Sber Creates AI Model to Forecast Customer Behavior
Sber researchers have developed a foundational AI model that forecasts the behavior of bank customers.
It combines financial transactions, in-app actions, inquiries, and communications into a single timeline and identifies patterns.
The multimodal approach improves forecast accuracy and makes it possible to use one model across different scenarios instead of separate solutions. The technology was tested in an A/B test across six products and three communication channels. It delivered around a 1% increase in NPV and has been put into commercial operation.
According to him, the new model makes it possible to rely on this context when working with different products and to choose more precisely who should receive a message. That way, a customer is more likely to learn about a product that will be useful to them.
