Worldwide Express Deploys Bedrock Multi-Agent System for Freight Billing Exceptions
AI assistants · Enterprise AI
Worldwide Express implemented an Amazon Bedrock multi-agent system to automate LTL freight billing disputes, cutting active exceptions by 75 percent in six months.
Worldwide Express, handling over 75 less-than-truckload carriers, worked with AWS and LTM to deploy a multi-agent AI framework on Amazon Bedrock and Amazon EC2. The system handles billing mismatches, pricing variances, and unexpected accessorial fees within its SpeedShip system. Six months after rollout across more than 40 carriers, the company reported closing roughly 10,000 exceptions weekly with an error rate below 1 percent, reducing the active exception backlog by 75 percent.
Key takeaways
- Three specialized agents handle invoice matching, pricing variances, and accessorial charges against carrier-specific knowledge bases and rules.
- The deployment operates on Amazon EC2 for orchestration and Amazon Bedrock for foundation model inference, interfacing via existing TMS APIs.
- A human-in-the-loop review was utilized initially, shifting toward bulk approvals as accuracy exceeded 99 percent.
- The company reported a 75 percent reduction in active exceptions and a four-week improvement in time-to-cash.
Source：AWS Machine Learning Read the original ↗
Prepared, checked and automatically published by AI; not individually reviewed by a human.