Shipping Document Processing with UiPath for an Industrial Distributor

We built a UiPath robot that turns each scanned batch into complete, checked records in Infor. UiPath Document Understanding reads the documents; UiPath Action Center lets people complete what the robot cannot.

From one scan to clean ERP data

1. Split the batch. Document Understanding classifies the pages of each multi-document PDF and splits it into its five or six separate documents.
2. Check completeness. The robot verifies that every document the shipment needs is present, including the bill of materials.
3. Extract the data. Document Understanding extracts the fields needed from each document type.
4. Check the signatures. The robot looks for each of the three to four required signatures, transport approval and receipt confirmation among them, and checks them against what the shipment needs.
5. People fill the gaps. A missing or uncertain field goes to Action Center for a person to complete or confirm.
6. Into Infor. The complete, validated data is entered into the web-based Infor ERP.
7. Reconcile. Invoices are reconciled against purchase orders and transportation data, and missing data is added to Infor from the documents.

Technology

Each shipment's paperwork now reaches Infor as separate, checked documents with complete data, and people see only the cases the robot cannot settle on its own.

• Multi-document scans split, checked and entered into Infor automatically.
• Every shipment checked for all required documents and signatures before reconciliation.
• Gaps in transportation and invoice data filled from the documents, with people only on the uncertain cases.
• Invoices reconciled with purchase orders and delivery data.
• The client did not record volumes or savings, so we do not quote any.

The risk most teams miss

Document AI projects are usually tested on clean, single documents. Real shipping paperwork is a scan of five or six documents from different senders, in any order, some rotated, some with handwritten signatures. Splitting the batch correctly is the step everything else depends on, and a signature that is "present" is not the same as the right person's signature in the right place. Building the split, the completeness check and the signature check first, with Action Center for every uncertain case, is what keeps bad data out of the ERP.

Two related builds: EnergyLink invoice reconciliation for a US oil producer, also on UiPath with Action Center, and AI invoice processing, where scanned batches of 200 mixed documents are split, read and posted.