The challenge
Vitalon receives PDF, faxed, scanned, and handwritten orders from different sales channels. Staff need to read each document and interpret its business meaning: a number may represent cases, units per case, or the total quantity. Organizing those details field by field takes time before an order can enter the next stage of processing.
Reported reduction in the time needed to process orders after the AI workflow was introduced.
Our approach
We deployed an AI document recognition workflow that extracts the ordering party, address, product code, product name, case quantity, units per case, total quantity, and unit of measure separately. Staff receive structured fields they can review before connecting the order to ERP operations. Uncertain or handwritten content remains easy to check against the original document.
What we delivered
- A document recognition workflow for PDF, faxed, scanned, and handwritten orders from different sales channels.
- Separate fields for ordering party, address, product code, product name, case quantity, units per case, total quantity, and unit of measure, with the original document available for review.
- Template mappings, section exclusions, and review rules that prepare confirmed order information for ERP processing.
The details we handled
During implementation, a sample review covered five representative documents and 183 core fields. It surfaced quantity-field confusion, information captured from quality inspection sections, and notes mixed into product names. Those findings guided template mappings, section exclusions, and review rules so the extracted data matched the needs of actual order processing.
- Numbers must be interpreted according to their business meaning: case quantity, units per case, and total quantity cannot be treated as interchangeable fields.
- Handwriting and differing layouts make field boundaries uncertain. Quality inspection sections and notes can be mistaken for order data or become part of a product name unless extraction and review rules distinguish them.
What changes
The AI workflow reduces Vitalon's order-processing time by 90%. Different document types now arrive for confirmation in a common structure, allowing staff to focus on fields that need attention and move confirmed orders forward. Formats from additional channels can be incorporated as the team continues refining the workflow.
How to verify delivery
Acceptance checks for a comparable implementation. Agree on representative inputs and acceptance criteria with your team before verification.
- Review representative documents from each supported format; compare the extracted ordering party, address, product code, and product name with the original document.
- Check case quantity, units per case, total quantity, and unit of measure separately; confirm that inspection content and notes do not populate order fields incorrectly.
- Include handwritten or unclear fields, resolve them through source-document review, and compare the confirmed fields with the data prepared for ERP processing.
Every thoughtful implementation starts with understanding the work.
