The strategic case for modernising core banking infrastructure is no longer contested. Most chief technology officers (CTOs) in the region have that conversation settled internally. What remains genuinely difficult is how, and specifically, how to introduce cloud-native architecture and AI capability into systems that serve millions of customers daily, where unplanned downtime is not an acceptable outcome.
Banks across Southeast Asia are under real pressure to modernise. Digital-native competitors have entered the market unburdened by legacy infrastructure, and established banks are watching closely. In response, Singapore and other nations across the region have already signalled openness to financial innovation through regulatory changes. The direction of travel is clear.
What slows things down is rarely a lack of conviction at the leadership level. Instead, it is usually the operational reality of making change inside a system that cannot be paused. A core banking platform processing transactions around the clock, across multiple markets, with compliance obligations in each, does not offer much room for error. The CTO who greenlights a modernisation programme is also the one accountable if it goes wrong.
Why the clean-slate approach keeps disappointing
The instinct to start fresh is understandable. Legacy architecture is genuinely complex, and the appeal of a purpose-built cloud-native system is real. But full-stack migration programmes have a poor track record in enterprise banking for reasons that are by now fairly well documented. Data from McKinsey tells the story plainly: fewer than one in three core banking migrations have successfully made it to full implementation.
The complexity that accumulates in a core banking system over decades does not simply vanish when you build something new. It relocates. Business rules that were never formally documented and edge cases that only surface under specific conditions have a way of reappearing as problems in the new system, often at the worst possible moment. The result is that large-scale replacement programmes tend to run long, cost more than projected, and absorb organisational energy that could have been directed at the business outcomes the modernisation was meant to deliver. By the time the new system is ready, the case that justified it has sometimes already shifted.
See also: Revised framework for single family offices to take effect June 15
Extend the core, don't replace it
The right approach starts from a different premise. Rather than treating legacy infrastructure as something to be displaced, banks’ best bet is to extend it by adding cloud and AI capability on top of foundations that already work, so that banks can access modern functionality without taking on the execution risk of a full migration.
In practice, this means building integration layers that allow newer tools to connect with existing core systems, and developing AI applications that can draw on core banking data without requiring that data to move or that underlying architecture to change. The bank gets the capability, while the infrastructure on which its operations depend stays stable.
See also: JPMorgan doubled private bankers in Singapore team to top 50 — Bloomberg
For a CTO weighing options, this changes the risk calculus materially. The question shifts from "can we afford to undergo this transformation" to "can we afford not to add this capability", which is a considerably easier decision to take to a board.
Legacy infrastructure is not just technical debt
There is a framing problem in how the industry often discusses legacy systems. The word itself carries a bias toward replacement. However, a core banking platform that has been running at scale for decades carries something that a newer architecture has to earn over time: production-hardened reliability, accumulated knowledge of edge cases, and the institutional trust that comes from having processed millions of transactions without failure.
Digital-native banks have real advantages in speed and flexibility. What they are also doing, as they grow, is building toward the operational depth that established systems already have. The gap between a well-designed architecture and a trusted one is measured in years of running in conditions that no test environment fully replicates.
What sustainable modernisation looks like
The banks making meaningful progress on digital transformation share a common characteristic. They are not the ones that underwent the most dramatic overhaul. They are the ones that embedded new capability into how daily operations already work: in credit decisioning, in customer engagement, in risk monitoring, with minimal disruption in operations but huge impact on business.
That kind of incremental, sustainable modernisation is harder to announce than a full transformation programme. It does not make for a clean narrative at a technology conference. But it tends to deliver results that the more dramatic approach does not, and it keeps the lights on while it does.
Cassandra Goh is the CEO of Silverlake
