The rise of extensive monitoring systems and the availability of low-cost sensors as well as affordable computing power has led to the development of various big data and simulation model applications in the building sector. Nevertheless, many of these promising approaches face a common hindrance for the widespread application. In case of the big data applications, training data is often limited. Much the same, simulation models often lack required input data and require extensive work for calibration. Standard practices are often preferred to innovative approaches because construction and commissioning businesses are highly cost-sensitive. Therefore, we identified the need for a holistic approach for the combined use of machine learning and simulation techniques. In this paper, we present a toolchain to generate models and data needed for the application of innovative building automation and control tools. Using the data available during the construction process, machine-learning algorithms are employed to determine the type and location of data points in devices. From the relations of data points, the system architecture is derived and simulation models are generated automatically. Using these models, the data needed for the training of big data machine-learning algorithms can be generated. We describe the toolchain, already existing components and discuss the possible future implementation.