Georgia-Pacific AWS AI Vision Saves $1M per Machine with 90-Day Deployment

Title

Georgia-Pacific AWS AI Vision Saves $1M per Machine with 90-Day Deployment

Content

Georgia-Pacific saved up to $1 million per machine by deploying machine learning for condition-based predictive monitoring of paper mills on AWS, with the journey starting a couple of years ago in shop-floor environments and now letting the company predict equipment failure 60 to 90 days in advance for selected assets and eliminating 40 percent of parent-roll tears on one converting line, with at least 150 converting lines seen as candidates for the same approach. At Hannover Messe 2026, Steven Blackwell, Head of Product Engineering & Services Center of Excellence at AWS, described how cloud infrastructure helps manufacturers transform across engineering, operations, supply chain, and customer-facing business models, and walked through three concrete deployments tied to Georgia-Pacific and the broader manufacturing market. The first is the predictive monitoring program itself, which combines ML on AWS IoT infrastructure with operator workflows; the second is GP Chat, a maintenance chatbot built on Amazon Bedrock that combines real-time IoT sensor data with operator queries to give maintenance engineers immediate access to all available knowledge for diagnosing problems; and the third is a defect-detection workflow using Amazon Nova, which AWS describes as a zero-training approach in which manufacturers define defect detection criteria through natural language prompts and the system compares a reference image with an image from the actual production line, removing the need for large labeled training data sets and allowing fast iteration across product lines. The interview, conducted by Lucian Fogoros of IIoT World at Hannover Messe 2026, also pointed to agentic AI as the next step for addressing the manufacturing skilled labor shortage, with engineers becoming citizen developers who deploy agents in their own ecosystems rather than waiting on central ML teams; Blackwell framed AI as evolving from traditional machine learning into agentic AI, and argued that the speed of adoption — not just the technology — is what manufacturers most need to change. Looking across AWS's manufacturing customers, Blackwell positioned Amazon Nova as a generalizable pattern for shop-floor computer vision while GP Chat serves as the on-shift knowledge interface for operators, with both expected to expand across Georgia-Pacific's converting line footprint as the predictive monitoring program matures.