What challenges did Oldcastle APG face before AI implementation?

Process a high volume of proof-of-delivery ship tickets with minimal human intervention. Scale to 200,000–300,000 documents per month. Handle rotated pages and variable formatting. Improve extraction accuracy and add signature validation. Provide real-time visibility into outstanding PODs and deliveries.

What was the impact of the AI solution?

Manual processing was reduced and automation increased. Extraction accuracy and reliability improved. Signature validation and rejection of incomplete documents were added. Real-time visibility into outstanding PODs and deliveries was provided. The process scaled to 200,000–300,000 documents per month. Processing cost was less than $0.04 per page.

Rank#1,554of 3,449

Source

Scaled productionEvidence: Low40/100

Oldcastle APG accelerates proof-of-delivery document processing with Amazon Textract and Amazon Bedrock

Use case typeIntelligent document processingUpdated Jun 13, 2026

Oldcastle APG, one of the largest global networks of manufacturers in the architectural products industry, was processing 100,000–300,000 ship tickets per month across more than 200 facilities. Its existing OCR system was unreliable, only reading 30–40% of documents accurately and requiring constant maintenance and manual intervention. AWS and Oldcastle APG built an event-driven document processing workflow to automate proof-of-delivery processing and improve visibility across deliveries.

Organization: Oldcastle APG
Industry: Manufacturing
Location: United States
Published: September 2025

Reported outcomes

Strategic outcomes

Speed & agilityAutomated proof-of-delivery processing workflowNew product / capabilityAdded signature validation and rejectionBetter decisions & insightReal-time delivery visibilityScale & capacityScaled document processing volume

Why do we believe this deployment?Customer identity, provider attribution, maturity, and source checks

Customer: Oldcastle APG
Provider: AWS
Maturity: Scaled Production
Linked source: AWS Machine Learning Blog

The process scaled to 200,000–300,000 documents per month

Customer identity supportedSource describes one deploymentMaturity supported

Primary read

Use case focus

Showing 3 of 3

1Intelligent document processing
2Proof of delivery automation
3Supply chain operations

Process a high volume of proof-of-delivery ship tickets with minimal human intervention.
Scale to 200,000–300,000 documents per month.
Handle rotated pages and variable formatting.
Improve extraction accuracy and add signature validation.
Provide real-time visibility into outstanding PODs and deliveries.

Amazon Simple Email Service receives ship tickets by email.
Amazon S3 Event Notifications trigger an event-based workflow and auto-scaling compute orchestration.
Amazon Textract StartDocumentAnalysis with Layout and Signature features extracts document text and signatures.
A microservice normalizes rotation issues and generates markdown from the extracted text.
Amazon Bedrock extracts key-value data from the markdown output.
Results are stored in Amazon RDS for PostgreSQL for downstream visibility and validation.

Technologies

Amazon Textract Amazon Bedrock Amazon Simple Email Service Amazon S3 Event Notifications Amazon RDS for PostgreSQL

Manual processing was reduced and automation increased.
Extraction accuracy and reliability improved.
Signature validation and rejection of incomplete documents were added.
Real-time visibility into outstanding PODs and deliveries was provided.
The process scaled to 200,000–300,000 documents per month.
Processing cost was less than $0.04 per page.

Architecture

Event-driven workflow using Amazon SES, Amazon S3 Event Notifications, Amazon Textract StartDocumentAnalysis with Layout and Signature, a markdown-generating microservice, Amazon Bedrock for key-value extraction, and Amazon RDS for PostgreSQL persistence.

Sources & evidence1

Evidence: Low40/100Evidence strength