GenAI, banks, and striking a balance

Designing the ideal artificial intelligence experience is all about balance.

"ANZ’s amie is a new, multi-agent chatbot that will provide both our commercial and institutional bankers answers to market-related questions. This will free them up to do more of what they do best: help our customers."

Earlier this month we revealed the upcoming launch of amie (short for ANZ Markets Insights Engine). At ANZ, we’ve known for a while generative artificial intelligence — or genAI — is not just changing how we work, but the way we interact with all our community and industry stakeholders. 

In April, a United Nations report suggested the value of the global AI market could grow to $US4.8 trillion by 2033, and enhance 27 per cent of all jobs in advanced economies. The technology’s potential is clear.

ANZ’s amie is a new, multi-agent chatbot that will provide both our commercial and institutional bankers answers to market-related questions, offering them guidance around research, news, and data. This will free them up to do more of what they do best: help our customers.

amie is the product of years’ worth of work across ANZ, as well as with valued partners like Amazon Web Services (AWS). The process taught us a lot of interesting things that will shape our approach around genAI — and how we’ll work with our customers in the future.

UX

Working with amie has taught us a lot about the user experience. The first is that balancing time against accuracy and quality of output is critical.

The importance of output quality in genAI has been known for some time. As far back as 2018, Gartner suggested 85 per cent of AI projects were at risk of errors due to unkept data. Models have improved dramatically since then, but since the launch of ChatGPT user expectations around genAI have also grown.

At ANZ, when our bankers speak with business customers, it’s clear to us they want relevant information in seconds, not minutes. But we know that speed can’t come at the cost of accuracy.

In order to strike the right balance, at ANZ, we’ve found using a variety of tools can help manage these many elements. amie uses separate but aligned processes for customer data requests, more complex analytics, news and research retrieval, and complex knowledge navigation, among others. The importance of that diversity has been a key lesson for us at ANZ through the process.

Bringing amie to life has also reinforced to us the importance of personalisation. At ANZ we know no matter what industry you operate in, customers expect a highly personalised experience. We’re actively seeking this in all our customer interactions, but conscious that it needs to be done in a safe and considered way – meaning balance is, once again, critical.

When it comes to genAI, neural networks are great at capturing complex interactions, but that often comes at the cost of transparency and accessibility. Large language models have improved accessibility, and we are excited by some of the research into better understanding large language models, particularly around interpretability and alignment.
Through our work so far, we have found genAI very much complements traditional machine-learning methods, helping create content that quickly responds to shifting customer preferences. We’re excited about what that means for the future.

What’s next?

Another key lesson for us was how genAI can demonstrate real business value with focused applications that scale progressively.

While success does require thoughtful data investment, the barrier to entry is much lower than previous technology waves.

The evolution toward agentic AI – where agents work to complete transactions on behalf of users - is very exciting. The scale of this change is not unlike how e-commerce applications transformed the internet from an infrastructure into a commercial platform, creating entirely new ways of doing business.

At ANZ, we're seeing strong results in our testing, and while we're still in the early stages, the learning curve is accelerating rapidly.  The combination of lower entry costs, increased speed, and proven wins creates a compelling path forward for businesses ready to embrace this future.

Perhaps the most exciting lesson for us is how agents like amie can help share the deep knowledge of the subject matter experts at organisations like ANZ. This critical experience and understanding can be distributed across a wide range of people in the organisation, scaling expertise rather than just the underlying technology.

We know genAI isn't just about efficiency.  When used well, it can be a tool that delivers personalised expertise to every customer interaction, helping us better understand complex client needs.

Intersection

At ANZ, we believe the future of customer experience lies at the intersection of personalisation, predictive analytics, and agentic AI. These things are powerful in isolation, but when combined have the potential to transform financial services.

ANZ’s markets business processes over three billion data points every day, running around 1,000 customer analytics across thousands of combinations. And every customer need is constantly changing in line with global markets. genAI is helping us solve this puzzle in real time, and at scale. 

Our approach through the entire process — on amie and with genAI as a rule — has been iterative, considered, and balanced, with ethics, security, and privacy all non-negotiable elements.

At ANZ, we're building for the future with deliberate, safe steps, creating an exceptional customer experience, and helping our customers focus on what they do best: growing their businesses. 

Tim Moloney is Head of Digital Strategy & Head of Financial Institutions Sales at ANZ