The challenge
The retailer received large volumes of supplier documents in different layouts, and staff keyed the data into the ERP by hand. Processing was slow, errors were common, and the workload grew with every new supplier.
Our approach
We designed for scale and accuracy from the start. Document layouts varied too much for fixed templates, so we combined OCR with AI interpretation. Every record is validated before it reaches the ERP, and anything uncertain is routed to a person.
The solution
- Azure Blob Storage to receive and store incoming documents
- Python Azure Functions that process each document using Azure AI OCR and the OpenAI API
- Azure Service Bus to queue work and handle volume spikes
- Azure SQL Database for extracted data, with validation and business rules
- Azure API Management to securely post validated records into the ERP
- Development accelerated with Claude Code
The outcome
Supplier data now flows into the ERP without manual capture. Processing time has dropped, errors are caught before they enter the ERP, and the solution scales with document volume rather than headcount.