The documents a forwarder runs on
Services
What we build, and what we promise.
Four things, built on your documents rather than on a general-purpose model that has never seen them.
01
Answers from your own documents
Tariff sheets, contracts, invoices, bills of lading and CMR notes, indexed so they can be searched by meaning and by exact reference — port codes and quote numbers included. Every answer arrives with the passage it came from.
02
Quote requests, prepared
An incoming request is read, the relevant past quotes and contract terms are pulled, and a draft is assembled. Rates are checked against your contracts by ordinary code, not by a model — that part is money, so it stays deterministic.
03
Accuracy we measure and report
We build a test set from questions your team actually asks, each paired with the passage that should answer it, and report how often the system finds it. Re-run as your archive grows — retrieval degrades quietly, with no error to warn you.
04
Nothing acts without approval
The system never gets write access to your own software. It proposes from a fixed set of actions we define with you, and anything consequential stops and waits for a named person. An interrupted review resumes exactly where it stopped.
Why it can be trusted
We show receipts, not promises.
There are no client results to point to yet, so here is what we can show: a measured number, the method behind it, and exactly what we will and won't claim.
Grounded in your documents
95%
found on the first try
100%
found within the top 3
38 questions, each paired with the passage that should answer it — the same method described in Services, run here on our own test corpus (what's usually called recall@1 and recall@3). Not a client result: we don't have one yet.
We can walk through a real example on request — a public tariff sheet or bill of lading, with the exact passage an answer would cite. Labelled as a demonstration, not a client engagement.
Retrieval trace
See how an answer earns its place.
Question
Which contracted rate applies to quote RTM–2048?
Retrieved source
TARIFF · PAGE 07 · LINES 18–21
Rotterdam → Hamburg · 40 ft container · contracted base rate
Source check
Expected passage found at result #1
Action
Draft prepared · waiting for approval
Illustrative interface using a synthetic document — not a client result.
What is measured
- 01Pair each question with its expected passage
- 02Run the question against the document archive
- 03Record where the correct passage appears
How we work & pricing
Three stages. Stop after any of them.
Each one ends with something you keep, whether or not you carry on to the next.
Free · about a week
Find out what is actually answerable
What documents you have, what state they are in, and what your team asks day to day. A week is enough at this size: the document set is small enough to see all the way through. We write it up, and we build a test set — real questions from your staff, each paired with the document that ought to answer it.
You keep
The written assessment and the test set — a testable definition of what working means for you.
Contact
Tell us what is slowing your team down.
One email is enough. Tell us where your team loses time and we will use the first call to find out whether it is worth investigating.
- 01
What to send
The process that is slowing you down, the documents involved, and roughly how often your team handles it. A short description is enough.
- 02
What happens next
We start with a focused call. If the problem fits, we agree what the free audit should answer before looking at your operation.
Scope and price are agreed before each paid stage. No automatic progression and no invoice before your team has tested the finished system for a week.
Opens a new message in your email app.