This paper explores the application of artificial intelligence (AI) in forecasting Saudi Arabia’s non-oil export trajectories, aligning with the Kingdom’s Vision 2030 objectives for economic diversification. A range of machine learning models, including LSTM, Transformers, Ensemble Stacking, Random Forest, and XGBRegressor, were employed to analyse historical export and GDP data. Among these, the Advanced Transformer model demonstrated superior predictive accuracy, achieving a MAPE of 0.73%, underscoring its ability to capture complex temporal dependencies. The non-linear Blending Ensemble, which integrates Random Forest, XGBRegressor, and AdaBoost, exhibited robust predictive performance (MAPE 1.23%), leveraging the diverse strengths of its base regressors. Additionally, the Temporal Fusion Transformer (TFT), incorporating GDP data, provided insights into macroeconomic influences, though with a higher MAPE of 5.48%, indicating the challenges associated with integrating economic indicators into forecasting models. An analysis of feature importance, utilizing SHAP and Partial Dependence Plots, revealed that recent export values (lag1, lag2, lag3, and lag10) were the most influential predictors. These findings reinforce the transformative role of AI-driven forecasting methodologies, offering data-driven insights to support decision makers and economic strategists in formulating policies that promote sustainable economic growth.