Kinetic modeling is a challenging aspect of biomass conversion due to its inherent complex reactions. Further, it is difficult to achieve the accuracy of predicting the biomass pyrolysis process at varying experimental conditions, particularly for more complex samples based on kinetic modeling alone. Therefore, this study aims to use artificial intelligence (AI) models [artificial neural network (ANN), support vector machine (SVM), and decision tree (DT)] to predict biomass thermogravimetric (TG) behaviour, derivative TG (DTG), product (volatile and char) yield, and kinetic triplets. Two oil palm biomass types-empty fruit bunches (EFB) and oil palm shells (OPS) – are used to generate a 72-experiment dataset from a thermogravimetric analyser (TGA) at different heating rates (5, 10, 15 and 20°C min^-1). The results reveal that each AI model can accurately predict the DTG profile with high coefficients of determination (R2) in the range of 0.94 and 0.99 and low mean square errors (MSE) between 0.09 and 5.12. The product yield prediction results are not as promising, as indicated by higher MSE values (4.27, 2.79, 6.67). However, the ANN models most capably predicted the activation energies of oil palm biomass pyrolysis (~260°C – 360°C at 20°C min^-1) for both model-free and model-fitting methods, followed by the DT and SVM models.