AI IN SUPPLY CHAIN

By contrast, consumption remains very low in many Middle Eastern and North African countries. These regional variations demand nuanced forecasting models that consider far more than simple averages.

For Diego, this complexity is precisely where AI shows its worth.“ Capturing patterns that are not straightforward to be observed is one of the key advantages brought by AI,” he explains. That said, it takes months of work with data scientists to make the models function effectively. Diego outlines a technique his team applies known as time series decomposition. This statistical method breaks down sales data into its underlying components: trend, seasonal, level and residual. In doing so, Southern Glazer’ s can detect item and location combinations that show strong seasonality and reclassify them into targeted clusters. These decomposed components are then used as explicit inputs in the AI model, giving it visibility over recurring spikes and dips that simpler models might miss.

At OpenText, Mark points to predictive AI as an essential tool for building resilience into supply chains, stating“ the business world is evolving as rapidly as ever.”

However, he stresses that clean, reliable data is key:“ A major difficulty arises from partners with separate data management processes, resulting in differently formatted data.” If inconsistencies and errors aren’ t dealt with, such as duplicates or mislabelled values, then AI systems risk returning

misleading outputs. Mark calls this risk“ hallucinations,” where the AI appears to provide confident forecasts that are, in fact, inaccurate due to faulty data inputs.

Meanwhile Gallup’ s 2022 report shows a noticeable shift in drinking patterns across generations. In the US, only 62 % of adults under 35 reported drinking alcohol in 2021, down from 72 % in 2001.

“ These trends are playing an important role on our long-term forecasting parameters,” Diego says. This trend is being picked up by Southern Glazer’ s AI models, which now analyse up to 36 months of data and project up to 24 months ahead. As forecasts for core categories like beer or spirits trend downwards, the company sees greater innovation activity – particularly in non-alcoholic options.

“ Innovation forecasting is an area we are working on,” Diego explains.“ It is much harder to generate

72 July 2025