Revolut Launches Dedicated AI Research Division
Revolut has launched Revolut Research, a dedicated AI research division focused on machine learning for financial services. The company announced the new division on August 25, 2026.
The unit sits within Revolut’s broader AI Department. Moreover, it will work with academic and technology institutions to develop proprietary AI systems and machine learning capabilities.
At the centre of the initiative is PRAGMA, Revolut’s proprietary foundation model developed with NVIDIA. The model targets financial applications including real-time risk assessment, fraud detection, platform operations and personalised product recommendations.
Revolut Builds Around Proprietary Financial AI
Revolut says Revolut Research will provide a central framework for its proprietary AI deployments. Therefore, the company intends to develop a shared intelligence layer rather than rely mainly on separate models for individual financial functions.
PRAGMA focuses specifically on financial behaviour and banking event data. In an April 2026 research paper, Revolut researchers described PRAGMA as a family of foundation models for multi-source banking event sequences. The research examined applications including credit scoring, fraud detection and customer value prediction.
Furthermore, Revolut says early deployments on historical data produced significant improvements against legacy baselines. The company reported 2.3 times higher accuracy in identifying credit default risk. It also reported 65% more fraud cases detected, with 17% greater precision in fraud alerts.
The company also reported a 41% improvement in the relevance of product recommendations across retail and business accounts. However, these figures come from Revolut’s own early testing and should therefore be treated as company-reported results rather than independent performance benchmarks.
Revolut’s strategy also builds on a large operational dataset. The company says it serves more than 80 million customers across more than 40 markets and processes billions of cross-border transactions. Consequently, Revolut sees transaction and behavioural data as an important source for improving its models.
PRAGMA Anchors Revolut’s AI Strategy
Revolut Research will also support the company’s wider AI infrastructure. In particular, Revolut has been developing internal systems for machine learning and large language models alongside its customer-facing AI tools.
The company’s 2025 annual report said its AI platform supports the lifecycle of machine learning and LLM-based models. It includes controlled releases, versioning, validation checks, data monitoring and evaluation systems.
Meanwhile, Revolut has continued developing AI capabilities for customer service and financial workflows. Its annual report said its LLM-powered Assistant resolved more than 75% of customer queries in 2025.
Beyond PRAGMA, Revolut says its advanced security models review nearly one billion transactions each month. These systems aim to detect fraudulent activity before it affects customers. Furthermore, the company’s AIR assistant can execute complex financial tasks through a single step, although the feature is currently available only in the UK.
The creation of a dedicated research division also places Revolut within a broader shift across financial technology. Banks and fintech companies are increasingly exploring proprietary AI systems for fraud prevention, lending, customer service and operational automation. Consequently, control over data, models and deployment infrastructure is becoming a strategic consideration.
Research Expansion Targets Global AI Leadership
Revolut Research plans to publish scientific findings and open-source technical frameworks regularly. In addition, the team will participate in international research conferences and engage with the wider scientific community.
The company plans to participate in NVIDIA GTC Berlin in October and ICAIF in November. It will also host quarterly scientific meet-ups at Revolut offices. Therefore, the division intends to position itself within the broader financial AI research community.
The move also extends work that is already visible in academic research. PRAGMA was publicly described in an April 2026 paper, which presented the model as a general-purpose representation layer for financial applications.
For Revolut, the next stage will involve moving from research benchmarks to production-scale financial applications. That transition will require continued attention to model reliability, data governance, security and regulatory controls.
Ultimately, Revolut Research signals a deeper investment in proprietary AI infrastructure. Rather than treating AI as an additional interface, the company is positioning machine learning closer to the core of its financial technology stack.
