Work · Purchase order automation · 2026
KiDaFlow
Scanned purchase orders read automatically and dropped into the fulfilment sheet, for a Kenyan tea distributor.
The outcome
2 hrs → 4 min
Per-order processing
Stack
- n8n
- OpenAI Vision
- Claude
- Google Drive
- Google Sheets
System
Purchase order automation
Year
2026
01 · The context
Built through KiDaFlow for a tea distributor that supplies retailers, hotels and corporate canteens across Kenya. Most LPOs (Local Purchase Orders) arrived as scanned PDFs over email or WhatsApp, and the ops team typed them into a fulfilment spreadsheet by hand.
02 · What was breaking
Each order took about two hours of someone's attention: open the PDF, read the line items, type them into the right columns, check the prices, send it to dispatch. At 30 orders a week, that was a full-time job spent on data entry.
03 · The approach
- 01The scanned PDF goes into a watched Drive folder.
- 02Two AI models read the line items, quantities, prices and customer details, and a third pass checks their work.
- 03An n8n flow checks the structure, flags anything odd for a person to review, and writes the order into the fulfilment sheet.
- 04Dispatch picks up the row as if someone had typed it.
04 · Shipped
Live in production. Handles English and the occasional handwritten note. Every order gets a confidence score, and anything below the line goes to a review queue instead of straight to fulfilment.
Note · Runs on the client's own accounts. No vendor lock-in.
05 · The receipts
2 hrs → 4 min
Time per LPO
11%
Human review rate
30+
Hours recovered per week
More work
Same shape of problem. Different industry.
Next
Got a workflow like this you want shipped?
The Automation Opportunity Audit takes 90 seconds and tells you which three workflows in your business would pay back fastest.