I got the impression that insurance companies build a large Bayesian network and do parameter estimation using their tons of data to estimate, given a person wi

I got the impression that insurance companies build a large Bayesian network and do parameter estimation using their tons of data to estimate, given a person with a specific profile, what his expected costs are (plus calibrated "error rate"). Perhaps I was too naive?