In today's IT environment, project implementation is characterized by a high level of complexity, dynamic changes, multifactorial nature, and significant uncertainty. Management decisions made in such conditions require rapid analysis of a large amount of heterogeneous information, as well as the ability to predict the future state of the project under existing resource and time constraints. Traditional project management methods are often insufficient to respond effectively to such challenges, as they are based on linear or rigidly formalized models. In this regard, the implementation of intelligent decision support systems that can adaptively respond to changes and take into account the complex relationships between project parameters is becoming increasingly important. The article investigates the possibilities of using Artificial Neural Networks, in particular the Long Short-Term Memory (LSTM) architecture, to improve the efficiency of management decisions in IT projects with a high degree of complexity. The theoretical foundations of building decision support systems based on deep learning are analyzed, and the feasibility of using the LSTM model for forecasting tasks in project management is substantiated. The architecture of such a system is presented, including modules for data collection, pre-processing, and analysis, an LSTM-based forecasting kernel, and a user interface. As part of an experimental study, an IT project was modeled using synthetic data that simulates the course of several sprints in the SCRUM methodology. The obtained results confirmed the model's ability to predict key indicators (execution duration, delay probability, resource utilization level) with high accuracy, which, in turn, allows project managers to make timely corrective decisions. Numerical metrics for assessing the quality of the model (Mean Absolute Error (MAE), Mean Squared Error (MSE)), as well as examples of forecast visualizations are presented. The advantages and limitations of using LSTM in project management are discussed, including the problems of interpretability, generalizability, and integration with existing corporate systems. Thus, neural networks can become an effective tool for supporting management decisions, providing adaptability, accuracy, and timeliness in predicting project parameters. The results of the work can be useful for project managers, developers of Decision Support System (DSS) systems, and researchers in the field of applied artificial intelligence.