Meet Cathy
Your AI freight operations manager.
From enquiry to invoice, Cathy knows the shipping tasks. Ask her anything, anytime — she checks your records, calls the right engines, and does the work: the quote drafted, the update written, the invoice prepared.
Works inside Catupult Forward and Catupult Sellarate.
The container completed its voyage and is in empty return at Rotterdam.
What she actually is
Most AI tells you about your operation. Cathy runs it.
An assistant that answers questions still leaves the work with you. You asked where the container was, you got an answer, and you still have to write the email.
Cathy completes the task and hands you the finished work. The quote is drafted and attached. The update is written. The invoice is prepared and the discrepancy is flagged. What's left for you is the decision, not the typing.
| An assistant | then | An operator |
|---|---|---|
| An assistant Tells you the rate | An operator Builds the quotation | |
| An assistant Summarises the document | An operator Files it against the shipment and updates the job | |
| An assistant Reports the delay | An operator Drafts the customer update and annotates the file | |
| An assistant Shows you the discrepancy | An operator Prepares the invoice with it flagged |
What she does
Everything a desk does, except decide.
- Sales
- Creates enquiries from a message or a forwarded email. Compares rates across contracts, spot and surcharges. Prepares quotations priced to your margin rules.
- Operations
- Opens orders, bookings and jobs. Captures carrier confirmations. Watches containers, milestones, free time and exceptions.
- Documents
- Reads booking confirmations, vendor invoices, shipping instructions, rate contracts and KYC documents. Generates SI, HBL, AWB and invoices on your letterhead.
- Finance
- Matches vendor invoices line by line against the job. Flags what doesn't reconcile. Prepares customer invoices.
- Analysis
- Answers from your records — lane margin, customer volumes, what's open — with every number traced to its source.
She doesn't wait to be asked twice.
Standing instructions run on their own: chase the carrier, flag the delay, prepare the invoice, tell the customer.
Cathy isn't about replacing your people. She handles the work that shouldn't need a person, so your people can do the work that does.
No guesswork
The AI understands the question. Your data answers it.
Cathy's facts come from one place: your records, and the engines that serve them. The language model's only job is to work out what you're asking for and which part of the system to use. It never supplies the answer from its own training.
No general knowledge. No web lookup. No plausible-sounding guess. If a rate isn't in your contracts, there isn't a rate — and she tells you that instead of producing one.
She can't make something up, because she was never the source.
Intent, not knowledge
The model reads what you meant. Everything it then does runs against your data.
Engines, not memory
Rates, schedules, tracking, documents and finance each answer from their own system of record.
Missing is missing
When the data isn't there, she says so. No estimate dressed as a fact.
Control
She inherits the role she's acting for. Nothing more.
Cathy doesn't have her own access. She operates as the person who asked, with exactly their permissions. If your documentation clerk can't approve an invoice, neither can Cathy on their behalf. If a branch manager only sees their own branch, that's all Cathy sees when working for them.
Every action is also checked against the role before it runs — not filtered from the screen afterwards.
Nothing irreversible happens without a confirmation.
Anything that leaves the building — a quotation to a customer, an order to a carrier — stops and asks. You get a confirm-or-cancel card showing exactly what's about to happen. Confirm it, cancel it, or leave it to expire. All three outcomes are recorded.
Which actions need a signature, and at what value, is yours to set.
AI assistants that auto-send things cause incidents.
This one asks first. Every time.
Every action, with the evidence behind it.
Each thing Cathy does is logged with what she was asked, what she used, which records she touched and when. Not a chat transcript — an operational record you can reconcile against the shipment.
If a quote went out at the wrong number, you can see which rate was selected, which margin rule applied, and who confirmed it.
Every number traces back to where it came from.
A rate traces to the contract. A milestone traces to the carrier feed. A charge traces to the document it was read from. And quotations are priced from a recorded rate selection, so a quote can still be explained months later — even after the rate book has moved on.
We deleted our own analytics screen because it was built on hardcoded numbers. Everything you see now reads live from your data.
She works on your live records.
Not a copy, not a nightly export, not a separate index that drifts out of date. Cathy reads and writes the same records your team is looking at, which is why what she tells you matches what's on the screen.
The limits are the point.
- Won't send without asking
- No quotation, order or customer email leaves without a confirmation.
- Won't invent a number
- The model interprets your request; your records supply every fact. If the data isn't there, she says so.
- Won't exceed the role
- She can't do anything the person she's acting for couldn't do themselves.
- Won't hide her working
- Every output traces to the record, document or feed behind it.
An operator you can't audit isn't an operator. It's a liability with a friendly interface.
Watch her work before you decide anything.
A live demo shows the workflow end to end. Then evaluate it with the people who'd be accountable for it — operations, finance and IT — in a workspace of your own.