The manufacturing industry are increasingly turning to artificial intelligence (AI), particularly machine learning (ML), to improve their operations and decision-making. Predictive modelling and ML hold significant promise in revolutionising forecasting and assessment activities by offering improved decision support for quicker and more reliable decisions. However, the preliminary results of the current multidisciplinary and interactive research project PredMod (short for predictive modelling) empirically shows that there are still many challenges to overcome when implementing predictive modelling and ML to produce forecasts and other assessments, such as capacity needs, in the manufacturing industry. The purpose of this paper is to enhance managerial comprehension and awareness of challenges associated with using predictive modelling and machine learning to produce forecasts and assessments within the manufacturing industry. Based on the empirically identified challenges, the paper also provides some practical lessons learned and articulate key questions that should guide predictive modelling projects within the manufacturing industry.