This study examines the association between Artificial Intelligence (AI) adoption and the financial performance of Indian commercial banks over a ten-year period (2014–2023), using a panel of 20 banks (10 classified as AI-adopting, 10 as minimally AI-adopting). Because AI-adoption status does not change within any bank over the sample period, a standard fixed-effects estimator cannot identify its coefficient —the treatment is time-invariant and gets absorbed into the bank-specific intercept. The analysis instead uses pooled OLS and random-effects models with standard errors clustered by bank, cross-checked against Welch's t-tests. AI-adopting banks show significantly higher Return on Assets (+0.50 percentage points) and Return on Equity (+5.63 percentage points) and a significantly lower Cost-to-Income Ratio (−11.32 percentage points) than non-adopting banks (all p < 0.001). A bootstrapped mediation test does not find a statistically significant indirect effect of AI adoption on profitability through cost efficiency, indicating that the profitability association operates at least partly through channels this dataset does not capture. The design cannot rule out reverse causality — that financially stronger or better-managed banks are simply more likely to adopt AI — and this is discussed as the study's central limitation rather than resolved by the availabledata. The contribution is a corrected identification strategy for a time-invariant adoption variable in bank