Artificial Intelligence Supports Customers and Employees of PKO Bank Polski

The most well-known AI solutions at PKO Bank Polski are voice bots, including the voice assistant in the IKO app, which helps customers use banking services. Customers can speak to it or communicate in natural language, just as they would with a human. It can, for example, facilitate fund transfers, provide BLIK codes, adjust card limits, or block cards. When a customer inquires about their expenses, the assistant helps analyze them through clear charts, and if they wish to exchange currency, it directs them to a currency exchange. Currently, it understands over 320 conversation topics and has already been used by over a million bank customers.

In total, PKO Bank Polski has 16 bots that perform tasks in various areas. They understand human speech, make calls to customers, or support them on the bank's helpline. Depending on the topic of the conversation, they help resolve issues or redirect to the appropriate consultant. They also assess the level of customer satisfaction with service or their inclination toward a sales conversation, enabling bank employees to contact only those who are genuinely interested in a particular offer. They can also verify the customer's identity, protecting them from potential fraud. If a bot detects a suspicious transaction, it calls the customer to verify whether they indeed initiated such an instruction.

The bank continually seeks new AI applications to enhance its services. It has solutions that can read and classify customer letters and emails, or automatically categorize thousands of customer opinions about the IKO app and bank branches. These are just some of the applications of AI at PKO Bank Polski, and the models used for these tasks are constantly being developed and learning new activities.

Machine learning has long been used by the bank in risk assessment processes. Advanced machine learning models for risk assessment, forecasting various phenomena, and customer behavior are created based on vast data from various sources. Currently, the bank is developing a Data Advanced Analytics Platform (MLOps, Machine Learning Operations), which allows analysts to build advanced and appropriately optimized models using Google cloud solutions.

PKO Bank Polski's MLOps platform is currently the most advanced tool of its kind in the Polish banking sector. This cloud-based solution streamlines the deployment and management of a range of machine learning models in the production environment while ensuring their quality. It serves as a model factory that enables bank analysts to select and apply the most efficient model for a given case. The bank aims to have up to 90% of risk assessment processes automated through artificial intelligence eventually.

Machine learning models in PKO Bank Polski also play a crucial role in supporting sales processes. Advanced predictive models based on machine learning algorithms are now an essential part of supporting the sales of banking products and services in the context of CRM activities. They allow for the preparation of product recommendations or personalized offers for customers, minimizing the risk of misselling—proposing financial services to customers that do not meet their needs. Thanks to the accumulated data, the bank optimizes the method, timing, and contact channel depending on the customer's level of digitization while maintaining omnichannel principles. There are many areas in which machine learning/AI can support current activities in the bank. We strongly focus on the Advanced Data Analytics Platform launched a few months ago, also known as ML Ops, which is the foundation for the implementation of new initiatives. This is an area that will undoubtedly continue to grow dynamically within the bank," says Jacek Kuban, Director of the Big Data Department at PKO Bank Polski.

***