A complete example

Background

Spent 5 years on the backend team of a Sydney freight-forwarding software company and knows customs declarations, commercial invoices and ICS entry well. Now helps 5–20 person businesses connect document handling to AI: test accuracy on real samples first, then decide which fields AI fills in and which stay with a human reviewer.

AI case study · 6-person freight forwarder: sorting quote-request emails and drafting quotes

Situation

The team received 50–70 quote requests a day. Two operators read each email, checked the rate sheet and wrote the quote by hand, so first replies often went out in the afternoon.

What was done

4 sessions, 5 hours in total. Session 1: went through 30 past emails and defined 5 request types. Session 2: used Claude to classify emails and extract fields, then tuned the prompt against the operators’ own answers. Session 3: wrote the results into the Google Sheets rate sheet to produce draft quotes. Session 4: trained the operators to review and edit the drafts.

Afterwards

Two weeks after go-live, the client’s own count showed first replies dropping from about 4 hours on average to under 1 hour. Operators still approve every quote before it is sent.

Limits and prerequisites

The client needs to be able to export emails and the rate sheet. Handwritten attachments and poor-quality scanned PDFs still need a person. Results depend on email formats and data quality.

Open slots

Pick a slot to see the price