Chapter 2: AI, data and innovation

Not everyone in fintech is onboard the AI hype train. While the survey does not explicitly ask how (or if) fintechs measure the return on investment of AI products, conversations with industry leaders suggest AI value is uneven. Some admit to implementing AI tools out of fear of missing out, only to wonder whether competitors are extracting greater value from the same technologies.

As one participant put it: “We’ve got Claude, but are we really getting the benefit?”

Another says, “AI is making people slightly more efficient, but if anyone is claiming a massive step change, they are either overstating the impact or paying a massive bill for tokens.”

Cost control is another reason for cautious AI adoption. Some participants express concerns about being 'Microsofted'; drawn in at low cost and then becoming dependent on a small number of providers before facing sustained price increases.

Others are less concerned, arguing that rapid advances in open-source and lower-cost models provide a natural competitive check on pricing power.

Regardless of which view proves correct, the message from industry leaders is that individual-tier subscriptions are more than sufficient for most current use cases, and that fintechs should avoid premature enterprise-level commitments or heavy token-based billing before the value case is proven.

Questions around AI accuracy and reliability also emerge repeatedly

Interestingly, some participants suggest that AI value is inversely proportional to a company's stage of growth. For start-ups and emerging fintechs, AI is largely viewed as an accelerator. Smaller teams can build products faster, test ideas more cheaply and compete above their weight.

For larger organisations, however, the equation becomes considerably more complex. Mature businesses must consider governance, integration with their existing tech stack, regulatory obligations, liability and cyber security.

A sole founder may view the ability to launch a new feature overnight as a competitive advantage. An organisation responsible for customer funds, regulated data and critical infrastructure may see that same capability and hear alarm bells.

As one participant observes:

“Where is the cyber insurance for that?”

Some leaders take an even more cynical view, arguing that the future value of human expertise may increasingly lie in accountability. Financial advice, lending decisions and compliance activities ultimately require someone to take responsibility for outcomes. An AI cannot be held accountable for an error or regulatory breach. A human can.

AI regulation is still evolving, but technology regularly outpaces legislative action, and fintechs cannot afford to wait. As AI moves from experimentation into production environments, governance is rapidly becoming a board-level issue.

This is particularly significant within financial services, where explainability, auditability and accountability are critical.

Data-focused fintechs appear highly conscious of governance, with established privacy, Consumer Data Right (CDR) and security frameworks widely adopted among respondents. Privacy-preserving techniques are also becoming relevant to product roadmaps, although approaches to explainable AI remain uneven, suggesting governance frameworks may currently be more mature than the mechanisms used to make AI models transparent to customers and stakeholders.

Cyber security is also a major concern, as clients increasingly demand assurance that embedded AI does not become a back door.

Multiple organisations adopt explicit rules prohibiting staff from entering sensitive information into public AI systems. Others create ring-fenced environments and cloud architectures to maintain control over data.

When discussing solutions to unreliable outputs, one roundtable participant describes an “AI sausage factory,” where highly trained agents operate in a production line, each layering on their field of expertise, with human checkpoints for quality control and a failsafe that scraps any output containing errors and sends it back to start over from the beginning. This may be an effective way to ensure accuracy but does raise questions about how many tokens it costs to produce a sausage.

For fintechs providing business tools and software as a service (SaaS), many already view AI as a bigger competitive threat than traditional rivals.

If every fintech has access to the same AI tools, it can’t be a differentiator. Data, on the other hand, is rapidly increasing in value.

As software becomes easier to build, unique datasets, relationships built on trust, and access to distribution channels will be the most valuable assets of the AI era.

As one leader notes, “many AI businesses are little more than wrappers around existing models, while organisations with access to unique proprietary datasets can create significantly more defensible value.”

This creates an opportunity for stronger collaboration between fintechs and established financial institutions that hold large volumes of proprietary data. Programs such as the Buy Australian AI Partnership Program can help facilitate these connections by bringing together innovators, data holders and potential customers to explore practical AI applications.

It is also a great example of the non-financial support government can provide to foster collaboration, encourage knowledge sharing and create pathways for homegrown innovators to engage with large organisations. We hope to see similar programs conducted in the future.

Of course, the ability to capitalise on this advantage hinges on the quality of information flowing through financial systems. One participant argues that the most important AI challenge is no longer model capability but data integrity itself.

This focus on data quality and accessibility naturally leads to one of the most significant infrastructure discussions in Australian fintech: the future of the CDR.