Why Digital Banks in Southeast Asia Are Rethinking Core Banking for the AI Era

Digital banking in Southeast Asia was once built on speed.

For quite some time, they held on to the formula of launch first, refine later, and it did work, especially during their early stages when their volumes were much lower, and regulators were still patching and adapting to digital models.

But the stale recipe seems to show cracks as these banking institutions grow, particularly when artificial intelligence moves from experimentation into their core operations.

That pressure is especially visible in larger ASEAN markets such as Indonesia and the Philippines, where things such as digital wallets and QR payments are becoming part of everyday financial activity rather than niche digital channels.

For instance, in the Philippines, BSP data showed that InstaPay and PESONet processed a combined US$401.8 billion (PHP24.74 trillion) in 2025, up 42.02% from 2024, while InstaPay transaction volume surged to 4.66 billion transactions.

In Indonesia, QRIS had reached 57 million users and 39.3 million merchants by the first half of 2025, with transactions hitting US$33.2 billion (IDR579 trillion), while fintech P2P lending outstanding financing grew 31.06% year-on-year to US$4.6 billion (IDR80.07 trillion) as of February 2025.

This is where the regional growth story becomes more complicated. More customers means more products and more products leads to more digital touchpoints which is a good thing as it creates new opportunities, but they also place more pressure on the systems underneath.

Thus, there’s no surprise when investment in AI shoots up, with spending by BFSIs in general expected to reach nearly US$97 billion by 2027.

The reason could also be that financial institutions that fully embrace AI could realise a 15-percentage-point improvement in their efficiency ratios, as per research by PwC suggests.

And yet, despite being in the 6th year after AI burst onto the scene, many of these initiatives are not making it past controlled environments.

With spendings reaching the billion-dollar mark, it does not make sense to say that the issue revolves around the lack of ambition. No, the root of the problem sits deeper.

When examined closely, it all boils down to how the systems are struggling to keep up.

And a gap now slowly forms. Institutions are now stuck between what they are trying to deploy and what their infrastructure can reliably support.

Growth is still happening, but it is bringing a different set of problems with it.

The test now is not only whether digital financial institutions can grow, but whether they can manage that growth under tighter scrutiny and rising operational complexity.

AI Is Moving Fast. Execution Is Not

Across the region, banks are actively exploring artificial intelligence, particularly in areas like credit, risk, fraud detection and customer engagement.

The early results can be encouraging where institutions can see that their models performing well in controlled environments, where data is clean and conditions are predictable.

However, the difficulties usually appear when those same models are pushed into production.

This is where many initiatives stall, often described within the industry as “pilot purgatory” which means that the limitation tends to sit in the surrounding environment rather than the models themselves.

The gap is showing up globally as seen in BCG’s 2025 research which found that only 5% of companies are achieving AI value at scale, while 60% are seeing little or no material value despite investment.

The finding is especially relevant for banks and fintech players in Southeast Asia, where AI ambitions often run into fragmented data, brittle workflows and systems that were never designed for real-time decisioning.

While models can be developed within a few months, it can take much longer for frontline staff to trust them, understand their outputs and integrate them into daily workflows.

Even when the technology is ready, organisational readiness often lags behind.

In practice, AI only works when the underlying environment can support it, something many institutions are still working towards.

When those conditions are not in place, performance becomes unreliable.

In that sense, AI is acting less as a silver bullet and more as a diagnostic tool. It exposes weaknesses that were previously easier to work around.

Despite these hurdles, leading institutions are beginning to demonstrate the financial upside of moving beyond experimentation.

For example, DBS Bank reported generating approximately US$565 million in 2024 from more than 350 specific AI use cases deployed across its operations.

Within wealth management, JP Morgan said that AI-powered coaching tools have helped advisors improve productivity, contributing to double-digit sales growth by reducing research time and enabling faster client engagement.

