The Business

• 10,000 sqm dry storage 3PL facility in Dubai Industrial Area
• 800–1,200 outbound orders per day
• 40+ inbound deliveries per day from clients and suppliers
• 22-person operations team across 2 shifts

The Problem

Approximately 12% of inbound deliveries resulted in some form of variance dispute between the 3PL and the client — items short, damaged, or mis-labelled. Most disputes were raised days after receipt, by which point CCTV footage had often rolled over and resolution required goodwill writes-offs. Annual cost: an estimated AED 480,000 in disputed variances plus opportunity cost of operations manager time spent investigating.

The Solution

Hardware Deployment

4 dock doors instrumented with overhead 4K cameras feeding into a single NVIDIA Jetson AGX Orin edge server. No cloud connectivity required for AI inference; only metadata events flowed off-site.

AI Models

Object detection model trained for carton counting (rectangular box detection with rotational handling), OCR for shipping label extraction, and pallet detection. Per-pallet item count and label data extracted at the moment of unload.

Odoo Integration

Custom integration layer pushed each receipt event into Odoo as a draft incoming transfer with attached photos, item counts, and label OCR. Variance from expected (PO line) was flagged automatically before goods were accepted. Receiving clerk signed off in Odoo before truck departed — variance disputes were resolved at the dock, not days later.

Results (Month 4)

Operational Improvements Beyond Disputes

• Average truck dwell time at dock reduced by 12 minutes (faster check-in)
• Receiving clerk’s daily paperwork time reduced by ~40%
• Damage claims now backed by timestamped video evidence — claim acceptance rate from suppliers up significantly
• Operations manager dashboard in Odoo shows live receiving status across all 4 docks