← Selected Work Industrial operations · Private engagement

StockGPT Engine

Inventory decisions under long lead times, variable demand and financial constraints.

StockGPT Engine is the decision system we adapt to each client’s data, constraints and workflows. It brings together inventory, sales and financial data, cleans and structures it, detects patterns and projects changes in demand and supply to support purchasing and stock strategy. In this engagement, it was deployed for a stone-processing business.

  • Industrial Operations
  • Agentic Harness
  • Inventory Planning

Product and engine

StockGPT

Our product for inventory and sales analysis. Businesses upload their spreadsheets and get answers on stock, margins and reorders.

StockGPT Engine

The decision harness underneath, adapted to a client’s systems, operational constraints and decision processes. This case describes an Engine deployment.

The challenge

In stone processing, stock decisions affect both working capital and the ability to fulfil future orders. Long replenishment times make shortages difficult to correct, demand is uneven, and unnecessary purchases tie up cash and space.

What we built

We adapted StockGPT Engine to the business’s data and constraints. Agents bring together inventory, sales and financial records, clean and reconcile them, and prepare them for analysis.

The engine then detects demand and supply patterns, projects likely changes and proposes purchasing strategies that respect lead times and financial limits. The team reviews each recommendation before acting on it.

Beyond the state of the art

  • Operations and finance in one decision

    Stock levels are not planned on their own. The engine plans purchases against lead times, demand and working-capital limits in the same decision.

  • An agentic data layer

    Messy ERP exports and spreadsheets are where most planning projects stall. Agents clean, reconcile and structure them, so every analysis starts from consistent data.

  • Scenarios, not single forecasts

    Changes in demand and supply are projected as scenarios, giving the team a range of outcomes to plan against rather than a single number to trust.

Architecture

  1. 01Data ingestionERP · Spreadsheets · Finance
  2. 02Cleaning and reconciliationAgentic
  3. 03Pattern detectionDemand · Supply · Lead times
  4. 04Demand and supply projectionScenarios
  5. 05Constraint-aware planningStock · Working capital
  6. 06Purchasing recommendationsHuman review

Built on the decision harness first developed for Blue World, extended for operational and financial data.