Nowadays, service-providing employment accounts for about 80% of jobs [1], most of which require low physical activity. This sedentary behavior is carried out by workers who are mainly seated for long periods and usually also includes postural behavior that can cause pain and chronic injuries. To increase worker well-being, in addition to encouraging and promoting physical activities, posture monitoring can be helpful in identifying wrong behaviors, leading to less postural stress and musculoskeletal injuries. This paper presents a non-intrusive posture monitoring system, based on a matrix of pressure sensors inserted on a chair cushion, where the worker is seated. The gathered data is then subject to a machine learning algorithm that can infer favorable/unfavorable postures. The results lead to more than 80% accuracy, which means that this setup can be helpful in informing users of their postures that lead to their correction, thus helping to mitigate pain and injuries.