Current human-AI collaboration (HAIC) research suffers from fragmented collaborative taxonomies and a lack of systematic alignment with task characteristics. To address these gaps, this study constructs a modular HAIC design framework grounded in the task-process model. On the basis of analyzing ten critical task characteristics, the framework defines eleven fundamental collaborative sub-patterns spanning task input, processing, and output stages, while clarifying their underlying combination rules. Finally, utilizing six representative team tasks — such as new product development and multidisciplinary consultation — we demonstrate how to sequentially select and adapt collaborative sub-patterns based on specific multi-stage and multi-dimensional task features. By providing a comprehensive, multi-feature, and modular theoretical collaboration framework, this research integrates fragmented collaborative models and advances HAIC design to an operational level. These findings offer both theoretical support and a practical roadmap for organizations to systematically plan and optimize human-AI collaborative workflows. 

针对当前人智协作研究中存在的协作模式体系碎片化、未能与任务特征系统关联的问题, 构建以任务过程模型为基础的模块化人智协作设计框架, 针对10项关键任务特征, 提出覆盖任务输入、处理和输出3个环节的11种基础协作子模式, 并明确其组合规则。以新产品开发和多学科诊疗等6个典型团队任务为例, 演示如何根据具体任务在多环节、多维度上的特征, 依次选择协作子模式。通过提出全过程、多特征、可组合的协作模式, 整合碎片化的协作模式, 将协作设计推进至操作化层面, 为组织系统规划和优化人智协作工作流程提供理论支撑与实践路径。