Maintaining oil production at long-term developed fields requires solving the problem of high production costs. This problem is associated with the need to withdraw significant volumes of produced water and a proportionally high need for injection in order to maintain reservoir pressure. It is noted that a 1% reduction in water cut in production makes it possible to reduce operating costs in oil production by up to 15%. It is shown that the problems of effective development of mature fields are associated with the solution of the optimization problem of distributing fluid production and water injection in the wells system. The authors argue the idea that at the later stages of development, the priority for hydrodynamic modeling should be tools based on solving the inverse problem of hydrodynamics, providing for the widespread use of material balance methods and allowing big data processing. A new concept of combining artificial intelligence methods and a hydrodynamic model is proposed. The concept provides for obtaining a functional relationship between the historical oil production rate and injectivity using a neural network, searching for the maximum oil production and its distribution. At the same time, only one calculation is performed on the hydrodynamic model, which significantly reduces time costs. An example of the application of the proposed technology is given. It is concluded that the set of methodological, mathematical and informational solutions presented in the article will allow formalizing the processes of designing hydrodynamic methods for enhanced oil recovery, clarifying the model for a feasibility study of profitable and technologically recoverable oil reserves.