Worldwide Express cuts active freight billing exceptions 75% with Amazon Bedrock agents

Worldwide Express, part of ShipStation Global, says a multi-agent AI system built on Amazon Bedrock cut its active freight billing exceptions by 75 percent within six months of launch. According to AWS, the company also improved time-to-cash by four weeks and gained a one-time $10M free cash flow infusion.
- Worldwide Express says active billing exceptions fell 75 percent, from $15M in June 2025 to $4M six months later.
- The system uses three specialized agents on Amazon Bedrock and Amazon EC2, integrated with the SpeedShip TMS.
- Agents close 10,000 exceptions weekly (about 50 percent) with a less than 1 percent error rate across over 40 LTL carriers.
- Human approval was used at first; bulk approval became the default once accuracy exceeded 99 percent.
- The company reports a four-week time-to-cash improvement, over $1M in annual savings and a $10M cash flow release.
A billing backlog that froze cash
Worldwide Express reports approximately $5 billion in systemwide revenue, more than 75 less-than-truckload (LTL) carriers and over 130,000 customers. According to AWS, that scale came with a cost: $15 million in free cash flow sat frozen every month in billing exceptions that had not yet been resolved.
The source says roughly 14 percent of LTL invoices fail the standard three-way match between the purchase order, the goods receipt and the carrier invoice. At Worldwide Express, the failures fell into three categories:
- Line-item mismatches: quantities or service codes that did not align with the shipment record.
- Pricing variances: the carrier billed a rate different from the contracted price.
- Unexpected accessorial charges: fees such as liftgate, residential surcharges or reweigh adjustments that were not on the original quote.
Each exception triggered a five-step manual process: identify the type, research the root cause in vendor records, email the carrier or warehouse, wait for a response, and execute the resolution in the transportation management system (TMS). Because every carrier has its own tariff structure and dispute workflow, the work consumed most of the FinOps team's bandwidth. AWS notes that rule-based automation handles predictable workflows, but exception resolution needs contextual judgment, which is why the company turned to agentic AI.
Three agents, one knowledge base per carrier
Worldwide Express worked with AWS and LTM on the solution, which plugs into its TMS, SpeedShip. The company engaged the AWS Business Transformation & Value Advisory team in June 2025 to identify 19 automation opportunities across the Order-to-Cash lifecycle, and moved from assessment to production in less than 6 months.
Amazon Bedrock provides the AI reasoning, while Amazon EC2 runs the orchestration. AWS says building the framework directly on EC2 keeps the orchestration logic in the team's hands and supports open source agent libraries. When an exception is detected, it is routed to one of three specialized agents:
- Invoice Matching Agent: finds the matching shipment record and flags what does not line up.
- Pricing Variance Agent: compares the carrier's charge with the contracted rate and decides whether the difference is valid.
- Accessorial Validation Agent: checks whether unexpected fees are supported by the actual delivery events.
Each agent has an instruction prompt, a set of actions that integrate with SpeedShip, and access to carrier-specific knowledge bases holding tariff rules, surcharge definitions, dispute steps and variance thresholds. The agents use the same APIs as human analysts, so they see the same contract rates, shipment details and delivery proof. According to the source, the same three agents handle all carriers without any carrier-specific code. After resolution, the system writes a plain-language note explaining what was found and what action was taken.
Human review and observability
During the initial rollout, a human reviewer approved each recommendation in the existing SpeedShip UI, a step AWS describes as necessary for FinOps adoption. Once accuracy exceeded 99 percent, bulk approval became the default.
The team built dashboards tracking automation rate by carrier, error rate by exception type, resolution time and human override frequency. When a reviewer rejects a recommendation, the system logs the feedback, and the team refines prompts and knowledge bases based on that rejection history to expand automation coverage.
Results and what comes next
Six months in, the solution is deployed across over 40 LTL carriers. The agents close 10,000 exceptions weekly, approximately 50 percent, with a less than 1 percent error rate, according to AWS. The backlog of active exceptions fell from $15M in June 2025 to $4M six months later. The company also reports over $1M in annual cost savings, a $10M cash flow release and 60x first-year ROI.
Nathanael Pearson, VP of FinOps at Worldwide Express, said in the post: "Today, we release approximately $10 million more in customer billings each week at a speed that simply wasn't possible before."
Worldwide Express is now expanding the approach to other parts of the Order-to-Cash cycle. For other logistics companies with high carrier volumes, AWS suggests starting with the most repetitive exception type, proving accuracy, and then scaling across the rest of the portfolio.