More broadly, generative AI-driven personalisation is starting to translate into measurable financial impact, with some institutions reporting modest gains in revenue and improvements in return on equity as adoption scales.

Where the Real Friction Sits

Much of that friction can be traced back to how systems have evolved.

Over the past decade, many institutions focused on improving what customers interact with, which has resulted to why most interactions we have today, feel faster and easier.

Underneath those improvements, complexity has been gently accumulating.

Much of the infrastructure still runs in silos, often disjointed and not fully connected, with each system maintaining its own version of the same information.

Plus, in order to keep them aligned requires constant reconciliation, which most certainly takes time and introduces risk.

Some technology leaders refer to this as an “integration tax”, a not-so a one-off issue that time and again accumulates as systems expand.

This is where technical debt begins to crowd out innovation.

Accenture estimates that banks can spend around 70% of their IT budgets on maintaining legacy systems, leaving a much smaller portion for new products, data infrastructure and AI enablement.

Integrating just 200 legacy services can cost a bank US$480,000 in upfront expenses alone. This process consumes up to 1,000 developer days and commits the institution to an additional US$100,000 in yearly maintenance costs.

Over time, banks slide into what some describe as “technical bankruptcy”, where maintenance work begins to paralyse innovation.

The strain becomes more visible when AI is introduced.

Legacy infrastructure, especially those built around batch processing, struggles to keep up with continuous data flows, and over time the results become harder to trust.

What once felt manageable starts to limit how far an institution can go.

True core banking modernisation tackles the problem at its starting point by replacing aging systems with a cloud-native, API-first platform.

A modern core communicates seamlessly with AI models, but only if a governed data layer provides a real-time, read-only replica of the production database.

Without this, AI risks drawing on poor-quality data, producing biased or inaccurate predictions.

Regulation Is Tightening the Boundaries

At the same time, regulators across Southeast Asia are raising expectations.

Authorities including the Monetary Authority of Singapore, Indonesia’s Otoritas Jasa Keuangan, and Bank Negara Malaysia are placing greater weight on how institutions manage their systems under stress, not just how quickly they innovate.

Technology now has migrated away from the back office directly onto the balance sheet.

Driven by recent enforcement actions, the perception of system failures has shifted from isolated incidents to material risks that carry direct consequences for capital requirements.

The banking industry now treats system fragility as a tangible financial risk with a measurable impact on the balance sheet.

The regulatory response to repeated service disruptions at DBS Bank illustrates this change in posture.

Following a series of outages, the Monetary Authority of Singapore required the bank to hold additional US$ 1.2 billion in capital by increasing its operational risk-weighted assets.

Similar signals are emerging across the region.

In Malaysia, Bank Negara Malaysia imposed significant administrative monetary penalties in 2025 after several institutions failed to maintain high availability for critical systems.

Bank Simpanan Nasional was fined US$251,000 (RM995,000), while Bank Kerjasama Rakyat Malaysia faced a penalty of US$719,000 (RM2.85 million) after repeated unplanned downtimes exceeded regulatory thresholds.

In Indonesia, new rules introduced for 2025 require financial aggregators and digital service providers to maintain data and recovery centres within the country, with potential fines of up to US$57,000 (IDR1 billion) for non-compliance.

The move made it clear that system fragility can directly affect a bank’s ability to deploy capital and expand.

This changes how institutions approach compliance.

It becomes embedded in day-to-day operations. Every change leaves a trace, every decision must be explainable, and if needed, reversible.

That level of visibility is becoming part of the operating baseline.

For Oradian, this is where the conversation around core banking modernisation becomes more urgent.

The issue is no longer whether financial institutions can launch digital services quickly, but whether they can keep those services consistent, explainable and resilient as volumes rise.

