Scholarlink

Artificial intelligence (AI) offers transformative opportunities across industries, but poses new challenges related to fairness, safety, and compliance. Despite the urgent need for governing AI development projects systematically, there is a research gap regarding practical frameworks. Established development frameworks, such as agile development or CRISP-ML(Q), either lack AI-specific aspects or are incompatible with firms’ stringent governance needs. Therefore, we designed and evaluated a phase model to govern and manage AI development projects, drawing on extant literature and close collaboration with AUDI AG, a leading car manufacturer. The resulting artifact defines six incremental phases to support procedural governance and specifies key activities and project checkpoints. The study contributes a rigorously developed artifact and additional recommendations to support firm-specific adaptations of the model. The results provide actionable guidance for practitioners to foster business/AI alignment, formalize responsible AI practices, and mitigate project risks.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Toward Procedural AI Governance: Designing a Phase Model for AI Development Projects

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Michael Weber,
  • Martin Biller,
  • Timo Phillip Böttcher,
  • Andreas Hein,
  • Helmut Krcmar

摘要

Artificial intelligence (AI) offers transformative opportunities across industries, but poses new challenges related to fairness, safety, and compliance. Despite the urgent need for governing AI development projects systematically, there is a research gap regarding practical frameworks. Established development frameworks, such as agile development or CRISP-ML(Q), either lack AI-specific aspects or are incompatible with firms’ stringent governance needs. Therefore, we designed and evaluated a phase model to govern and manage AI development projects, drawing on extant literature and close collaboration with AUDI AG, a leading car manufacturer. The resulting artifact defines six incremental phases to support procedural governance and specifies key activities and project checkpoints. The study contributes a rigorously developed artifact and additional recommendations to support firm-specific adaptations of the model. The results provide actionable guidance for practitioners to foster business/AI alignment, formalize responsible AI practices, and mitigate project risks.