Southwest Airlines' crew scheduling system SkySolver was overwhelmed during a winter storm, forcing manual scheduling that led to over 13,000 flight cancellations and stranded passengers nationwide.
During Christmas week 2022, Southwest Airlines experienced a massive operational failure when its crew scheduling AI system called SkySolver was overwhelmed by the scale of flight disruptions caused by a winter storm. The system, which normally helps reassign crews after flight disruptions, failed to handle the complex task of matching available pilots and flight attendants with flights across Southwest's network. This forced crew schedulers to resort to manual processes, combing through records by hand. Phone lines became jammed with hold times of eight hours or more as crew members couldn't get assignments. The airline canceled more than 13,000 flights, stranded passengers and bags across the country, and drew criticism from federal officials. Southwest's CEO Bob Jordan publicly apologized, and the incident was expected to cost the airline 3-5% of fourth-quarter earnings. The failure highlighted inadequate technology infrastructure that hadn't kept pace with the airline's growth. Union leaders had previously complained that SkySolver often produced nonsensical crew assignments, such as sending pilots on circuitous deadhead journeys without actually flying planes. The incident prompted calls for technology modernization, with executives acknowledging they needed better tools to handle the airline's scale and complexity.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
AI systems that fail to perform reliably or effectively under varying conditions, exposing them to errors and failures that can have significant consequences, especially in critical applications or areas that require moral reasoning.
AI system
Due to a decision or action made by an AI system
Unintentional
Due to an unexpected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed