AI + Supply Chain
Stock health score, dynamic safety stock, reorder points, automatic purchase suggestions and demand anomaly detection.
The AI + Supply Chain page recalculates everything from stock history every time it opens. Nothing needs setting up apart from the lead time and minimum order on each item.
What is calculated
- Demand is read from stock issued over the last 12 weeks.
- Dynamic safety stock = 1.65 × standard deviation of demand over the last 8 weeks × √lead time. That is a 95% service level: enough stock in 95% of weeks.
- Reorder point = average weekly demand × lead time + safety stock.
- Stock health: every item is grouped as stock-out risk, needs reordering, overstock (enough for more than 16 weeks) or healthy. The health score is the percentage of healthy items.
- Demand anomalies: last week is flagged if it is more than 2.5 standard deviations away from the 11 weeks before. The change is broken down by sales channel (Modern Trade, E-commerce and so on) so you can see where it came from.
Purchase suggestions
When stock on hand plus stock on order falls below the reorder point, Air-M suggests a quantity that covers the lead time plus one week, rounded up to the vendor's minimum order.
- Click Order n on one suggestion to create a PO for that item only, or Act on n suggestions for all of them at once. Air-M creates one PO per vendor and submits it straight away: below the approval limit it is approved immediately, above the limit it waits for approval.
- For finished goods that have a bill of material, the button is Plan production n and the result is a production work order, not a PO.
- POs and work orders created from AI suggestions can be found in Supply Chain and Manufacturing.
If the numbers look odd
- Items with no sales history cannot be calculated yet; wait for a few weeks of data.
- A wrong lead time or minimum order gives wrong suggestions. Check the item data.
- Physical stock differs from the system: do a stock count first.