One Shipment,Many Stakeholders
How a freight forwarder processes shipment documents the moment they arrive, each routed to the right system in near real-time.
Each document is processed the moment it arrives: no waiting for the full set. Four systems update in near real-time as documents come in. Zero re-keying, zero delays.
A freight forwarder processes international shipments. Each shipment involves multiple documents (commercial invoices, packing lists, bills of lading, certificates of origin, customs declarations), but they don't arrive together. They're generated at different points in time by different parties: the shipper sends the commercial invoice first, the carrier issues the bill of lading days later, customs declarations follow separately. Four departments need different data from these documents, and each needs it as soon as the relevant document is available, not after the full set is assembled. Data entry clerks currently re-key the same data into four separate systems, often delayed by waiting for documents that haven't arrived yet.
Ingestion
Documents arrive individually as they're generated: shipping agents drop files via FTP/SFTP, and the logistics platform pushes them via S3. Each document enters the pipeline the moment it's available, not when the full shipment set is complete. OCR makes every format machine-readable on arrival (bill of lading OCR for the carrier's scans, clean text from digital PDFs), so freight document processing starts the moment a file lands.
Classification
Four Multi-Structure Classifiers evaluate each incoming document in parallel, one per stakeholder concern. Each classifier matches against multiple tables and structures within the document. The customs classifier needs to find both HS code tables and declared value headers. The warehouse classifier looks for item tables with quantities, weights, and dimensions. The billing classifier matches charge tables and surcharge summaries. The tracking classifier identifies routing tables with vessel and port data. A document only matches a classifier when all its required structures are present, not just one table.
Extraction
Each classifier feeds its own Extractor, which pulls data from the multiple tables identified during classification and maps it to the format the downstream system expects. The customs extractor reads HS codes and declared values and maps them to the customs declaration API schema. The warehouse extractor combines items, quantities, and dimensions into the warehouse management system's inbound format. The billing extractor maps freight charges and surcharges to the invoicing system's line-item structure. The tracking extractor formats vessel, port, and ETA data for the tracking portal's API. Each extractor produces a ready-to-send payload: no transformation needed downstream.
Delivery
The pipeline delivers each extraction directly to its target system in near real-time: customs declaration system, warehouse management, billing, and customer-facing tracking portal. The moment a bill of lading arrives, the tracking portal updates. When a commercial invoice lands, customs and billing get their data. Each department acts on information as it becomes available, not after waiting for the complete document set.
Multi-Table Classification
A document is only classified as a match when multiple tables and structures are found within it; a single table match isn't enough. The classifier validates the document's full structure.
Cross-Table Extract & Map
Each extractor pulls data from multiple tables within the same document and maps it to the target system's API schema; the downstream system receives a ready-to-send payload.
Per-Extraction Sink Routing
Each extractor delivers its output to a dedicated sink. Every stakeholder gets exactly the data they need, from the same source documents.
Near real-time, per-document processing: documents are classified and extracted individually as they arrive, with each stakeholder system receiving its data immediately rather than waiting for a complete bundle.
Freight Questions
What logistics teams ask about freight-document automation.
Yes, the moment they arrive. OCR makes every format machine-readable, from scanned bills of lading to digital PDFs, and extraction fills the shipment schema while the truck is still at the dock. Delivery notes, customs documents and freight invoices follow the same path.
Near-real-time: each document is processed as it arrives, not when the batch completes, and the extracted data is delivered to four systems in parallel: billing, tracking, customs and warehouse management.
See a Pattern That Fits Your Workflow?
Every pattern is built from the same modular components: classifiers, extractors, validators, and connectors. Let's find the right combination for your documents.