The user positioning data from regional GNSS Correction Services can be an unconventional source for cartographic updates. These data include user geographical coordinates, date, and connection duration. This analysis aims to automate the identification and classification of newly constructed civil works using regional data from the Spanish Andalusian Positioning Network between 2008 and 2020. The methodology was developed in a QGIS and Python environment, following a two-step process. First step involves the geometrical and temporal characterization of polygons. These polygons were generated from filtered and clustered user positions using their convex hull. They were characterized using a geometric index (to distinguish between polygonal works, linear works, and other activities), sinuosity (to discard areas with insufficient inner points or negligible perimeter values), temporal parameters (start, and end dates of connections) and bounding box coordinates. The second step stars with exporting the data into a CSV file and testing of supervised Machine Learning approach to classify work types. A training dataset was categorized into four classes: unidentified, linear works, polygonal works, and activities outside public works. This classification was manually validated by cross-referencing hulls with geospatial layers from the Andalusian Spatial Reference Data. Newly constructed elements were confirmed using recent and historical orthophotos from the National Plan for Aerial Orthophotography. The Decision Trees method achieved the highest performance with 75% accuracy. This highlights the potential of integrating unconventional sources in cartographic updates and using Artificial Intelligence to accelerate the detection of new construction works.