The largest Polish bank – PKO Bank Polski – uses artificial intelligence to assess the risk of most consumer loans and the SME sector, as well as for 30 percent mortgage loans. AI also supports precise analysis of corporate loans. Currently, over 80 percent of credit decisions are made with the support of advanced machine learning algorithms.
Thanks to modern technologies, the bank can assess the customer’s ability to repay the loan more precisely and offer him a better-tailored product. Also, machine learning is used to monitor loan portfolios and timely repayment of liabilities.
– Effective risk management is a key element in maintaining bank stability. Thanks to modern technologies, we have implemented a cloud-based risk assessment tool, which allowed us to increase data processing power. Currently, over 80 percent of credit decisions are made using machine learning algorithms. Based on a huge amount of information from various sources, we can create increasingly advanced models for risk assessment and forecasting of various phenomena and customer behavior. As a result, we grant more loans with the lowest risk costs in the sector, – says Marek Łach, Director of the Retail Customer Credit Risk Division at PKO Bank Polski.
Thanks to the use of artificial intelligence, the bank can create more accurate risk assessment models that take into account hundreds of variables, including new data from external sources or available only to a specific group of customers. Credit assessment also uses information about how the customer behaves in mobile and online banking. Machine Learning, thanks to its accuracy, increases the acceptance level of loan applications by 2 percentage points. PKO Bank Polski also uses this technology for precise analysis of corporate loans. Decisions regarding the 8.6 billion portfolio of these loans are made using precise analysis of purchase transactions.
Faster and customer-friendly credit decision
AI in risk assessment means not only a more precise analysis of loan applications but also a faster and simpler loan granting process. At PKO Bank Polski, in the case of the best customers with a good repayment history, the decision to grant further financing is made automatically and practically without any formalities. Artificial intelligence also supports the processes of monitoring the timeliness of repayments and helps select the appropriate form of contact with a customer who has problems with settling their liabilities.
ML Ops platform – model factory
In 2022, PKO Bank Polski implemented the ML Ops Advanced Data Analytics Platform. It is a solution based on cloud technology that allows analysts to build advanced risk assessment models. It streamlines their implementation and ensures appropriate quality. It constitutes the so-called model factory that allows bank analysts to select and apply the most effective model for a given case.
– Our ML Ops platform is the most innovative solution of this type in the Polish banking sector. It works like a modern production line and allows you to create models on a mass scale. This allows Data Science teams to build and implement them efficiently, with minimal IT involvement. The use of the platform has significantly increased the number of models used in customer credit analysis. We are constantly developing this tool and working to increase the available range of data. The results related to improving the quality of models trained using graphics cards are also very promising. We also want to introduce further significant improvements in this area, – says Bogumił Brela, Director of the Department of Development of Data Engineering and Analytics Platforms at PKO Bank Polski.
PKO Bank Polski aims to ensure that, with the support of AI, as many as 90 percent of risk assessment processes are carried out automatically.
Source: Bank’s Press Release from July 01, 2024
