Bayesian networks are probabilistic models that can be used for applications like diagnosis, trouble-shooting, prediction under uncertainty, user-modeling customer-modeling, and so on. We need suitable Bayesian network models and probabilistic reasoning methods for practical applications. Our software (BayoNet) can construct Bayesian networks from statistical data in SQL DBs or CSV files. After constructing the model, BayoNet executes probabilistic reasoning via a fast approximate algorithm for practical application systems. BayoNet has been licensed to companies. We apply Bayesian network modeling to marketing, constructing user models that can predict and apply user intentions for interactive systems.