Revolut’s PRAGMA Model Highlights Data’s Dominance Over AI in FinTech, Says Linas Beliūnas

In a recent LinkedIn post, Linas Beliūnas discusses the evolving landscape of Artificial Intelligence in FinTech, arguing that the true competitive advantage lies not in the AI models themselves, but in the vast and well-structured data behind them. Beliūnas highlights Revolut’s new foundation model, PRAGMA, as a prime example of this shift.

As Linas Beliūnas notes:

“Everyone talks about AI models. Revolut just dropped PRAGMA and showed the real moat is the data behind them.”

Beliūnas elaborates on PRAGMA, explaining that it was trained on an immense dataset of 40 billion banking events from 25 million users across 111 countries. This model differs from traditional approaches by focusing on user behavior over time rather than individual transactions or hand-built features. This allows for a single, adaptable model that can be applied across various functions.

The Power of Behavioral Data in FinTech

Linas Beliūnas points out the significant performance improvements achieved by Revolut through this data-centric approach. The post details substantial gains in key areas:

• +130% in credit scoring
• +65% in fraud detection
• +79% in engagement

However, Beliūnas also identifies a crucial trade-off. While PRAGMA excels at understanding user behavior, it shows a notable weakness in rule-based tasks, performing approximately 47% worse in Anti-Money Laundering (AML) efforts. This highlights the inherent challenge in building models that can master both nuanced behavior and rigid regulatory compliance.

“In other words, great at behavior, yet struggles with rules. And that’s the trade-off.”

This observation is particularly relevant given the strategic direction of major players in the financial industry. Beliūnas argues that companies like Stripe, Visa, and Mastercard are also moving towards a similar philosophy.

Data Scale as the New Moat

The core argument presented by Linas Beliūnas is that the emphasis is shifting from the sophistication of AI models to the sheer scale and quality of the data used to train them. He suggests that Revolut possesses a distinct advantage due to its Super App ecosystem, where nearly every user interaction generates valuable training data.

“Models matter less → Data scale matters more.”

In contrast, Beliūnas observes that many traditional banks suffer from siloed systems and fragmented data signals. This fragmentation makes it significantly more challenging for them to build comprehensive user profiles and develop competitive AI-driven solutions compared to platforms that can aggregate data seamlessly.

The Future of FinTech Leadership

Concluding his analysis, Linas Beliūnas posits that the future leaders in the FinTech sector will be those who can most effectively understand and leverage customer behavior through data. The ability to interpret and act upon behavioral patterns, rather than solely relying on transactional data or complex, task-specific models, will likely define success.

“If this holds, the winner in fintech will be whoever understands behavior best.”

Beliūnas’s insights underscore a critical pivot in FinTech strategy, emphasizing that a unified data strategy and a deep understanding of user behavior are becoming the paramount differentiators in an increasingly AI-driven financial world.