Three stages. The middle one is where the language model lives — and it is metered openly.
Carrier or terminal invoice, B/L, booking confirmation, gate-in/gate-out and vessel events (EDI, portal export or screenshot), your contract or tariff with free-time terms, and the email thread with the carrier.
Extracts dates and references from every document, builds a single timeline, computes free time and chargeable days under the clause that applies, compares with the invoice line by line, and flags inconsistencies (events the carrier's invoice ignores, holidays miscounted, wrong tariff).
A dispute file: timeline with sources, clause-by-clause check, computed vs invoiced days, list of missing evidence to request, and a dispute letter draft in the carrier's preferred format. Export to PDF or your TMS notes.
Illustrative output.
A dispute file reads 5–15 documents and an email thread per shipment. Teams handling 10–50 disputed shipments a day run roughly 20–100 million tokens a month. Day-count maths is deterministic code; the model is spent on reading documents and drafting.
We are onboarding a small number of design partners with real documents and real deadlines. Tell us what you work on and we will set up a workspace with your material.