“Across Southeast Asia, we’re seeing a new phase of growth in digital financial services, one where institutions are moving from experimentation to building truly scalable, intelligent operations. AI is accelerating this shift, but it also highlights the importance of having the right foundation in place. At Oradian, we see that the institutions making the most progress are those investing in modern, cloud-native cores that act as a single source of truth, connecting data, operations, and compliance in real time. This foundation gives them the confidence to innovate faster, scale sustainably, and operate reliably as they grow,” said Antonio Separovic, CEO of Oradian.

Why Growth Now Requires More Discipline

Earlier phases of digital banking rewarded speed. Today, speed without control introduces a different kind of risk.

As institutions scale, small inconsistencies can compound quickly.

What works at a smaller volume can become unstable under pressure thus, fixing those issues later is rarely straightforward.

Consequently, the strategic priority for core systems has shifted from rapid deployment to structural integrity.

Rather than treating them as background infrastructure, there is a growing recognition that the core acts as the control centre of the organisation.

It has become the definitive environment for operational integrity, ensuring that data, logic, and compliance are inseparable.

Ultimately, the core system must serve as the primary site of institutional control.

A stable foundation simplifies change management, and institutions that strike this balance prove their resilience by thriving under pressure.

A More Deliberate Architecture Is Taking Shape

Taken together, these pressures point to the same conclusion. AI needs trusted, real-time data. Scale requires systems that do not fracture as volumes rise.

Regulation demands traceability, resilience and control.

Core modernisation sits at the centre of all three because it determines whether innovation can move safely from experiment to production.

In response, a clearer structure is starting to emerge.

Core systems are increasingly being treated as the control layer of the institution, keeping records consistent while giving banks the auditability regulators now expect.

Around that core, banks can build more flexible layers for customer engagement and decisioning without putting critical records or compliance processes at risk.

The separation is deliberate.

It allows teams to iterate on products and models without placing critical records or compliance processes at risk. Changes can happen, but within defined boundaries.

This approach follows broader shifts toward API-driven, composable architectures, enabling different components to interact without tight coupling.

It may sound subtle, but in practice, it changes how a bank operates under growth.

Infrastructure Is Becoming a Strategic Question

Decisions about infrastructure now also carry weight beyond technology teams.

These choices effectively set the ceiling for a bank’s growth. Not only that, it can aid in determining its ability to pivot when the market shifts without compromising on safety.

Cloud-native, API-based platforms let banks deploy and scale systems with greater flexibility, driving their growing adoption.

They allow institutions to handle fluctuations in demand and introduce updates without extended downtime.

Furthermore, access to data is just as important.

Older workflows often depend on manual extraction or scheduled transfers, which slows things down and increases the chance of errors.

Newer approaches can help provide controlled access to live data environments. It will allow teams to work with current information without compromising stability.

Sustainable growth now hinges on basic feasibility rather than minor optimisations.

The Next Phase Will Favour Those Who Can Operate Under Pressure

All of this is contributing to a broader shift in how core banking is understood, as it is no longer seen merely as a system of record.

Most institutions now treat their core as a complete operating model that defines how the bank actually functions while it grows.

Using the core as a blueprint ensures that data stays accurate and different systems work together properly as the business expands.

Performance still matters, but the focus has shifted toward long-term stability.

A new category of platforms is now emerging to help banks scale under heavy regulation without losing control of their daily work.

These providers such as Oradian go beyond just replacing old software. They can help create a reliable environment where banks can safely launch new tools like AI without the risk of system failures.

The framing now becomes different because the problem has changed.

As the global digital banking market is forecast to reach US$87.8 billion by 2034, the institutions that thrive will be those that view unified data and cloud-native architecture as their strategic core.

The future of banking is no longer a question of digital presence, but of digital intelligence.

And that intelligence is only as strong as the core that powers it.

Banks that handle this shift well often move forward with fewer surprises.

Those remaining institutions who don’t, may find that the shortcuts they once relied on become harder to ignore as they grow.

Featured image: Edited by Fintech News Singapore based on images by 21ST, rawpixel.com and Who is Danny via Freepik.