Artificial intelligence has largely passed the experimentation threshold in financial services. Finastra’s 2026 survey of 1,509 senior leaders across 11 markets found that only 2% of institutions report no AI use. The more revealing question is no longer whether banks are testing AI, but which foundations allow it to become dependable operational infrastructure.
Singapore provides the clearest example in the survey. Nearly two-thirds of its institutions have AI in production, and 73% said they had deployed or improved AI use cases in payments technology during the previous year. The global equivalent was 38%.
That lead is not explained by enthusiasm alone. Singapore’s institutions combine modern infrastructure, deployment discipline, and substantial security investment. Their experience also reveals the constraints that become more visible once pilots begin to scale.
Production AI Rests on Modern Infrastructure
Globally, 31% of institutions report scaled AI across multiple functions and 30% have limited production deployments. Another 27% remain in pilots. Singapore is further ahead, with an additional 35% researching or piloting applications beyond those already operating.
Its cloud position is an important part of that progress. Fifty-five percent of Singapore respondents host all or most infrastructure in the cloud, while 30% use hybrid environments. Scalable compute and accessible data make it easier to move a model from a contained trial into payments, compliance, or risk workflows.
The survey connects AI with a broader modernisation agenda. Eighty-seven percent of institutions globally plan to increase modernisation investment, while Singapore leads in the proportion planning spending increases above 50%. This matters because AI cannot repair fragmented data, brittle interfaces, or unclear ownership by itself. It often exposes those problems more quickly.
Singapore and US institutions most frequently cite compliance and regulatory improvement as an objective, at 43%. Globally, accuracy and error reduction lead at 40%, followed by employee productivity at 37% and risk management at 34%. These are operational goals with measurable baselines, not generic ambitions to “use AI.”
Security Investment Rises With Deployment
Financial AI creates new attack surfaces while also strengthening defence. The research projects a 40% average increase in global security spending during 2026. Singapore institutions are already moving quickly: 62% upgraded fraud detection or transaction monitoring, compared with 48% globally.
Sixty percent also modernised security information, orchestration, and automated-response systems. More than half deployed or upgraded multifactor authentication and biometrics. These investments recognise that deepfakes and generative fraud can raise the quality and volume of attacks.
The next weak point may be the connection between systems. API security and gateway hardening are a priority for 34% globally. As agents access data and act across organisational boundaries, an insecure interface can turn useful automation into a route towards accounts, payments, or customer records.
Security therefore cannot be added after a successful pilot. It is part of the architecture that makes production possible.
Talent Becomes the Hardest Scaling Limit
Singapore’s progress creates intense demand for specialists who can integrate models, govern them, secure cloud systems, and understand regulated workflows. Fifty-four percent of its institutions identify talent shortages as a barrier, the joint-highest result in the survey. Fifty-two percent cite budget constraints.
Partnerships are one response. Globally, 54% of institutions use fintech providers as their default route to AI capability. Buying expertise can accelerate deployment, but banks still need internal ownership of data, risk decisions, and performance.
The survey describes a sector that has crossed the adoption threshold without completing the transformation. Singapore’s advantage is that modernisation, governance, and security are moving alongside AI rather than behind it.
The next measure of leadership will not be the number of systems labelled intelligent. It will be whether institutions can show that production AI improves accuracy, resilience, and customer outcomes while keeping financial trust intact. At that stage, the pilot is over—and operational accountability begins.
