In cooperation with a consumer credit company based in Stockholm, this bachelor thesis investigates if the customer profitability in the consumer credit market can be predicted with multiple linear regression. Data collected before the initial credit was accepted and data connected to the account activity of the customers' first nine months are analyzed. Further, it is examined if the findings could be useful in a profitability analysis and as a reduction of adverse selection. The findings show that a number of covariates express promising correlations with the costumer profitability. However, the prediction error is high and not efficient in individual cases. Further, some reduction in adverse selection, due to a decrease in asymmetric information between the customers and the company, can be identified, but further research is encouraged. Finally, potential improvements are discussed, especially concerning the choice of regression algorithm. / I samarbete med ett konsumentkreditbolag baserat i Stockholm undersöker detta kandidatexamensarbete om kundlönsamhet inom marknaden för konsumentkrediter kan förutsägas med hjälp av multipel linjär regression. Data består av information som insamlades innan den initiala kreditförfrågningen accepterades, och av kontoaktivitet under kundens nio första månader. Vidare undersöks om resultatet kan användas i en lönsamhetsanalys och som en metod för att minska snedvridet urval.  Resultatet visar att ett antal kovariat uttrycker en lovande korrelation med kundlönsamheten. Dessvärre är felen från förutsägelserna stora och därför ineffektiva gällande estimering av individuella kunder. Fortsättningsvis kan det identifieras viss reduktion av snedvridet urval som en följd av minskad informationsasymmetri mellan kunderna och företaget, men vidare undersökning uppmuntras. Avslutningsvis diskuteras ett antal förbättringsmöjligheter, framför allt gällande val av regressionsalgoritm.

Background: Managing credit risks is an integral part of the banking sector and is crucial for banks’ success. Effective risk management ensures stable and profitable operations, addressing challenges like information asymmetry between lenders and borrowers. To combat these challenges, banks are shifting from manual methods to automated processes in credit assessment and credit risk management.Purpose: The purpose of the study was to investigate how the use of AI has contributed to credit risk management and the handling of risk assessments within Swedish banks. Additionally, the study explored the factors driving the use of AI in this area.  Methodology: An abductive research approach was employed within the framework of a qualitative research method. Four banks were included in the study: two major banks and two niche banks. Semi structured interviews provided the primary data for the study, while secondary data, such as articles and literature, were used to support and explain the findings during the analysis and discussion.  Theory: The study was based on two models and the theory of information asymmetry. The first model focuses on the credit assessment process, while the second addresses critical success factors for the implementation of AI. The theory of information asymmetry consists of moral hazard and adverse selection. Conclusions: The study’s conclusion indicated that AI has contributed to increased efficiency and precision in credit risk management. Furthermore, AI supports addressing information asymmetry by automating data collection, analysis, and fraud detection. The study concludes that effective AI usage necessitates a balanced combination of management support, strategic vision, organizational culture, and structure.