Industrial Distributor Shipping Paperwork, Automated with UiPath

The client is a specialty distributor in the US. Its range is valves, pipe and flow-control products, sold for industrial maintenance. It runs over 50 stocking locations and employs about 1,000 people.

Paperwork for each inbound shipment came from two sources: the vendor and the delivery partner, UPS for example. It covered the bill of materials, the invoice and purchase order, and the documents for transport and delivery. All of it was scanned into a single multi-page PDF. Matching each invoice against its purchase order and the delivery, then paying it, had to wait until someone taken the scan apart, checked it and keyed the missing data into Infor, the company's ERP.

Why the manual process hurt

• Six documents in one file. Up to five or six separate documents per scan, spread over many pages, in no particular order.
• Gaps found by looking. If the bill of materials or a delivery document is missing, the batch is incomplete. Spotting that manually takes time.
• Payment without confirmation. Each shipment needed three to four signatures: some to approve transport, some to confirm receipt. If they were missing, the company could end up paying for goods no one had signed for.
• Incomplete ERP records. Transportation and invoice records in Infor had holes until a person keyed in the data from the paperwork.
• Reconciliation stuck. Until all of this was done, no invoice could be checked against its purchase order and delivery data.

Paperwork for every shipment now arrives in Infor as separate, verified documents with full data. People only handle the cases the robot cannot resolve by itself.

• Scans holding several documents are separated, verified and posted to Infor with no manual work.
• Before reconciliation, each shipment is checked for every required document and signature.
• Missing transportation and invoice data is filled in from the paperwork, and people only see the uncertain cases.
• Invoices are reconciled with delivery data and purchase orders.
• The client kept no record of volumes or savings, so we quote none.

What teams usually overlook

Most document AI projects are tested on tidy, one-document files. Real shipping paperwork is one scan of five or six documents from several senders, in random order, some pages rotated, some signed by hand. Everything downstream depends on splitting that batch correctly. And a signature being there does not prove that the right person signed in the right spot. We built batch splitting first, followed by the completeness and signature checks, and routed every uncertain case to Action Center. That is what stops bad data from reaching the ERP.

Two related projects: reconciling EnergyLink invoices for an oil producer in the US, which also uses UiPath and Action Center, and invoice processing with AI, which splits, reads and posts scanned batches of 200 mixed documents.