OTP Bank attributed a 2% increase in net profits last year to its use of 36 machine learning models in risk management, president Ilya Chizhevsky revealed at the 2026 International Financial Congress. Rather than pausing its AI rollout, the bank is carefully selecting scenarios where models deliver consistent and verifiable outcomes.
The biggest benefits, according to Chizhevsky, come from retail-focused models, which show a 5-20% improvement in Gini coefficients compared to previous methods. In banking credit scoring, the Gini coefficient measures how effectively a model distinguishes reliable borrowers from risky ones-the higher, the better the separation and the lower the cost of errors.
However, Chizhevsky noted that performance varies across use cases. AI models analyzing corporate borrowers using public data remain unstable and require frequent recalibration. Here, AI mainly helps gather and condense information, while final lending decisions stay with human analysts. Similarly, in atypical processes with few repeatable actions, training models is harder to scale.
This focus on measuring AI’s return on investment aligns with wider industry trends. Russia’s top banks have shifted over the past two years from experimental AI projects toward practical applications in credit scoring, fraud detection, and customer service. The Bank of Russia highlights credit institutions as among the country’s most digitized sectors, while the FinTech Association ranks AI alongside biometrics and open APIs as core technologies for banking platforms.
How OTP Bank uses machine learning in risk management
OTP Bank’s deployment of 36 machine learning models focuses primarily on retail credit scoring, where improvements in accuracy have led directly to profit gains. The bank carefully tests and implements models only in scenarios where they deliver reliable and measurable benefits. For corporate lending, AI serves as a tool for information gathering rather than decision-making, with human analysts maintaining control over final loan approvals.
AI impact on Russian banking sector’s credit risk and customer service
Russia’s leading banks, including OTP, Sberbank, T-Bank, and Alfa-Bank, are rapidly adopting AI-powered solutions for credit scoring, fraud detection, and personalized customer service. The integration of AI assistants and recommendation engines aims to boost efficiency and enhance customer experiences by anticipating needs and automating next steps without increasing operational or regulatory risks.
Challenges and outlook for AI in banking
Applying AI across diverse lending segments reveals challenges such as unstable performance in corporate borrower analysis and difficulties in scaling models for atypical processes. OTP Bank’s experience underscores the importance of balancing automated decision support with human oversight to ensure regulatory compliance and customer satisfaction. Going forward, banks that deploy stable, scalable AI models will gain competitive advantage by aligning product offers more closely with actual customer needs.