The Problem
The company operates 11 US manufacturing facilities including the flagship 750-acre Appliance Park and 14 distribution centers across the nation, shipping over 131,000 freight loads annually worth $115M across domestic truckload, LTL, ocean, and air modes. Its freight audit ran through Uber Freight supplemented by SPS Commerce — but with a critical configuration flaw: invoices under $3,000 were approved automatically without line-item validation. Carriers knew the threshold. It was not exploited randomly — it was a predictable gap that enabled carriers to bill accessorial charges, surcharges, and dimensional errors below the trigger level without systematic detection.
Detention charges were managed entirely outside the standard audit flow via email approvals, with manual validation against TrackX and Kaleris YMS data. 68 logistics providers generated detention events across the network. Verifying each one required someone to pull the YMS timestamp records, match them to the carrier’s billing window, and confirm or dispute the charge by email. The process was slow, inconsistent, and dependent on the individual analyst managing the queue. International ocean and air accruals used Infor Nexus with 5% tolerance thresholds and manual quarterly fuel updates — meaning accruals could be 5% wrong by design, and fuel surcharge mismatches accumulated for an entire quarter before anyone corrected them.
GL cost allocation was a persistent problem downstream. Manual GL code assignment by origin-destination pairs from Uber Freight caused frequent rejections due to formatting mismatches. 5–6 FTEs supported domestic audit alone. The allocation errors that propagated into SAP required manual correction at close, consuming additional finance team capacity and producing month-end accruals that didn’t match audited actuals.
What Freehand Did
Freehand replaced both Uber Freight and SPS Commerce with an AI audit layer that validates every invoice at the line level — no threshold below which charges auto-approve without validation. The Audit Agent runs a multi-way match on every invoice across all 131,000 annual freight loads: base freight against the contracted rate in the Rate Manager, accessorial charges against entitlement rules per carrier, and billed shipment data against the actual shipment records from Oracle via SPS/SFTP. The sub-$3,000 blind spot is closed. The $6M in annual recovery is the systematic capture of what the threshold-based model had been releasing without review.
Detention validation now runs automatically against TrackX and Kaleris YMS timestamps. When a carrier bills detention, the Audit Agent validates that the dwell time billed matches the recorded dwell time at the facility — using the same YMS data that had previously required manual email coordination to retrieve. Charges without YMS support are flagged before payment. The Collaboration Agent manages the carrier dispute communication with structured evidence assembled from the YMS records. The 68 carriers generating detention events across the network are all in scope, consistently, without the variability that came from individual analyst capacity.
The GL Coding Agent eliminated the manual allocation process entirely. Every invoice is allocated to the correct origin-destination pair, cost center, and SAP account automatically based on configurable business rules — with self-correcting logic for formatting variations that had previously caused rejections. The 5–6 FTEs supporting domestic audit alone are redeployed to strategic carrier management and network optimization. The 2–3 FTEs freed from GL coding work on analysis rather than data entry. GE’s zero-distance supply chain strategy — faster inventory turns through enhanced freight visibility — is supported by a 50% improvement in inventory turns through the real-time spend intelligence the Audit Agent’s outputs feed into planning.
