Scrum has become the dominant Agile framework for managing software projects, emphasizing adaptability, collaboration, and continuous delivery. The recent emergence of Large Language Models (LLMs) has created new opportunities to support the knowledge-intensive nature of Scrum work. While prior studies have vastly examined LLMs in technical tasks (e.g., coding and testing), their role in management activities (e.g., refining artifacts, supporting events, and assisting team roles) remains quite scarce. In this study, we investigate the use of LLMs in Scrum management activities through a survey of 70 Brazilian professionals. Among them, 49 actively use Scrum, and 33 reported using LLM-based assistants in their Scrum practices. The results reveal high proficiency and frequent use of LLMs among respondents: 85% report intermediate or advanced proficiency with LLM tools, and 52% use them daily. Exploring Scrum dominates adoption, artifacts and events show strong but selective use, and other management tasks remain the most cautiously assisted by LLMs. The main benefits include increased productivity (78%) and reduced manual effort (75%). However, critical risks persist: many respondents report ‘almost correct’ outputs (81%), express confidentiality concerns (63%), and encounter hallucinations during use (59%). This work provides one of the first empirical characterizations of LLM use in Scrum management, identifying current practices, quantifying benefits and risks, and outlining directions for responsible adoption and integration in Agile environments.