Proposal week 26-31 May. 78 stores, 240 priority SKUs. 235 automatic, 3 critical. Softshell Jacket M stockout in Milan/Turin by Thursday, central warehouse 140 units: transfer 60 Milan + 40 Turin. Running Shoe 38 overstock Rome South +180 units: redistribute Naples +90, Bari +50.
AGENT · STOCK ALLOCATION OPTIMIZER
Stock reaches the stores where demand is waiting for it.
Stock Allocation Optimizer produces weekly the proposed stock allocation between the retailer's central warehouses and stores. It combines demand forecasts by store, available stock, and logistics constraints. Structured output for the retail supply chain manager.
Stock Allocation Optimizer at work.
Approved. Proceed with the critical transfers. Recording approval.
Transfer orders generated. Warehouse managers notified. Registry updated.
Why it exists.
Stock allocation is one of the core operations in mid-to-large retail. A chain with fifty or two hundred stores has demand patterns that vary by geography, seasonality, and local events. Manual allocation takes time and regularly produces overstock at some stores alongside stockouts at others.
How it works each week.
Stock Allocation Optimizer produces the allocation proposal every week. It reads the demand forecast for each store — built on historical sales data and configured contingency factors — the stock available in central warehouses and regional depots, and the logistics constraints on transfers (capacity, delivery windows, transfer costs). It calculates the allocation consistent with those constraints.
Approval stays with the manager.
The retail supply chain manager approves the proposal, adjusts exceptional cases, and triggers the transfers. The agent does not generate movement orders without approval.
The teams that optimise stock availability across the retail network.
Retail supply chain manager
Receives the ready allocation proposal every week, with critical cases (imminent stockout, localised overstock) highlighted for approval. Manual preparation time drops; analysis time focuses on exceptions.
Weekly allocation 42 stores
Store Madrid center +18% demand
Warehouse MI buffer stock ok
SKU-2210 push to 6 stores
Estimated saving: 14% inventory
Merchandise manager or buyer
Sees stock distribution aligned to the demand forecast, not to historical allocation rules that have calcified over time. Seasonality and local events enter the calculation rather than staying as afterthoughts.
Brief 07:00 12 active segments
Gold segment +8% engagement
Silver segment redemption -3%
Anomaly · campaign 14 discarded
Monday action map
Store manager
No longer handles urgent stock requests outside the standard cycle. The weekly allocation arrives with enough lead time to plan the store layout. Exceptional cases — a local event, an unplanned promotion — are handled as point requests to supply chain.
Line L3 · SKU-441 7-day fcst
Week 23 8,400 units
Week 24 9,100 units
MES · MRP fed
Variance vs plan: +6%
A weekly allocation proposal that prevents stockouts and overstock.
The agent prepares the proposal. The manager receives Monday.
For a fashion retailer with seventy-eight stores across three areas (North, Central, South), the weekly allocation proposal arrives every Monday morning in the supply chain manager's work channel. Stock Allocation Optimizer worked over the weekend: it read the previous week's sales data, the demand forecast for each store, the current stock in central and regional warehouses, and the logistics transfer constraints.
Two hundred and thirty-five automatic. Five critical.
For two hundred and thirty-five of the two hundred and forty SKUs, the proposal is automatic. For five critical cases — imminent stockout at two high-traffic stores, localised overstock at three stores in zones with lower-than-forecast demand — the agent highlights the situation with the proposed action and recommended quantities.
Fifteen minutes to approve. Transfers depart.
The supply chain manager approves in fifteen minutes on the work channel. Transfer orders go to the warehouse managers. Stock arrives at the stores before the weekend. Decisions stay in the registry for allocation effectiveness analysis.
Declarative rules from the customer's retail supply chain team.
The rules of Stock Allocation Optimizer are declarative. The customer's retail supply chain team defines in a readable format the demand forecast model for each store (factors to include: sales history, seasonality, local events, planned promotions), the logistics constraints (transport capacity per route, delivery windows, transfer costs), and the thresholds for flagging critical situations (imminent stockout threshold, overstock threshold). WMS and ERP integration is built during delivery via a dedicated adapter.
- Language
- TypeScript (Node.js)
- LLM model
- customer's choice: Anthropic, OpenAI, Mistral, open source models hosted internally, AWS Bedrock for a private model
- Built-in controls used
- pii-detector, tool-rate-limit
- Native delivery channels
- Slack, Telegram, WhatsApp, OpenAI-compatible HTTP
- Scheduling
- configurable per instance (typical weekly, weekend run for Monday brief)
- WMS (Warehouse Management System) integration
- dedicated adapter built during delivery by the Exelab team
- ERP integration
- dedicated adapter built during delivery by the Exelab team
- Memory
- persistent per instance, pgvector + PostgreSQL FTS
- Registry
- immutable, queryable with a standard SQL client
How Stock Allocation Optimizer works in detail.
Not necessarily. Stock Allocation Optimizer can build the demand forecast from the historical sales data per store available in the customer's WMS or ERP. Alternatively, if the customer already has a forecast model (e.g. a dedicated ERP module, a demand planning tool), the system integrates with that as an input source. The approach is defined during discovery.
The standard frequency is weekly, suitable for most retailers. For high-turnover categories (fresh goods, fast fashion with weekly new arrivals) it is possible to configure twice-weekly or daily frequency for priority SKUs. The optimal frequency depends on transfer lead times and warehouse operational capacity.
The system can handle point transfer requests (e.g. an unplanned promotion at a store, an unexpected local event) as off-cycle operations, always subject to supply chain manager approval. The optimization logic applies to urgent requests too, within the available logistics constraints.
The typical pattern is 12-18 weeks. Discovery and store network and management system mapping two weeks, forecast model and allocation rule configuration four weeks, WMS and ERP integration four weeks, testing with real data and hand-off to the supply chain team two to three weeks. The range depends primarily on the complexity of the WMS and the variety of SKUs to manage.
From a 30-minute conversation to the squad in production.
A 30-45 minute conversation to understand how Stock Allocation Optimizer would configure to the customer's case. Number of stores, WMS and ERP in use, demand forecast model already available.