Research Summary: Why Organizational Rules Fail AI: O-I-B-A-R and the Externalization of Decision Boundaries

Research identifies a recurring class of organizational AI failures stemming from a fundamental mismatch between how AI systems receive formal procedures and how human organizations execute them. AI systems are provided with explicit procedures, but human practice incorporates unstated elements such as negative boundaries, runtime judgments, responsibility assignments, and learning history. This discrepancy in knowledge representation at the sociotechnical interface leads to suboptimal AI integration and performance. The O-I-B-A-R framework (OPEN, IS, BUT, ACTION, RESULT) is introduced as a method to externalize these critical, often implicit, decision boundaries, thereby enhancing AI's operational effectiveness.

Why it matters

This analysis highlights a critical challenge in integrating AI systems into organizational processes, emphasizing that technical AI deployment is insufficient without addressing the nuances of human operational knowledge. Successfully bridging the gap between formal procedures and situated practice is vital for realizing the full strategic benefits of AI, avoiding costly failures, and ensuring AI systems perform as intended within complex organizational contexts.

Key insights

• Organizational AI failures frequently arise from a knowledge representation problem at the sociotechnical interface.
• AI systems primarily receive explicit procedures (policies, prompts) while human organizations operate on these procedures augmented by implicit contextual factors.
• Critical implicit factors include negative boundaries, runtime judgments, responsibility assignments, and learning history, which are often absent from AI's operational understanding.
• The discrepancy between formal descriptions of work and situated practice contributes significantly to AI deployment challenges.
• The O-I-B-A-R framework is proposed as a scaffold to externalize these missing decision boundaries, improving AI's alignment with organizational practice.

Source

arXiv — Computers and Society — https://arxiv.org/abs/2608.29055