AI System for Checking Construction Material Delivery Documents

The problem

A delivery may contain the correct product but have a quantity or document missing. When purchase orders, delivery notes, products, quantities, lots and conformity certificates are scattered across different screens, documents or team records, issues do not become visible in time.

This fragmented data delays preparation of the material receipt discrepancy report and requires staff to search for supporting records again by hand. Before making a decision, the store manager and quality engineer have to complete missing fields and verify conflicts one by one.

A common exception is a partial delivery or an approved equivalent. If an incomplete or incorrect workflow runs without checking this context, non-conforming material may be accepted or a valid delivery rejected; the output therefore cannot be applied directly.

What we set out to improve

• Number of verified delivery discrepancies
• Rate of outputs corrected or rejected by the store manager and quality engineer
• Time between the source record and the review output

How the system works

1. 01 Define the review scope and the authorised store manager and quality engineer roles
2. 02 Receive the inputs: purchase order, delivery note, product, quantity, lot and conformity certificate
3. 03 Check source identity, date, version and data freshness
4. 04 Match the purchase order with the delivery note
5. 05 Check the quantity, lot and certificate
6. 06 Check the exception: there may be a partial delivery or an approved equivalent
7. 07 Place low-confidence or conflicting records in a separate human review queue
8. 08 Write the source identity, rule, output and time to the audit trail
9. 09 Have the store manager and quality engineer approve, correct, defer or reject the report

Methods we used

• Document and product-code reconciliation
• Source, format, version and required-field checks for the purchase order, delivery note, product, quantity, lot and conformity certificate
• A missing-data warning instead of a prediction when inputs are absent or conflicting
• A separate check for the exception outside the main rule
• Joint storage of the source identity, rule result, material receipt discrepancy report and human decision

Where people stay involved

Data and security

Access to purchase orders, delivery notes, products, quantities, lots and conformity certificates is restricted to the minimum permissions required for each role. Identifiers and sensitive fields are masked where possible, and only the necessary section of a file is sent to the model rather than the complete raw file. Access, output and the reviewers' decision are logged; production use does not begin until the organisation has approved the retention period and data location.

Who this suits

Frequently asked questions