Governing generative AI in organizations: a design theory and quasi-experimental field study of sociotechnical guardrails — Leon (2026)
Abstract Generative AI adoption has outpaced organizational governance capabilities. We conceptualize AI guardrails as sociotechnical governance mechanisms, comprising policy, technical, and workflow components that embed organizational norms in deployed AI systems. Extending norm-based coordination accounts, we specify three mechanisms (norm encoding, output monitoring, and escalation) targeting four properties: predictability, fairness, safety, and auditability. We instantiate the theory in a three-layer artifact at a Fortune 500 firm and evaluate it through a stepped-wedge quasi-experiment…
The skinny
The skinny isn't ready yet — notes appear once the transcript is processed.