Researchers conducted a study on how experienced software developers use AI agents in their work. They found that while developers value the productivity boost from agents, they still maintain control over software design and implementation to ensure fundamental quality attributes. Developers employ strategies to control agent behavior using their expertise and only delegate tasks suitable for automation. The study suggests that effective use of agents requires adherence to best practices and highlights opportunities for improving agentic interfaces and guidelines. --- Why it matters: This research matters because it provides insight into how experienced developers actually use AI agents in practice, rather than relying on hypothetical scenarios or idealized expectations. Understanding the real-world applications and limitations of agent-based development can help engineers and researchers design more effective tools and workflows. Source: https://arxiv.org/abs/2512.14012