FIRSTVAL

Automating the Bill of Lading

90,000 bills a day, read in seconds instead of minutes.

FirstVal customer — global shipping data company

Shipping & freight documentation

80%
Fewer documentation errors
8x
Productivity per operator
6x
Cost saving per document

01 — The intro

Where the story starts

A bill of lading is the document that holds a shipment together. It is the receipt that goods were loaded, the contract of carriage, and the title to the goods themselves.

Automating it means pulling the key facts off the page and into the payables system fast enough to clear approvals and payments. Traditionally this has been paper-driven work — slow, manual, and error-prone.

Daily volume
90,000

Bills handled across operating centres before automation.

Manual effort
4,500+ hours

Daily documentation and verification effort at roughly 20 minutes per bill.

Quality exposure
4%

Reported manual documentation error rate before model-assisted validation.

02 — The challenge

Read bills of lading in any format, extract every relevant fact accurately, and do it at a cost and speed that scales.

  • Close to 90,000 bills of lading processed every day across operating centres.
  • Each bill carries multiple line items — roughly a million item-level records to document.
  • 4,500+ man hours consumed daily, around 20 minutes per bill, across a regional workforce.
  • A reported error rate of around 4% during manual documentation.

Improve

Throughput and accuracy

Reduce

Cost per document and rework

03 — The solution

A document intelligence pipeline that ingests any layout, extracts named entities, reads context, and validates itself before storage.

  1. 01Engineered an extraction pipeline that accepts documents across formats after annotation.
  2. 02Trained ensemble models specifically on bill of lading structures.
  3. 03Named-entity extraction for names and addresses, PO number, instructions, pickup date, item number and attributes, packaging type, freight class.
  4. 04Context analysis for free-text notes and handwritten instructions.
  5. 05Auto-validation and storage on cloud with audit trail.
  6. 06Model optimisation so scans run on a phone in the yard.
  7. 07Micro-services architecture for scale, with PII handled at every hop.
  8. Impact

    1,000 bills of lading processed in under three minutes.

04 — Value logic

How operating change reaches financial value

01

Straight-through extraction

Documents move from scan to structured fields without re-keying.

Less processing labour per bill and greater volume capacity.

02

Context validation

Entities, instructions and line items are checked before storage.

Lower correction cost and fewer payment delays.

03

Mobile inference

Teams capture documents at the point of work.

Shorter cycle time without adding fixed terminals or handoffs.

05 — Feature delight

What people actually felt

01

Any format, no rules

Varied layouts handled by the same models — the customer was freed from maintaining brittle rule-based templates.

02

Mobility with AI

A light but accurate model that runs on mobile, so a scan can happen wherever the paper is.

06 — Ground impact

What moved, and by how much

6x

Cost saving

Cost per person for documentation, processing and validation.

90 days

First pilot

Then 12 locations live in under six months.

80%

Error reduction

Models fine-tuned against the human error baseline.

7 sec

Per bill

Extraction, processing and documentation, down from 20 minutes.

07 — Measurement

How the result stays accountable

Speed

Compare the documented 20-minute manual cycle with the seven-second automated processing time.

Quality

Track exceptions requiring human correction against the reported 4% manual error baseline.

Capacity

Measure bills completed per operator and cost per completed document at each rollout location.

08 — CSAT score

Scored by the customer, A+ to D

A responsive, intuitive document AI solution delivered against a measurable financial outcome.

A+
AI model and deployment pipeline
A+
Customer service and PII data management
A
User experience
A
Overall impact