Аннотация:The paper addresses the issue of overcoming the limitations of classical methods of improving the quality of education, with regard to the forecasting of long-term future social needs. Firstly, these limitations arise due to the unreliability of forecasting needs, since the market can fluctuate in an unpredictable way. Consequently, teaching from kindergarten onwards is carried out in an environment of very high uncertainty, and is mainly based on retrospective knowledge and methods. Long-term needs forecasting does not yet allow adequate educational trajectories to be developed, and the use of a multilevel educational management system means that this development is unstable. This paper demonstrates that digital education and strategic planning can provide fundamentally new opportunities for improving the quality of future education. This requires the use of non-classical and post-non-classical technologies for predicting educational services needs and synthesising long-term educational trajectories. This synthesis is a form of inverse problem solving, that is, it arises from an imprecise image of the future. The author's convergent approach is used here, which combines methods of inverse problem solving in topological space, cognitive modelling, genetic algorithms, network expertise, controlled thermodynamics and quantum semantics. Practical testing of the proposed approach has shown its fruitfulness.