By Remus Lim (pictured), Senior Vice President, Asia Pacific & Japan, Cloudera
In this new phase of Artificial Intelligence (AI) adoption, ideas and pilot models are no longer enough. Increasingly, operations leaders and boards want to see AI at full-scale production with measurable returns. However, that’s proving to be a more difficult task than anticipated.
IDC’s Technology Leaders Survey 2026 showed that across Asia Pacific, 37% of organizations are investing aggressively in AI due to fear of falling behind, often with limited evaluation. At the same time, poor data quality is the top reason AI fails to deliver ROI globally, cited by 51% of organizations globally where AI underperformed expectations.
AI value depends on more than investment appetite. For financial services, the gap between “having data” and “driving value” often boils down to latency. While many institutions have spent the last decade perfecting “lakehouse” models for static data, the strongest AI use cases require a fundamental shift toward real-time data or data in motion.
The Case for Real-Time AI in Financial Services
The driver for real-time data goes deeper than technical speed; it’s about repairing a massive operational leak. Financial institutions have long tolerated “dark hours” where data sits idle, waiting for overnight batch processing. In recent years, this delay has become a competitive liability. Real-time AI can help financial institutions act faster across critical use cases, from fraud prevention and security to customer experience, data management, platform modernization and reporting. By applying AI to these areas, institutions can improve forecasting, streamline complex reporting and support faster, more efficient operations.
For the financial service sector, scaling real-time AI responsibly requires three core capabilities: hybrid flexibility, strong governance, and AI and model sovereignty.
Running AI where it makes the most sense
To scale AI responsibly, financial institutions need flexibility over where workloads run. Real-time AI in financial services often demands “always-on” compute to support use cases like payments processing, risk modeling, and trading operations. While cloud environments offer agility for experimentation, the total cost of ownership (TCO) for stable, high-throughput workloads like transaction processing or regulatory reporting can be significantly lower on premises.
A hybrid data platform enables data and application portability so institutions can run latency-sensitive and cost-intensive workloads where they make the most financial and operational sense. This is why hybrid AI deployment has become an essential strategy for the sector, as revealed in Cloudera’s report ‘How Financial Services Institutions Are Scaling AI’ which found that 91% of financial services organizations rate a hybrid approach as highly valuable.
Governance as the foundation for trusted AI
A major obstacle for AI in financial services is the difficulty data scientists and risk teams face in discovering, trusting, and governing data in motion. As financial institutions operate under constant regulatory scrutiny, AI-driven decisions must be accurate, explainable and protected. Governance gives institutions the confidence to scale AI safely, balancing speed and innovation with the oversight expected by regulators, boards and customers.
By extending consistent governance, lineage, cataloging, and security controls to real-time data, Cloudera ensures that the data used for decisions is as auditable and trustworthy as data at rest. This is critical for meeting compliance requirements and supporting explainable AI.
Protecting data, models and decision-making
As AI adoption matures, institutions are moving beyond data residency into the era of AI and model sovereignty. IDC expects 80% of the 2,000 largest enterprises in Asia Pacific (excluding Japan) to prioritize AI sovereignty for mission-critical workloads by 2028. For financial institutions, this means ensuring that both data and models remain within required geographic or regulatory boundaries,supporting compliance with evolving data protection and financial regulations. Enterprise-grade models with clear provenance can help institutions improve transparency, reduce risk and meet rising expectations for accountable AI.
Bringing AI closer to the point of decision
To enable real-time decisioning, such as fraud prevention, credit adjudication, and trade validation, financial institutions need to move beyond batch processing to event-driven architectures that ensure that data changes and updates are propagated in real-time.
Edge AI can support this shift by moving decision-making closer to the point of interaction such as the point of sale, an ATM, or within a mobile app. This enables real-time fraud detection and transaction validation, allowing institutions to stop fraudulent activity before a transaction is completed, rather than identifying it after settlement.
Not every financial services use case requires a large-scale model. Small Language Models (SLMs)under 10B parameters can be deployed at the edge or within controlled environments to support customer authentication, document processing, and compliance checks, delivering lower latency, improved privacy and reduced infrastructure costs.
Building the foundation for AI at scale
Real-time data is now the essential foundation of modern banking, payments, insurance, and capital markets operations. It transforms static reporting into continuous, event-driven decisioning, enabling dynamic workflows that adapt in real-time. Financial institutions can turn real-time data into a permanent competitive advantage without losing the control, governance, and resilience this sector demands will be best placed to unlock AI’s full value at scale.