Simulations show AI adoption mechanically reduces firms' CAPM betas, but realistic sample noise typically masks the effect; AI-based characteristic tilts change allocations and can deliver competitive returns across regimes, yet they do not reliably outperform minimum-variance strategies.

Purpose This study aims to examine whether organizational artificial intelligence (AI) adoption is associated with firms' exposure to systematic market risk, as captured by CAPM beta, and whether AI-oriented portfolio allocations exhibit economically meaningful characteristics. Rather than introducing AI as a new systematic risk factor, the study conceptualizes AI adoption as a firm-level organizational characteristic that influences exposure to the existing market factor. A further objective is to evaluate the statistical detectability of this relationship under realistic finite-sample conditions using a controlled Monte Carlo simulation framework. Design/methodology/approach A simulation-based research design is employed using a CAPM-consistent data-generating process for 50 firms observed over 60 months. Firm-level AI adoption scores influence market beta within a controlled Monte Carlo framework comprising 1,000 replications. Systematic risk is estimated using CAPM regressions, followed by cross-sectional OLS and weighted least squares (WLS) estimation. Sensitivity analyses examine alternative beta-estimation windows, while portfolio analysis compares characteristic-based AI-tilted portfolios with minimum-variance portfolios using both in-sample and rolling out-of-sample benchmarking across baseline, bear and bull market regimes. Findings The simulated data consistently produce a negative association between AI adoption and market beta, although statistical detectability remains limited under realistic noise conditions. WLS modestly improves estimator performance relative to OLS but does not eliminate finite-sample limitations. Longer beta-estimation windows reduce coefficient dispersion while producing only limited gains in statistical power. Rolling out-of-sample portfolio analysis shows that AI-tilted portfolios generate economically meaningful allocation differences and competitive performance across alternative market regimes without consistently dominating the minimum-variance benchmark. Overall, the results highlight the distinction between structural effects and their empirical detectability. Research limitations/implications The study relies on simulated data and a single-factor CAPM framework in which AI adoption is treated as a time-invariant firm characteristic. Consequently, the results should be interpreted as methodological rather than empirical evidence. The framework isolates one theoretical mechanism under controlled conditions and is not intended to establish AI as an independent priced risk factor. Future research may extend the approach to multifactor asset-pricing models, time-varying AI adoption, alternative channels through which AI affects financial risk, and empirical validation using observed firm-level data. Practical implications The proposed framework provides researchers and quantitative analysts with a methodology for evaluating whether AI-related effects on systematic risk can be reliably identified under finite-sample conditions. For portfolio managers, the findings suggest that AI-based characteristic tilts may produce distinct allocation structures and competitive out-of-sample risk-adjusted performance across different market environments, while not guaranteeing systematic outperformance over optimized portfolios. More broadly, the framework demonstrates the importance of accounting for estimation uncertainty, measurement error and statistical power when interpreting AI-related evidence in financial applications. Social implications As organizations increasingly invest in AI-driven digital transformation, understanding its relationship with financial risk becomes important for investors, managers and policymakers. The study highlights that statistically weak empirical evidence should not necessarily be interpreted as evidence of absent economic effects, particularly when measurement error and finite-sample limitations are substantial. By providing a transparent methodological framework for evaluating AI-related financial relationships, the research contributes to more informed interpretation of emerging technologies and supports evidence-based decision making regarding digital transformation and risk management. Originality/value This study contributes by integrating organizational AI adoption into a simulation-based asset-pricing framework while explicitly distinguishing firm characteristics from systematic risk factors. Rather than proposing a new pricing factor, it evaluates whether AI-related differences in market beta can be empirically recovered under realistic estimation noise. The study further combines Monte Carlo analysis, estimator comparison (OLS versus WLS), statistical power assessment and rolling out-of-sample portfolio benchmarking within a unified methodological framework. This provides a structured platform for investigating the empirical detectability of AI-related financial effects under controlled conditions.