The article considers scenarios of application of neural networks of different architectures to fulfill tasks in the foresight process. The purpose of the article is to determine at what stages of the foresight procedure the application of neural networks is justified and with what architecture. The differences between foresight and the process of forecasting are revealed. In addition, the concept of foresight, its main stages and phases, classification are considered. It is substantiated that the application of neural networks can significantly facilitate the foresight procedure at such stages as collection and processing of primary information, development of scenarios and solutions to problems, communication, and report preparation. It is shown that different types of neural networks are suitable for different foresight tasks. It is revealed that neural networks can process a larger amount of data and automatically detect complex patterns, which makes them more effective under conditions of environmental uncertainty and variability. The article ephasises the importance of further research and development of methods for applying neural networks in foresight processes with consideration to the specifics of particular industries and types of tasks. In the course of the research, the authors used analytical methods of diagnostics, establishing cause-and-effect relationships, etc.