Logistics planning long relied on simple statistical models, Excel spreadsheets and the experience of planners. AI has introduced a capacity for complexity that is structurally changing the discipline.
The problem AI solves
Classical demand forecasting models fail in the face of discontinuous events: promotions, weather, supply crises, viral social media behaviour. They work on historical averages and miss ruptures.
AI models can integrate dozens of external variables in real time and adjust forecasts dynamically. It is on this specific point that the gains are best documented.
What Carrefour put in place
Carrefour deployed an AI-based demand forecasting system across its entire French network, covering more than 5,000 fresh food references. The model integrates weather data, local event calendars, sales histories and promotional signals.
Measured result after 18 months: a 19% reduction in unsold fresh produce, and a 12% reduction in shelf stockouts across the same categories.
What a component manufacturer learned
A European electronics supplier used a stock optimisation model to reduce its safety stock across 800 references. After six months, average stock had fallen by 22% with no degradation in service levels.
But the logistics team describes an unexpected side effect: planners struggled to understand, and therefore trust, the model’s decisions. The rollout required six additional months of work on decision explainability and team training.
Key takeaway
AI demonstrably improves forecast accuracy and inventory optimisation. Its operational adoption requires explainability work and change management that is consistently underestimated in project plans.