This research aims to provide an insight about the applicability of Artificial Intelligence (AI) in making a forecast about decision-making duration of a new infrastructure project by the government during infrastructure planning procedure prior to the market introduction. The decision-making duration in this research is the duration which the government requires to decide on the preferred alternative for a certain infrastructure problem. By identifying the likely timeframe of a new infrastructure project introduction from the government, the construction firms can prepare itself better by allocating its resources accordingly to the profile and requirement of the upcoming projects. This research decided to explore further about the applicability of AI to forecast the duration within this context based on the potentials of AI that shown by previous researches. With regard to the scope of the research, the Netherlands construction industry is chosen with the perspective of a construction firm, which is BAM Infra BV. With regard to the legal framework of the industry, Dutch Infrastructure Planning Act (Tracewet) is the main reference of the research, which acts as the foundation of the proposed forecasting model. Lastly, the research focuses on the road infrastructure project from the Netherlands government, which falls under Tracewet. This research applies a combination of three methods to fulfill the aforementioned objective, which are literature study, experts interview, and model simulation. This research is divided into four main steps to answer the main research question: [1] AI Theory and System Design for Forecasting, [2] Data and Variable Exploratory Study, [3] AI Forecasting Model Implementation, and [4] Result Discussion. The result shows that the AI technology, specifically ANN, is not applicable to be used as a forecasting model to predict a new infrastructure project introduction based on the duration of decision-making on the preferred design alternative by the Dutch Government. Two approaches have been explored in this research, namely the ANN regression model and ANN classification model. It is found that neither models produce a reliable prediction, which indicated by high RMSE and low model accuracy on the dataset. This reliability is evaluated by interviewing BAM’s commercial manager about the acceptable range of error for the prediction made. The optimization effort has been done to address these results by iterating several different variables combinations into the models. These optimization results revealed that some factors influenced the performance of the models. A further discussion has been done on these factors; namely number of data entries, input variables influence, and representation of the world by the model. It is found out that the combination of these factors has an impact to a certain degree on the model performance. Besides that, a comparison with other conventional methods (i.e., Multiple Linear Regression and Logistic Regression) has been done. The result shows that the ANN models do not perform better than the conventional methods being compared in term of model prediction accuracy, which transcends into its ability to handle imprecise data and non-linear approach. This comparison result indicates the importance of dataset quality over the decision of a forecasting model to be used.