$11M of inventory carry out, fill rate up, at a $420M specialty distributor.
A 14-warehouse distributor cut order-exception cycle time by 78%, reduced inventory carry by $11M, and surfaced category shifts six weeks earlier than the prior quarterly cycle.
.01Exception cycle time−78%across orders and shipments
.02Inventory carry−$11Mwith fill rate held or improved
.03Demand signal lead+6 weeksvs. the prior quarterly cycle
The Situation
Order exceptions across fourteen warehouses moved through email and a heavily customized WMS. Demand planning ran on a monthly cycle that lagged real signal by weeks. Procurement coordination consumed two full headcount per region. Quality patterns in returns and tickets surfaced only at the quarterly business review, four weeks after they should have.
The COO had recently lost three regional ops leads to retirement and competing offers. The team coming in behind them was less seasoned and less tolerant of the email-and-spreadsheet operating model. Something had to change before the holiday peak.
The Engagement
Two-week audit phase mapped exception types by warehouse, the demand planning cadence and source data, the procurement coordination chains by region, and the quality feedback loop from returns and tickets back into category management. We pulled three years of order, shipment, and return data into Snowflake and ran category-shift analysis on the same data the team had only ever seen at quarter-end.
Build ran eight weeks. AgentPrime principal engineer, solutions architect, and ops lead embedded with the COO, the lead category manager, and the WMS systems owner. Phased deployment started with exception detection (highest volume, fastest payback) and expanded into demand sensing and supplier scorecards.
What We Built
A logistics agent across the exception, demand, and procurement chains. Order and shipment exceptions are detected at ingest, classified, and routed by region against the actual approval matrix. Demand sensing tunes against live POS signal plus three years of historical patterns, surfacing category shifts continuously instead of at quarter-end. Supplier scorecards generate automatically with auto-drafted reviews ahead of supplier conversations.
Inventory optimization runs across SKUs and locations with policy bounds that reflect the company's actual service-level commitments — not a generic safety-stock formula. Quality pattern detection from returns and tickets feeds back into the category team so problems surface in weeks instead of quarters.
The Outcome
Order-exception cycle time fell 78%. Inventory carry reduced by $11M without degrading fill rate — in two product lines fill rate actually improved against the prior baseline. Demand sensing surfaced category shifts six weeks earlier than the prior quarterly cycle, which let category management get in front of two emerging trends rather than chasing them after the fact.
Two regions reabsorbed their procurement coordinators into category management roles — work that was higher-leverage, higher-paying, and more durable. The COO went into the holiday peak with the highest fill-rate forecast he'd ever signed off on, and beat it.
"We took eleven million dollars of carry out without breaking service. Six months earlier I would have told you those two things couldn't happen at the same time."
