Ergonomics evaluation methods can be used to assess risks for work-related musculoskeletal disorders and promote physical well-being of people However, research has shown that different assessors often comes to different conclusions in regard to risks. Hence, the subjective nature of observation-based ergonomics evaluations can cause reliability issues, while also being time-consuming to perform. Recent developments in technologies such as camera-based and inertial measurement unit (IMU) sensor-based motion capture systems facilitate the measurement and digitalization of human postures over time. Hence, it is assumed that such technology can become integrated into the process of performing ergonomics evaluations, to evaluate workers’ well-being more objectively and efficiently. This study investigates the use of a motion capture system to record hand and finger motions, and the application of the random forest machine learning algorithm to classify hand postures into categories of grip types. The results show that random forests can, based on the motion capture data, automatically and successfully classify hand postures into three grip types defined by the HandPak ergonomics evaluation method. The random forest models did not exhibit the overfitting issues typically associated with decision trees in similar classification problems. However, the training and test data were obtained from only two subjects. Including more subjects in the training and test data to account for posture variation could improve the accuracy of the random forest models.