[Sung-hyun Cho (조성현), head of technology at Databricks Korea] Over the past decade, AI in financial services has been seen as a next-generation technology that will determine future competitiveness. But that view now needs to change.

As of 2026, AI has effectively become universal across the financial industry. Most major banks, insurers and asset managers are piloting AI or applying it in day-to-day work, and generative AI in particular is becoming a practical business tool much faster than many expected. About 94 percent of financial services companies worldwide are piloting or running generative AI in core tasks such as cybersecurity, pricing, risk management and personalised products.

But the impact of adopting AI is not the same for every company. Financial institutions that have successfully embedded AI are increasing the speed of decision-making and cutting operating costs by as much as 20 percent. Most institutions, by contrast, are still not seeing the results they expected. The issue is not model performance or a lack of strategy, but the execution capability to scale into real working environments.

The real bottleneck is not technology, but a fragmented structure.

The belief that AI projects fail because the model itself lacks performance is a common misconception. Many prototypes and proofs of concept work well in constrained environments, but only a limited number make it into production and translate into business results. The gap between lab settings and real operations stems from complex, fragmented data infrastructure. South Korean financial institutions face a particular environment in which they must comply with network separation, strict security rules and financial regulators' AI governance guidelines.

Existing financial IT systems were not designed to support AI workflows that operate in real time while applying consistent governance. When institutions try to scale enterprise-wide capabilities such as fraud detection systems, dynamic pricing and personalised services, gaps emerge in data consistency, lineage and control frameworks. That undermines trust in AI.

What sets leading financial institutions apart.

Institutions that deliver results do not obsess over building superior AI models alone. They focus on building the foundational environment that allows AI to operate across the enterprise.

They manage data as a strategic core asset, embed governance across the entire data and model pipeline rather than at the final stage, and build integrated operating frameworks across data, analytics and AI organisations to collaborate closely. This approach speeds the shift into operations and wins trust from business teams, leading to tangible results in which generative AI moves beyond detecting security threats to automating responses within a consistent data environment.

One operating system reflected in eight trends.

Databricks' "2026 Financial Services Outlook" report presents eight key trends that are changing the competitive landscape in finance. Each trend may already sound familiar. But when linked together, they point to a common direction: the financial industry is shifting toward a new operating system.

For example, real-time anomaly transaction detection is built on streaming data and a governed data environment. Customer 360, which integrates and manages customer data, also requires consistent data definitions across the organisation. Agentic AI, which plans and executes multi-step tasks on its own, only works properly when governance, data lineage and observability are embedded across the entire AI operations process. If these elements are built only as separate components, limits will inevitably arise when scaling AI across the organisation.

In the end, the core is the platform.

An AI strategy eventually arrives at a single question: does the current data platform have the foundation to run AI across the enterprise? Traditional data stacks were designed for reporting and batch analytics, leaving storage, governance, modelling and deployment tools fragmented. This disconnection complicates governance and makes audits and regulatory compliance harder. It also creates inefficiency by forcing repeated work.

Companies generating tangible results, by contrast, are adopting modern data and AI platforms that integrate data, analytics and AI into a single environment.

On a single Lakehouse foundation, they integrate data storage, computing, governance and AI workflows into one environment to minimise unnecessary data movement and duplicated work.

Through Unity Catalog, they provide consistent access control, data lineage and audit frameworks across data, AI models and applications. They support the full AI development cycle in a single environment, from data exploration and feature engineering to model deployment, monitoring and model drift detection.

They integrate ETL, streaming and AI model pipelines into repeatable and auditable workflows to improve operational efficiency and reliability. Using governed enterprise data, they implement AI agents and conversational AI to support autonomous workflows that go beyond simple question-and-answer to performing actual tasks.

This is not just theory. BC Card, a major South Korean payment company, unified its previously fragmented data environment into a single integrated platform based on Databricks. By bringing large-scale payment data analysis and an AI and machine-learning environment together and strengthening data governance, it sharply increased its ability to execute by applying AI to real business.

In other words, competitiveness does not come from one specific technology or a single feature. It comes from the integration and consistency of a platform that can run data, analytics and AI seamlessly as one organic environment.

Where will the financial industry gap widen in 2026?

By the end of 2026, the competitive landscape in finance will be reshaped not around whether a firm has adopted AI, but whether it has embedded AI in real business. Leading companies will widen the gap by building AI into core tasks such as risk judgement, pricing, customer service and anomaly transaction detection. Companies that cannot do so are likely to remain at the pilot stage and stay focused on discussing potential.

At first, the difference may not seem large. But as time passes, the gap will widen further, and competitiveness once established will be hard to catch up with. South Korea's financial sector is also moving faster to apply AI in real business as financial authorities actively encourage the use of generative AI and seek to establish risk management frameworks. With higher customer expectations and competitive pressure in the market, execution capability to move beyond pilots and into operating environments has emerged as an essential task for South Korean financial institutions.

Simply adopting AI earlier than others is no longer enough to secure a competitive edge. Execution capability that takes root in core tasks and decision-making processes is the most powerful weapon that will determine the success or failure of financial institutions in 2026.