AI engineering · Vehicle import, distribution and logistics · India
Five document jobs, one way of working.
People opened documents one at a time, read them on screen, and typed what they found into spreadsheets. We built one system that reads those documents, applies each team’s own checks, marks anything that looks wrong, and produces the file they already use.
At a glance
- Teams
- Operations, logistics and finance
- Status
- Built and demonstrated. Not live yet.
- What we built
- Five document jobs on one extraction and checking pipeline
- Evidence
- Everything below was observed while the system ran. No production measurements exist yet.
What changed
What changed
Before
Three departments, each with its own manual routine.
Now
Documents go in once. Only the exceptions come back to the team.
A number can be traced back to the document it came from.
Before
Important checks depended on someone remembering to do them
Documents arrived by email, in shared folders, on memory sticks, and by hand. The tools were ordinary ones: a PDF reader, spreadsheets, the annual tariff, and two finance templates.
Operations: typing up Bills of Lading
Staff opened Bills of Lading one at a time, retyped the details into a spreadsheet, and checked each vehicle’s VIN by counting the characters on screen.
Logistics: comparing documents
Documents for the same shipment were read side by side, which made disagreements slow to find and hard to record the same way twice.
Logistics: checking freight bills
Every line on the carrier’s bill was checked by hand against the annual tariff, which left room for wrong rates and charges nobody recognised.
Finance: the same bill, typed twice
Supplier bills arrived in different layouts, so the same details were typed out twice. Once into the finance system’s import file, once into the payment voucher.
All teams: getting attachments out of email
Attachments had to be saved out of email, unlocked, renamed and filed before any of the work above could start.
The shared problem
Slow review, the same details typed again and again, avoidable mistakes, results scattered across files, and little way to trace a number back to its document.
What was at stake
None of it looked urgent on the day. The work was routine and any one mistake was small. What made it worth taking seriously is what a small mistake turns into three weeks later.
- A shipment held up: a discrepancy nobody caught can delay a shipment being cleared.
- A charge nobody questioned: a rate that does not match the annual tariff gets paid, because nothing in the process forced the comparison.
- No way back to the source: if somebody questioned a row later, the filename was often the only route back to the document.
The method
One approach, reused
The system reads the document, pulls out the details, applies that team’s rules, marks anything that looks wrong, and produces the file they already use. Each of the five jobs runs all five steps on its own.
01
Send them in. Staff upload a batch of documents, or the system picks the attachment up from an email.
02
Read them. The system works out what each document is and pulls out the details that job needs.
03
Apply the checks. Rules check VINs, compare records against each other, and test charges against the annual tariff.
04
Mark the exceptions. Anything missing, anything that disagrees, and anything the system was unsure about is flagged clearly.
05
Produce the file. A structured spreadsheet is generated and sent to the people who need it.
The hard part
Reading the document was the easy part
Software has been able to pull text off a scan for years. The difficulty is that each of the five jobs has its own idea of what a right answer looks like: a VIN that is exactly 17 characters; the same value appearing on all four shipment documents; a charge that matches the annual tariff; one supplier bill mapped correctly into two different finance templates; an attachment pulled out of an email, unlocked, and named from what is inside it.
And the documents do not cooperate. Scanned and digital documents varied in quality and layout, carried stamps and handwriting, and repeated line items down the page. The same detail could appear under a different label from one document to the next.
What we built
The five jobs
01
Bill of Lading extraction
The whole batch is read in one go. Every VIN is checked against the rule that it must be exactly 17 characters. Sample run: 9 documents read, 6 valid VINs, 1 trimmed to 17, 1 short, 1 needs review.
02
Comparing documents
The same fields from the Bill of Lading, commercial invoice, packing list and certificate of origin go into one table. Sample: eight fields across four documents, three disagree, two missing. Those five are what the team looks at.
03
Checking freight bills
The tariff is loaded once and every bill line is matched against it. Sample: 8 lines checked, 4 correct, 4 charged too much, 8.18 over in total. None large on its own, which is exactly why they used to get through.
04
One bill, two finance files
The bill is read once and both the finance import file and the payment voucher are filled from the same rows. Sample: 2 bills, 5 lines, 0 failed.
05
Email attachment intake
The system takes the email, finds the attachment, unlocks it where allowed, names it from its contents, and files it so the next job can find it without anybody renaming anything.
The outcome
Demonstrated, and reusable
Validated through implementation
- All five jobs ran from start to finish
- Documents uploaded through the browser
- Reading and comparison produced structured results
- Each job generated the spreadsheet it needed
- Finished files sent out by email automatically
What it makes possible
- Repetitive typing replaced by reviewing the exceptions
- VIN, tariff and cross document checks done the same way every time
- Finance fills two templates from one reading
- A number traced back to its document
- The same approach extended to other paperwork heavy teams
Beyond this client
The job transfers
The same problem turns up well beyond this one client. These are the teams it fits.
Anyone typing documents into spreadsheets
- Freight forwarders and customs brokers
- Importers and distributors
- Accounts payable teams
Anyone checking bills against a rate card
- Logistics and transport operators
- Procurement teams
- Any team reconciling supplier invoices by hand
Next case study
Talent Carriage · AI engineering
The voice agent that writes the ticket while you talk
Facing something similar?