A hybrid approach combining AI and classical optimization is needed for promotional planning, as promotional events compress planning cycles into "just a few days," often causing inventory imbalances. Research from IHL Group puts the global cost of inventory distortion at "$1.73 trillion annually," or "roughly 6.5% of global retail sales," with supply chain disruption as the largest contributor at "$301 billion." While "76% of the retailers currently using machine learning for demand planning report positive results," fewer than "one in four retailers have actually rolled out AI" in areas like promotional planning. The article argues that most companies piloting AI for promotional inventory use a setup built around a "large language model (LLM)," but pure LLM approaches "cannot be trusted to respect hard constraints," whereas classical methods like "mixed‑integer programming" enforce constraints but cannot interpret unstructured inputs. The proposed solution is a "hybrid approach" that wires both tools together.