AI Failures: A Sociotechnical Theory of Brand Crisis

As artificial intelligence systems become more integrated into market-facing roles—from chatbots to recommendation engines—their failures can trigger significant brand crises. Traditional crisis management models often fall short in these scenarios because responsibility for an AI incident is typically distributed across multiple entities: the AI system itself, its developers, deployers, vendors, and even end-users. This paper introduces a sociotechnical process theory to clarify how stakeholders attribute blame in such complex situations. It distinguishes between an AI incident, an organizational crisis, and a full-blown scandal, arguing that the specific configuration of an incident influences who is held responsible. This attribution, in turn, shapes public perceptions of a firm's capability, integrity, fairness, and relationships. The theory proposes "accountable transparency" as a crucial response mechanism. This involves providing timely notice of an incident, offering an intelligible explanation, acknowledging role-specific responsibilities, providing remedies, demonstrating corrective actions, and offering recourse. This framework helps reconcile conflicting findings on how AI involvement impacts brand reactions, offering a more nuanced understanding of managing AI-related reputational risks.