The main asset and a central engine of any organization are its employees, its team. Business success depends to a large extent on their well-being and effective collaborative work. But very often such a factor as high stress at work leads to burnout of valuable employees, worsens the performance and quality of work results, badly affects the health of employees and even colleagues. However, not all managers especially in companies with a distributed territorial structure have objective information about subordinates - deterioration in working atmosphere, severe work-load imbalance among teams and team members, sharp negative changes in relationships with colleagues, customers and supervisors, etc. The study proofs that stress-related issues can be detected using passive corporate data. In order to do so, first of all, an extensive list of stress factors, possible stress predictors and compensation mechanisms from literature is compiled. Then, based on findings a framework is proposed to categorize parameters that can be calculated using data from corporate systems. The framework is illustrated and tested via an application prototype using selected parameters from theoretical framework, open-source datasets such as DReddit and Enron, advanced NLP techniques such as Recurrent neural network, Bidirectional Encoder Representations from Transformers model, and Linguistic Inquiry and Word Count. Upon detecting elevated stress levels, the developed application prototype generates a personalized list of identified stress factors and stress compensators. By leveraging existing corporate systems and requiring minimal user input, the proposed approach offers a non-intrusive and efficient tool for identifying and tracing workplace stress and promoting employee well-being.