Demand Planning Agent
A planning agent that ingests sales, weather, supplier lead times, and marketing signals to produce daily replenishment plans with confidence intervals.
Demand planning has been stuck in weekly cycles and Excel macros for two decades. The Demand Planning Agent breaks that cadence — running continuously, learning from every shipment, and producing plans planners can override.
How it works
The agent maintains a rolling forecast for every SKU at every node, updated whenever a new signal lands:
- Point-of-sale data refreshes hourly
- Weather and event signals refresh daily
- Supplier lead-time changes feed in real time
- Marketing and pricing actions are simulated as counterfactuals
When the forecast crosses a threshold of confidence, the agent emits a replenishment recommendation. Planners review a queue of recommendations, not a sea of cells.
Pilot results
In a recent pilot across a regional grocery network, stock-outs dropped 32% and inventory carrying cost dropped 11% in the first 60 days — primarily from catching slow-moving SKUs the weekly cycle had been missing.