The problem of predicting crystal structures is discussed in the context of artificial intelligence systems. The steps of creation of an expert system are considered as applied to crystal design, where the crucial step is invention of new structure descriptors. A number of such descriptors proposed quite recently are listed; most of them characterize local coordination or overall topology of the structure network. An important part of the expert system is the knowledge database that contains correlations between the descriptors; it is used by a computer analyzer, the inference machine, to make a conclusion about possibility of obtaining a particular structure motif. All the steps of developing the expert system are illustrated with the analysis of 811 cyanide complexes and examples of the structure prediction are given.