Artificial neural networks are currently widely used in economics and finance. First of all, they are an alternative to econometric estimation methods aimed at building forecasts, and solve a wide range of applications related to the preprocessing of data for research. It may be useful for a modern economist to have this tool in the arsenal of methods. The purpose of the proposed work is a popular explanation of the principle of operation of this tool, a demonstration of some areas of its application, features of development and conditions under which the use of this type of machine learning can be of practical benefit.