In laser beam welding (LBW) complex physical phenomena occur during laser-material interaction, which prolong the parameterisation of the process due to the extended trial and error tests. Therefore, predicting the process behaviour leads to a more productive and cost-reduced process experimental tuning. Besides, as LBW is a thermal process, part distortion plays a relevant role in the final result, and hence, thermal and mechanical analysis are both required. In view of this, in the present research, a novel multiscale numerical model, capable of predicting the thermomechanical behaviour of the LBW process has been developed. The significance of the model is its capability of forecasting the part distortion and thermal field employing two fully coupled modules, where the laser heat source automatically adapts to the welding regime without the need to consider the melt pool dynamics and at a low computational cost. The local model determines the melt pool dimensions and thermal field. Besides, the second module, the global model, figures out the part distortion based on the thermal results. Finally, the presented numerical simulations are experimentally validated with the corresponding temperature monitoring during the welding process, the posterior metallography inspection, and the part deformation measurement. The results present a high accuracy, with a maximum error below 10 % at the temperature measurements, an average dimensional deviation of 0.14 mm and 0.18 mm respectively for the weld bead depth and width, and a vertical deformation average error of 0.15 mm.