As banks struggle to transform AI from individual tools to business intelligence, Revolut’s PRAGMA model offers a glimpse of where digital banking and financial services may be headed next.
Adobe Stock
Revolut is taking a very different approach to AI in banking.
While many financial services firms use off-the-shelf AI tools or build separate models for fraud, credit risk and customer service, the UK-based digital bank is developing its own foundation model designed specifically for finance. The idea is not just to apply a linguistic model to financial data, but to teach AI to understand financial behavior itself.
Its proprietary model, PRAGMA, treats transactions, customer interactions, app usage, investments, bill payments and support requests as signals in a connected system. Instead of looking at each banking function in isolation, it learns patterns across the customer journey, creating a single AI foundation that can support fraud detection, credit decisions, product recommendations and customer service.
Financial services are fast becoming one of the most important reasons for business AI. The industry has massive amounts of structured data, repetitive processes, complex risk decisions and heavy compliance requirements, which make it fertile ground for intelligent automation.
Revolut’s approach offers a glimpse of where business AI strategies are headed. For businesses, the next stage of AI is to move beyond individual workflow improvements and build the foundation for building intelligence across the entire organization.
A unified AI for banking and financial services
As financial services institutions began working with AI, the common practice was to develop separate tools, models, and processes for the required functions, each requiring its own engineering, training data, and maintenance processes.
Revolut has replaced this stack with a common base model trained on 40 billion events and interactions with 25 million users.
This means it can learn patterns of behavior across the entire customer journey: app and website usage, transactions, transactions and investment activities, and paying bills and subscriptions.
The exhibitions impressive results. PRAGMA has improved by 64.7 percent in detecting and stopping fraud, a 16 percent increase in credit risk prediction performance, and is 41 percent more effective at recommending relevant products.
at technical sidethe stack is powered by 200 NVIDIA H100 GPUs, which allowed it to scale from 38 million users in 2023 to more than 70 million today.
What sets it apart from competitors that use artificial intelligence for similar tasks is that all of these improvements come from a single algorithmic and data pipeline. Lessons from a single task allow it to improve performance across the board as models learn in parallel, adapting to changing economic and behavioral trends.
PRAGMA is the foundation of an organization-wide AI strategy that allows the challenger bank to react to change with greater agility than larger incumbents.
This includes the AI customer service assistantwhich currently handles 75 percent of support requests without human intervention.
There is also his first customer facing artificial intelligence system, AIR (AI by Revolut; their acronyms are getting better), recently rolled out to UK customers. In addition to performing routine actions such as managing subscriptions or canceling lost cards, it is designed to be a “lifestyle companion”, helping with budgeting and even organizing trips.
So what lessons can business leaders and practitioners take from this and apply to AI strategy in financial services and beyond?
What can we learn from Revolut’s AI strategy?
Many organizations still run fragmented AI stacks, with different models and data pipelines for specific tasks or functions.
This can quickly become a liability, leading to high maintenance costs and a lack of transparency and best practices within the company. Just as crucial, it prevents learning and improvement in one function from achieving better performance across the business.
The Revolut alternative is a unified AI architecture based on a common model. This makes it capable of combining the intelligence and benefits of automated decision-making across the enterprise, rather than simply driving efficiency in individual tasks.
With billions of data points generated every day, the volume and velocity of data in Revolut’s hands means it is strategically positioned to deliver cutting-edge services: personalized insights, personalized customer support and agent tools capable of taking autonomous action on behalf of customers.
In banking, financial services and beyond, the ability to better understand and respond to customer needs is fast emerging as a huge competitive differentiator.
Taken together, these changes mean that businesses that invest in unified AI and data infrastructure are better able to solve problems and pain points, leading to an improved customer experience and driving long-term growth.
