This chapter explores how large language models (LLMs) can be practically and responsibly integrated into assessment practices in early elementary education (Kindergarten to Grade 3). Focusing on the realities of time-constrained classrooms, we examine how generative artificial intelligence (AI) can support teachers by automating high-effort tasks while preserving the central role of professional judgment and human interaction. The chapter centers on two complementary applications: automatic item generation and automatic feedback generation (AFG). In the first section, we describe how LLMs can assist teachers in designing age-appropriate assessment items while addressing developmental, linguistic, and equity-related considerations. In the second section, we examine the evolution of AFG systems and discuss how LLM-based feedback can be designed to be timely, supportive, and instructionally meaningful for young students. Throughout the chapter, we emphasize practical strategies, current research evidence, and ethical considerations, equipping educators with actionable guidance for leveraging AI to enhance assessment quality and feedback in early elementary classrooms.