Profitability of AI adoption in the Banking Sector

Sammanfattning

This thesis examines the relationship between generative AI adoption and firm profitability in the Nordic banking sector, motivated by the growing societal and commercial impact of generative AI, despite limited empirical evidence regarding its measurable financial benefits. To assess and compare the level of AI adoption across firms, a proxy was employed to capture each firm’s level of AI intensity (the extent to which firms utilize AI ) for each year in the study’s time frame (from 2022 to 2025), consisting of a textual analysis of each firm and year’s annual report, counting the proportion of a select number of AI-related words. This was found to be the most reliable way of measuring AI intensity, as there are no common or comparable financial metrics to represent it. A sample of the six largest (by total assets) Nordic banks were then compared over the period 2022 to 2025 with a fixed-effects panel regression (6 banks × 4 years = 24 observations), using control variables such as company size and leverage, with return on equity (ROE) as the dependent variable. One contemporaneous model was used, and one with a one-year lag effect. In the contemporaneous model, AI_intensity showed a positive but statistically insignificant coefficient (β = 0.143, p = 0.267), presenting no noticeable same-year relationship between AI_Intensity and ROE. In the lagged model, however, the coefficient became negative and marginally significant (β = -0.109, p = 0.095) at the 10% level. Overall, the findings provide limited evidence that generative AI adoption improves short-term profitability in Nordic banks. Instead, the results suggest that AI-related investments or initiatives may involve short-term costs that outweigh immediate financial returns.