What Asia Pacific banks learned about AI transformation

Financial services spends more on IT than any other industry, yet the sector gets better results when the business, not IT, drives AI adoption. That tells you something important about where AI transformation stalls.

The numbers bear this out. Productivity typically declines right after organizations adopt new technologies, then recovers, but only for firms that successfully navigate the organizational transformation around the technology (McElheran et al, 2025; Brynjolfsson et al, 2020). The difference between firms that capture value from AI and those that don't is more about organizational readiness than technical capability.

So what does organizational readiness for AI actually look like? Financial institutions in the Japan and Asia-Pacific (JAPAC) region offer a useful answer.

Two decades of contrasting choices

Over the past twenty years, many financial institutions bet on standardized processes to manage workflows. Centralized systems and prescribed rulebooks gave senior leadership tight control over operations. Individual employees and middle managers had limited discretion outside what was documented. This created efficient, predictable organizations, well suited for a stable operating environment. Consider what happens when a relationship manager at one of these institutions spots a way to use AI for financial analysis. The idea goes to a department head, then to a technology steering committee, then through a vendor review, then into a quarterly planning cycle. Six months later, the market has moved on.

Financial institutions in JAPAC, broadly, made a different bet. Rapid economic change across the region demanded adaptability over predictability. A population where the average age is roughly 30 (compared to 40–45 in developed markets) expected digital-first experiences and pushed institutions to evolve constantly. The Asian Development Bank has reported that the region is home to 60% of the world's young people (Morris, 2019), and these consumers have high expectations for how their banks use technology. HDFC Ergo responded by launching its "1UP" and "Here" apps, delivering personalized, Gemini-powered financial services across India. Government-owned Bank Rakyat Indonesia used Apigee to launch Agent BRILink, a nationwide network of branchless agents serving underbanked communities.

The result: two different organizational models facing AI transformation. Process-driven institutions, where prescribed workflows tend to be slow and multilayered, and relationship-driven institutions, where trust-based delegation gives employees room to adapt. Chinese business culture, for instance, is grounded in guanxi, interpersonal connections built on trust, reciprocity, and mutual obligation. These network-based relationships naturally make overreliance on rigid rule books impractical and keep decision-making closer to the people who understand the customer. That cultural inclination toward delegation didn't develop as an AI strategy. But it turned out to be one.

Why the organizational model matters now

Agentic AI operates in an unpredictable, continuously shifting environment. Customer behaviors change as people adopt AI for their own needs, while sophisticated security threats emerge just as quickly. In this environment, prescribed processes require constant revision to keep pace. Organizations need employees who can adapt in real time, and that requires genuine delegation. Recent research reinforces this: human-AI systems don't necessarily outperform the best of humans or AI alone, which suggests organizations should evaluate AI not just on productivity metrics but on whether it enhances creativity and reveals opportunities that weren't visible before. Those outcomes require organizational flexibility, not tighter controls.

What JAPAC financial institutions built

Digital and AI adoption in JAPAC accelerated well before AI became a strategic priority for most financial institutions in developed markets. Agile banks and fintechs continually raised the standard of innovation to serve a diverse, mobile-first population. FOMO Pay and Airwallex partnered with Google Cloud to power cashless payments across the region. In Hong Kong, Bank of East Asia built a unified data lakehouse on BigQuery to enable real-time customer decisioning.

What connects these examples is that these organizations delegate decisions to employees working directly with customers. They move quickly, and treat AI as a business transformation rather than a traditional IT project.

The institutions gaining traction in JAPAC understood that the real work is combining three capabilities under a single leader: deep knowledge of how the business actually makes money, the transformation skillset to redesign workflows and incentivize genuine adoption, and broad technology awareness to make sound architectural choices (Afanasyev, Milind, 2026). That technology awareness matters because JAPAC institutions operate at a scale and transaction diversity that demands practical architectural decisions. The region accounts for roughly half of non-cash transactions globally (Capgemini, 2025), and technology choices there were shaped by volume and speed requirements, not by legacy vendor relationships — which kept architectures leaner and more adaptable.

A diagnostic for your organization

You don't need to adopt guanxi to build an AI-ready organization. But you do need to honestly assess where your institution falls on a few dimensions, and take specific steps based on what you find:

Decision speed. When a frontline team identifies an opportunity to apply AI to a customer problem, how many approvals are required before they can test it? If the answer involves a steering committee and a quarterly review cycle, that's a structural constraint. Start by giving two or three business teams a defined budget and authority to test AI applications without enterprise-wide approval.

Employee discretion. Do your people have genuine authority to adapt how they work, or are they executing documented processes? AI transformation requires hundreds of small judgment calls per week. Organizations that concentrate decision-making at the top create bottlenecks that slow experimentation. Identify the five decisions your frontline teams escalate most often and formally delegate them.

How you define "AI project." If your AI initiatives live in IT and are measured on technical KPIs like model accuracy, you're likely missing the business transformation that drives actual returns. The institutions succeeding in JAPAC treat AI as a business strategy with technology components, led by someone who understands the business domain, the organizational transformation required, and the technology. Appoint AI initiative owners from business lines, not from technology, and measure them on customer and revenue outcomes.

Feedback loops. How quickly do results from AI experiments reach the people who can act on them? Process-driven organizations tend to route feedback through management hierarchies. Relationship-driven organizations share results laterally and adjust faster. Create direct channels between teams running AI experiments and the business leaders who fund them, weekly, not quarterly.

The gap between firms that capture value from AI and those that don't is an organizational one. Not every JAPAC institution succeeded; those that treated AI as purely a technology project faced the same stalls as their counterparts in developed markets. But the financial institutions that learned this earliest didn't succeed because they had better technology. They succeeded because their organizations were already built for rapid, decentralized adaptation.