Building Information Modeling has become an industry standard for designing, documenting, and collaborating within the architecture, engineering, and construction industry. Architects spend most of their time developing projects from the feasibility study phase through post occupancy operation using BIM software such as Autodesk Revit. There are many tools created to automate tedious tasks or speed up the design process, however, most tasks require manual work, therefore taking longer to complete and increasing the probability of design errors. With each new release software developers are adding more features and tools to increase application capabilities; however, these are often underutilized since the proper training and resources for adopting the tools are not always available. Due to this limitation, designers will tend to stick to the tools they are already comfortable that are not necessarily best fit for all tasks. This thesis aims to investigate using Autodesk Revit Journals as a non-intrusive method to extract sequential data about user interactions with the software and use collected data to conduct analysis on design and work patterns of users. Additionally, collected data is used to develop a neural network that takes sequence of user commands and trains to predict the next command. The developed neural network can serve as a recommender system suggesting the most suitable commands for the user, enabling improved and efficient workflows and enforcement of best modeling practices.