The aim of this editorial is to increase the acceptance of neural net modeling in the actuarial community. Neural nets may substantially improve classical actuarial models, if appropriately applied. We illustrate this on a toy example but, in fact, this should be understood as a universal concept. Assume we have a classical regression problem where the distribution of a response Y = Y(x) can be described by covariates x. A common actuarial problem is to determine the premium $\mu ({\textit{\textbf{x}}}) = {\mathbb E} [Y({\textit{\textbf{x}}})]$ as a function of the covariates x. Actuaries have developed excellent skills to solve such problems through finding appropriate regression functions x ↦ μ(x). This editorial shows how these skills can further be improved using the toolbox of neural nets.