Context: Generative Artificial Intelligence (GenAI) tools, such as GitHub Copilot and GPT tools, represent a paradigm shift in software engineering. While their impact is clear, most studies are short-term, focused on individual experiments. The sustained, team-level effects on productivity within industrial agile environments remain largely uncharacterized. Goal: This study aims to provide a longitudinal evaluation of GenAI's impact on agile software teams. We characterize its effect on developers' productivity by applying the multi-dimensional SPACE framework. Method: We conducted a multi-case longitudinal study involving 3 agile teams at a large technology consulting firm for around 13 months. We collected and compared quantitative telemetry (Jira, SonarQube, Git) and qualitative survey data from historical (pre-adoption) and research (post-adoption) sprints. Conclusion: GenAI tools can significantly improve team performance and well-being. Our key finding is a sharp increase in Performance and perceived Efficiency concurrent with flat developer Activity. This suggests GenAI increases the value density of development work, not its volume. This finding validates the necessity of multi-dimensional frameworks like SPACE to capture the true, nuanced impact of GenAI in situ, which would be invisible to studies measuring Activity alone.

翻译：背景：生成式人工智能（GenAI）工具，如GitHub Copilot和GPT工具，代表了软件工程领域的范式转变。尽管其影响显而易见，但大多数研究都是短期的，侧重于个体实验。其在工业敏捷环境中对团队层面生产力的持续影响在很大程度上仍未得到充分描述。目标：本研究旨在对GenAI在敏捷软件团队中的影响进行纵向评估。我们通过应用多维度的SPACE框架来刻画其对开发者生产力的影响。方法：我们在某大型技术咨询公司进行了为期约13个月的多案例纵向研究，涉及3个敏捷团队。我们收集并比较了历史（采用前）与研究（采用后）冲刺阶段的定量遥测数据（Jira、SonarQube、Git）和定性调查数据。结论：GenAI工具可以显著提升团队绩效和幸福感。我们的关键发现是，在开发者活动量保持平稳的同时，其绩效和感知效率出现了急剧上升。这表明GenAI提高了开发工作的价值密度，而非其工作量。这一发现验证了采用SPACE等多维框架来捕捉GenAI在真实情境中微妙影响的必要性，仅测量活动量的研究将无法揭示这些影响。