Store stockout response

The scenario and numbers in this demonstration project are simulated data.

Industry: Retail
What was built: Agents, skills, semantic models and write tools, built on demonstration data
Where the knowledge came from: Demonstration data from the POS, distribution centre, ERP, e-commerce, supplier and membership systems

Coordination that waits for Monday

An influencer video and an online promotion drive sales together, and the flagship store's stock falls below safety stock.
At retailers like this, a transfer order used to come only after the store manager reported on Monday and the operations team met.

The agent uses the store semantic model to check the flagship store's on-hand stock, safety stock and daily sales, and confirms the stockout risk.
The agent compares sales with the pre-promotion baseline, and separates the demand from the influencer video and from the promotion.
The agent checks online that the video's views are still climbing, and takes the upper end of the demand forecast.
The agent uses the store semantic model to scan every store, and finds the stores that can still transfer stock after safety stock and one day of sales.
The agent uses the distribution centre semantic model to check on-hand stock, safety stock and pending inbound, and confirms the ERP has no open purchase order.
The agent uses the e-commerce semantic model to check how many units the promotion takes each week, and proposes a limited-quantity promotion.
The agent checks the membership system and confirms that members have signed up for a back-in-stock notice at the flagship store.

Sindri gathers the transfer, the restock, the limited promotion and the restock note for the store manager into one decision pack.
Before transfer orders, the restock purchase order, the promotion change and the reply to the store manager are written back, the agent stops. Only after the operations manager approves them together on Monday morning does the agent write to the WMS, ERP, e-commerce and POS.

At retailers like this, promotions and product prices will keep changing.
The agent will collect feedback on its answers from users. The feedback will go to Odin.
Wrong answers will become tests. A change will be kept only if scores improve.

The scenario and figures on the slides are simulated data from the demonstration project. The Tyr screens are product prototypes.

このスライドは英語です。

スライドをクリックすると矢印キーで操作できます。スマートフォンでは左右にスワイプします。

Get started

AI feasibility review of a live workflow