Human health risk assessment has begun to depart from the traditional methods by replacement of the default assessment factors by more reasonable, data-driven, so-called chemical-specific adjustment factors (CSAFs). This study illustrates a scheme for deriving CSAFs in the general and occupationally exposed populations by quantifying the intraspecies toxicokinetic variability in surrogate dose using probabilistic methods. Acetone was used as a model substance. The CSAFs were derived by Monte Carlo simulation, combining a physiologically based pharmacokinetic model for acetone, probability distributions of the model parameters from a Bayesian analysis of male volunteer experimental data, and published distributions of physiological and anatomical parameters for females and children. The simulations covered how factors such as age, gender, endogenous acetone production, and fluctuations in workplace air concentration and workload influence peak and average acetone levels in blood, used as surrogate doses. According to the simulations, CSAFs of 2.1, 2.9, and 3.8 are sufficient to cover the differences in surrogate dose at the upper 90th, 95th, and 97.5th percentile, respectively, of the general population. However, higher factors were needed to cover the same percentiles of children. The corresponding CSAFs for the occupationally exposed population were 1.6, 1.8, and 1.9. The methodology presented herein allows for derivation of CSAFs not only for populations as a whole but also for subpopulations of interest. Moreover, various types of experimental data can readily be incorporated in the model.