This project studies work-related conflicts in the Swedish labor market using a newly constructed dataset of labor-law disputes adjudicated in Swedish district courts since 1990, linked to rich employer–employee register data. A central objective is to understand how the incidence, costs, and resolution of work-related conflicts vary with collective bargaining agreements (CBAs), a core institution of the Swedish labor-market model. The first research question examines how the prevalence and types of work-related conflicts differ between workplaces with and without CBAs. While collective agreements are commonly assumed to reduce legal conflict by providing clear rules and dispute-resolution mechanisms, systematic evidence on this claim is scarce. By combining court judgments with detailed firm-level and worker-level registers, the project analyzes whether CBAs reduce formal litigation or instead shift conflicts toward informal negotiations outside the court system. The second research question focuses on the economic and career costs of work-related conflicts for both employers and employees. For employers, conflicts may entail legal costs, management time, productivity losses, and reputational risks. For employees, disputes may lead to income losses, job separations, prolonged uncertainty, and persistent effects on wages and employment trajectories. Using longitudinal register data, the project quantifies both short- and medium-term consequences of conflicts for the involved parties. The third research question studies how labor disputes are adjudicated by the courts, with particular emphasis on judicial behavior. By exploiting quasi-random assignment of judges to cases, the project examines whether judges differ systematically in their rulings, remedies, and interpretation of labor law, and whether such heterogeneity interacts with the presence of collective agreements. This allows an assessment of the predictability and consistency of legal enforcement in the labor market. A key methodological component of the project is the use of a locally hosted large language model (LLM) to process and structure unstructured legal text from court judgments. The LLM is used to extract relevant information—such as parties, claims, outcomes, judges, and legal grounds—while ensuring that sensitive personal data never leave secure academic computing environments. This enables large-scale empirical analysis of work-related conflicts while maintaining strict compliance with data-protection regulations. Overall, the project contributes new evidence on how collective bargaining, legal institutions, and judicial decision-making shape conflict, costs, and stability in the Swedish labor market.