The design problem
Relay Freight is a fictional freight forwarding company. In this concept, its operations team tracks shipments in shared spreadsheets, coordinates through email, and re-keys details from delivery documents into its accounting system.
That approach creates predictable problems:
- No single view of status. Managers ask colleagues for updates because the spreadsheet is rarely current.
- Exceptions surface late. A missing customs document or a delayed handover is noticed only when a customer asks.
- Manual data entry. Details from delivery notes and invoices are typed in by hand, which is slow and error-prone.
Proposed solution
The concept is a focused internal web application, not a full transport management system. It covers the steps that cause the most friction and connects to the tools the business already uses.
Shipment pipeline
Shipments are shown by status: booked, in transit, exception, and delivered. They can be filtered by lane, customer, and owner. Each shipment has a detail view with its timeline, documents, and notes, replacing scattered email threads.
Exception queue
Rules flag shipments that need attention, such as missing documents before a cutoff or no status update within an expected window. Each exception has an owner and a resolution note, so nothing depends on memory.
AI-assisted document intake, with human review
When a delivery note or invoice is uploaded, an AI step extracts key fields such as the consignee, weights, and reference numbers. The extracted values are shown for review, not saved automatically. A team member confirms or corrects them, and low-confidence fields are highlighted. Corrections are logged so extraction quality can be evaluated over time.
Integrations
Approved records sync to the accounting system. Carrier status updates arrive through APIs or webhooks where carriers provide them. Every sync is logged, and failures alert the shipment owner.
Features demonstrated
- Role-based internal application with an auditable timeline
- Rules-based exception detection and ownership
- AI extraction with confidence thresholds and mandatory review
- Integration design with retries, logging, and alerts
- Dashboard visuals that use clearly labeled sample data
Notes on the sample data
All figures, shipment IDs, and routes in the illustrations are fictional sample data. They do not describe a real company or real performance.