AI-Powered Forecasting Under Disruption: Lessons from the Covid-19 Crisis for Adaptive Supply Chains

The Covid-19 crisis disrupted global supply chains, exposing critical weaknesses in traditional demand forecasting models that rely on historical data and seasonal trends. As demand patterns became unpredictable, companies struggled to adapt, leading to frequent manual overrides and inefficiencies. AI-driven forecasting emerged as a key solution, improving responsiveness and integrating real-time data. Yet challenges such as model opacity, data volatility, and the inability to handle unprecedented disruptions limited its effectiveness. This study examines how five companies from different industries adapted their forecasting processes during the crisis. The findings show that companies relying solely on statistical models faced severe forecasting inaccuracies, while those integrating AI saw improved adaptability. However, AI systems required human oversight, as unexplained predictions and reliance on past data made full automation impractical. Companies that successfully integrated real-time external data, such as consumer sentiment analysis and macroeconomic indicators, responded more effectively to demand shifts but faced challenges in filtering out noise. Firms also adjusted forecasting frequency, moving from quarterly to weekly or even daily updates, and restructured inventory strategies, shifting from lean models to increased safety stock. The study highlights the importance of hybrid AI-human forecasting approaches, where AI provides baseline predictions, but planners retain final decision authority. Companies that fostered cross-functional collaboration between forecasting, supply chain, and production teams were better able to align forecasts with operational constraints. The results suggest that future forecasting systems should enhance AI explainability, integrate real-time market signals dynamically, and develop more flexible, adaptive frameworks to improve resilience in volatile markets.