The integration of AI into software development is transforming coding practices, with tools like GitHub Copilot marking the beginning of AI-assisted development. However, the interaction between developers and AI assistants is not yet well understood, presenting both opportunities and challenges. This PhD thesis addresses three core objectives: (1) improving the effectiveness of developer-AI interactions, (2) redefining development practices to incorporate AI as an essential collaborator, and (3) exploring collaborative dynamics within mixed teams of developers and AI agents. The aim is to establish methodologies that can boost productivity and software quality, paving the way for seamless human-AI collaboration in development.