From 185,000 manual stamp entries
to autonomous extraction
JTI Ukraine manually extracted data from 185,000 tobacco excise stamps per year for tax refund filing. A Nanonets agent now monitors Google Drive, splits stamp files, deduplicates, extracts all fields, and exports structured data end-to-end.
Japan Tobacco International (JTI) is a leading international tobacco company and a member of the Japan Tobacco Group. JTI Ukraine holds a 22% market share in the Ukrainian tobacco market, selling iconic brands including Winston, Camel, and Sobranie.
The Problem
JTI Ukraine is required to submit excise stamp data to Ukrainian tax authorities annually to claim tax refunds on tobacco products. Each of the ~185,000 stamps requires three fields to be manually entered: Date, Number, and Series.
- —555,000 manual data points entered into Excel, one stamp at a time.
- —24+ week turnaround from stamp collection to tax refund approval.
- —Error-prone process with no deduplication — duplicate stamps went undetected.
The setup was seamless with Nanonets team guidance. The system interface is intuitive and customer support is top-notch.
The solution
One orchestrator agent. Zero manual touchpoints.
Nanonets replaced the manual loop with one autonomous agent. The agent orchestrates five sub-agents which each own one step of the pipeline — monitoring Drive, splitting stamps, detecting duplicates, extracting fields, and exporting structured data — with no human in the loop.
Pipeline overview
Impact
AGENT ANATOMY
See it run on your process, with your documents.
Start free. No credit card. Or talk to our team about your workflow.