Purchase Order AI for a Steel Distributor
Maharashtra, India · Metals & Steel
Overview
StancorIQ reading purchase orders arriving in 40+ customer formats and posting them into ERP, replacing manual data entry across a 9-person order desk.
The challenge
Orders arrived as PDFs, scans, emails and spreadsheets in over 40 layouts. Nine people keyed them in, order acknowledgement took 6-8 hours, and roughly 4% carried a transcription error that was only discovered at dispatch.
What we did
Extraction was trained on the 12 formats covering 80% of volume first, with the long tail routed to a review queue rather than forced through. Fields are validated against the customer master and price list before posting, so an error is caught at entry rather than at dispatch. Acknowledgement time fell to 11 minutes and the order desk moved to exception handling.
Project data
| Formats handled | 40+ (12 trained, rest via review queue) |
|---|---|
| Acknowledgement time | 6-8 hours → 11 minutes |
| Extraction accuracy | 98.4% on trained formats |
| Transcription errors | 4.0% → 0.3% |
| Team redeployed | 9 → 3 on exceptions |
| Payback | 7 months |
Standards applied
- ISO 27001
- SOC 2 Type II
- DPDP Act 2023
Scope supplied
- StancorIQ
- StancorERP
Related
- StancorIQ
- StancorERP