01
Extraction tuned to freight reality
Vision-model extraction handled the crumpled photos and low-res scans legacy OCR choked on, with field-level confidence scoring on every value — shipment numbers, weights, charges, signatures.
Mid-size freight forwarder · Back-Office Automation
1,100 hours a year returned to operations
A freight forwarder processed thousands of BOLs, PODs, and carrier invoices a month — every one keyed by hand from a different format. We built an extraction pipeline that reads them all, matches them to shipments, and posts them to the TMS automatically.
- 1,100h
- staff hours saved per year
- 4 days→same-day
- POD-to-invoice cycle
- 99.2%
- field-level extraction accuracy
/ 01 — The problem
What was breaking.
Documents arrived as email PDFs, driver phone photos, and the occasional fax — in hundreds of carrier-specific layouts. Ops staff spent entire afternoons re-keying, and typos surfaced weeks later as billing disputes.
Invoicing waited on POD paperwork being manually matched to shipments, stretching DSO four days past delivery on average.
/ 02 — The build
How we approached it.
02
Match, validate, post
Extracted documents were matched to open shipments in the TMS, validated against expected values, and posted automatically. Mismatches queued for review with the exact discrepancy highlighted — a 30-second decision instead of a 12-minute hunt.
03
Fail-safe by design
Low-confidence extractions never post silently. The system's honesty about what it can't read is why the operations team came to trust what it does.
Timeline4 weeks to first document type live, 10 weeks to 90% coverageStackClaudeAzure Document IntelligencePythonTMS APIn8n
/ 03 — The results
What changed.
- 90% of document intake now processes without human touch; the remainder arrives pre-analyzed with discrepancies flagged.
- Roughly 1,100 staff-hours a year returned to customer-facing operations work.
- POD-matched invoices now go out same-day, pulling DSO in by four days and measurably improving cash position.
- Billing disputes from keying errors dropped to near zero — the audit trail resolves the rest in minutes.
“We assumed we'd need an EDI mandate our carriers would never accept. Instead the system just reads whatever they send — including the driver photos.”
/ Free AI audit
Get results like these in your business.
Every engagement here started with the free audit. Sennova Labs maps your workflows, prices the waste, and shows you what to automate first.
Request a free AI audit
No obligation · Roadmap is yours to keep