The financial institutions that aren’t using AI can now be counted in the low single digits. That’s not a milestone — it’s a verdict.
According to Finastra’s Financial Services State of the Nation 2026 report, which surveyed 1,509 senior executives across 11 markets, just 2% of financial institutions globally report no AI use whatsoever. The adoption debate, which consumed boardrooms and conference stages for the better part of a decade, is functionally closed. What has replaced it is considerably harder: the question of whether AI deployments actually work at scale, and whether institutions can govern them well enough to survive the consequences when they don’t.
Six in ten institutions improved their AI capabilities over the past year. Nearly half — 43% — identify AI as their single most important innovation lever. The technology has embedded itself across fraud detection, compliance automation, document processing, and customer engagement. These are not experimental sandboxes. They are operational cores.
But near-universal adoption creates its own problem. When everyone has deployed, deployment itself stops being a competitive advantage. The institutions pulling ahead are those that have moved beyond isolated pilots toward enterprise-wide integration — and that distinction is proving harder to achieve than most expected.
The report’s top four active AI use cases reflect where the real work is happening: risk management and fraud detection at 71%, data analysis and reporting at 71%, customer service and support at 69%, and document intelligence at 69%. The next phase of investment is oriented around AI-driven personalisation, agentic AI for workflow automation, and — critically — AI model governance and explainability. That final priority signals something important. As AI decisions grow more consequential and regulatory scrutiny intensifies, the ability to audit, explain, and defend algorithmic outputs is no longer optional infrastructure. It is the price of operating responsibly.
None of this happens without the systems underneath. Finastra’s data makes the dependency explicit: 87% of institutions plan infrastructure modernisation investments over the next 12 months, not as parallel initiatives but as the foundation that determines how far AI can actually go. Cloud adoption, data platform upgrades, and core banking modernisation are all accelerating in direct service of AI scalability.
The barriers, however, are stubbornly human. Talent shortages are cited by 43% of institutions as the primary obstacle — acutely so in Singapore at 54%, the UAE at 51%, and both Japan and the United States at 50%. Budget constraints follow closely. Faced with those constraints, 54% of respondents have made fintech partnerships their default modernisation strategy, trading build costs for speed and specialisation.
Regional differences reveal how uneven this moment actually is. Vietnam leads the Asia-Pacific on active AI deployment at 74%, propelled by financial inclusion pressures and demand for faster payment and lending infrastructure. Singapore is scaling cloud and personalisation investment aggressively, with planned spending increases exceeding 50% year-on-year. Japan sits at the opposite end — only 39% report active AI deployment, reflecting legacy system constraints and a cultural inclination toward deliberate, incremental change rather than accelerated transformation.
With 63% of institutions already running or piloting agentic AI — systems capable of autonomous, multi-step decision-making — the trajectory is unmistakable. So is the weight it carries. Autonomous systems that act without human intervention at each step demand a fundamentally different standard of accountability, transparency, and control. For enterprise leaders, as Finastra’s CEO Chris Walters framed it, the imperative is to move quickly and responsibly simultaneously, as regulators tighten scrutiny and customers expect services that are reliable, secure, and personal every single time. Speed without governance is not a strategy. It is an exposure.
Photo by Nguyen Dang Hoang Nhu on Unsplash
Source: Original reporting
