Split freight invoice exceptions across three AI agents: one matches the shipment, one checks the contracted rate, one validates surprise fees, then a person approves the note before it posts: AI agent prompt
AWS published an Oct 30, 2026 case study on Worldwide Express (ShipStation Global), which used Amazon Bedrock multi-agent exception handling inside its SpeedShip TMS. Three agents cover invoice matching, pricing variance and accessorial validation against carrier-specific rules. Across 40+ LTL carriers the agents close about 10,000 exceptions a week (~50%) with under 1% error; active exception backlog fell from $15M to $4M, with ~$10M freed cash flow and a 4-week faster time-to-cash.
The prompt
1. Pull the last 60–90 days of freight or carrier invoice exceptions and group them into: line-item mismatch, pricing variance vs contracted rate, and unexpected accessorials (liftgate, residential, reweigh). 2. For each group, write the decision rules a reviewer already uses, plus where the proof lives (purchase order, goods receipt, delivery events, carrier tariff). 3. Set up three AI agents (or three clear roles in one agent workflow) on Amazon Bedrock, Claude or ChatGPT enterprise connected to your TMS: Matching, Pricing Variance, Accessorial Validation. 4. Give each agent only its tools and that carrier's billing rules. Require a plain-language note: what was found, what action is recommended, and the evidence. 5. Keep a human approve step in the same UI your FinOps team already uses. Start with one exception type and one carrier; raise automation only after error rate stays under your threshold. 6. Dashboard automation rate, error rate, resolution time and human overrides by carrier every week, and feed rejections back into the prompts and knowledge base.
How to use it
- Copy the prompt above.
- Replace every [bracket] with your own details (or fill in the blanks on AgentPlaybook Real Estate, which remembers your answers for the next prompt).
- Paste it into Amazon Bedrock, your TMS (SpeedShip or similar) APIs, carrier tariff/knowledge docs and review the answer before you use it.
Works with: Amazon Bedrock, your TMS (SpeedShip or similar) APIs, carrier tariff/knowledge docs, FinOps approval UI
Real-world example: aws.amazon.com/blogs/machine-learning/how-worldwide-express-automated-
Before you use it: what to check
Already built into this prompt
- Keep a human approve step in the same UI your FinOps team already uses.
- Dashboard automation rate, error rate, resolution time and human overrides by carrier every week, and feed rejections back into the prompts and knowledge base.
Check yourself
- Check every figure, price and total against your own records before anything is sent, posted or booked. A person approves refunds, discounts, payments and price changes.
- Read and edit the draft before it reaches a customer or goes public. If you later automate this prompt, start with "draft and notify me", not "send automatically".
- Check names, dates, numbers and any claim against your own source material before you use the answer. AI can misread or invent details.
Using an AI agent? Download this prompt as a SKILL.md agent skill with these checks built in, or get the .zip for Claude.ai. How to install skills.
Want to make this prompt even better? Grade it with the free prompt grader or build your own with the AI prompt generator.
32 copy-paste agent prompts (PDF) plus 5 new prompts every week.
Get it free