A large after-market automotive component manufacturer based in Detroit came to us with a mission to drive expansion and facilitate growth into new markets. To accomplish this, however, they needed to drastically reduce forecast errors and more accurately predict demand across their product portfolio. The company currently operates 2 full-scale manufacturing facilities with 3 distribution centers supplying retail customers across the country and manufacture 200,000 individual product SKU’s.
Forecast errors were creating an enormous surplus inventory and costing the company $25 million per year in carrying and obsolescence costs. The company was carrying $250 million in total inventory across the 200,000 SKU product portfolio. Growing into new markets would amplify the problem and it made predictive modeling virtually impossible.
We deployed individual deep learning AI agents for each of the 200,000 product SKU’s, increasing forecast accuracy by a factor of 5, enabling a 15% reduction in surplus inventory and reducing carrying and obsolescence costs by 80% annually.
Deployment of AI Agents
Deep learning models were trained on the trailing 12 months of ERP system data and POS data directly from their retail customers
Custom models are deployed as individual, on-premise AI agents running autonomously on top of ongoing ERP system data
Outcomes are incorporated into existing company workflows around supply chain management
The custom models continue to learn and improve from new and emerging data over time
By The Numbers
$35M reduction in surplus inventory
$20M annual decrease in carrying & obsolescence costs
5X increase in demand forecast accuracy