PRAGMA is trained on a vast dataset of 40 billion events from 25 million users, enabling a unified AI pipeline for fraud detection, credit risk assessment, product recommendations, and customer service.

Leaders should note that moving to a unified AI and data infrastructure can elevate customer experience, enable autonomous actions, and drive long-term growth across financial services and beyond.

The system analyzes patterns across transactions, app usage, investments, bill payments, and support requests to enable cross-task improvements and quicker adaptation to evolving economic and behavioral trends.

Revolut’s customer service AI now handles about three-quarters of support requests without human intervention, while AIR acts as a customer-facing lifestyle companion for budgeting and travel.

The technical setup runs on a unified stack powered by 200 NVIDIA H100 GPUs, supporting user growth from 38 million to over 70 million and sustaining a single algorithmic/data pipeline.

Revolut is building PRAGMA, a proprietary foundation model tailored for finance to understand financial behavior across the entire customer journey, not just isolated tasks.

A unified AI architecture offers advantages over fragmented stacks, including cross-functional learning, reduced maintenance, greater transparency, and compounded intelligence across the organization.

Reported results show a 64.7% improvement in fraud detection, a 16% improvement in credit risk prediction, and a 41% increase in the relevance of product recommendations.