Labour and Material Cost Advisor (Agentic AI model) 

Procter & Gamble’s Ben Cat Plant packages various razors at high volume and tracks labour and material cost per SKU by hand across multiple Excel workbooks. When seeking to reduce costs, an employee has to go through these workbooks and create a PowerPoint presentation to simulate and test various possible cost simulations and outcomes to select the most appropriate simulation for a particular product. As the portfolio grew and wages rose, that process became increasingly difficult to scale.



Cost Advisor is an agentic AI system that makes cost visibility continuous. It reproduces the plant’s own $/SU calculation in fixed precision arithmetic, generates ranked, costed interventions across six optimization goals, and monitors 623 SKUs against a ±5% volume swing deviation threshold.

The key engineering contribution is the separation between deterministic computation and LLM-generated explanations. Deterministic engines calculate all the numbers displayed across the interface, and the language model may only explain them. A grounding gate traces each numeric claim back to the calculation that produced it. Parity with reference Excel workbooks is ensured. This results in a transparent system where users can audit every figure that is presented to them.