streamline. How Westpac and its peers manage that transition, whether they redeploy and retrain staff or simply shrink headcount, will feed directly into national debates about the labour market impact of AI. Unions, regulators and policymakers will be paying close attention to whether the promised cost savings come from smarter processes or from thinner ranks.
Regulators have their own stake. The Australian Prudential Regulation Authority and the corporate watchdog have both signalled growing interest in how financial institutions govern automated decision-making. An AI agent that touches lending, hardship assessments or complaints handling raises questions about accountability, auditability and fairness that existing rules were not written to answer. As banks push agents into more consequential territory, the pressure for clear guidance will only build.
What comes next
The immediate test is a reporting one. Investors will be looking at Westpac’s upcoming results and market updates for concrete indicators, lower operating costs, improved cost-to-income ratios, or specific processes that have been automated and quantified. Vague references to AI initiatives are unlikely to satisfy a market that has watched valuations swing sharply on questions about whether AI spending is disciplined or speculative.
The broader story is about credibility. The banks that can point to hard numbers will find it easier to justify continued investment and to keep shareholders onside. Those that cannot may face uncomfortable questions about whether their AI ambitions are strategy or theatre. For Westpac, the coming months are a chance to move the conversation from potential to proof, and the rest of corporate Australia will be taking notes on how convincingly it manages that shift.
Sources: Kalkine
