Catching Defects Before the Waste Piles Up
The platform, called the Visual Inspection Cockpit, targets a problem that has long stumped conventional machine vision. Diapers, wipes and other textured consumer goods shift, stretch and wrinkle at production speed, and rule-based cameras had to be reprogrammed whenever a material, package design or line setting changed. P&G’s deep learning models handle those variations, and Siemens’ Industrial Edge hardware runs the inference next to the equipment so a defective product can be pulled from the line before more packaging, labor and machine time go into it.
“We engineered this solution to solve a myriad of industry challenges traditional vision systems couldn’t touch,” said Paul Thomas, P&G’s director of machine vision and applied AI. The companies said plant engineers can configure, train and update inspection models themselves, without a data science team on site.
Faster to Copy Than to Rebuild
We’d love to be your preferred source for news.
Please add us to your preferred sources list so our news, data and interviews show up in your feed. Thanks!
The commissioning speed is what turns a plant win into a corporate program. A bespoke vision system is engineered for one line and rebuilt for the next. A standardized AI model on shared edge hardware can be installed, trained on the new product and switched on in a fraction of the time, which is why P&G can replicate the tool across factories instead of funding a fresh integration at each one.
That approach fits P&G’s broader Supply Chain 3.0 program, which has moved into full scaling mode across the company, CFO Andre Schulten said on the April 24 earnings call, Supply Chain Dive reported. P&G is targeting up to $1.5 billion in cost of goods sold savings from the program, and it has piloted a fully automated four-hour night shift in Berlin. Schulten said each automated shift can deliver productivity gains of 15-60%.
Savings That Don’t Depend on Sales
Scrap reduction is margin that arrives without a price increase or a volume gain. For a consumer goods company facing tariffs, commodity costs and cautious shoppers, that is the kind of return an industrial AI project has to show to keep its budget.
Most manufacturers have not reached that point. The World Economic Forum’s Global Lighthouse Network, which recognizes factories that have scaled advanced technology, counted 238 sites worldwide and described its purpose as helping the rest of industry escape “pilot purgatory.” Analytical AI and machine learning accounted for about 62% of the solutions those sites ran in 2025, and the network’s latest cohort, named in June, was singled out for scaling across factory networks rather than single plants.
P&G’s scrap figures are the companies’ own and have not been independently verified, but the mechanism they describe—one model and one hardware platform repeated across plants—is what separates a factory experiment from infrastructure.
P&G has set 2030 as the target for completing Supply Chain 3.0, and Schulten framed what is left as a question of speed. “We know it works, and we know what to do,” he said on the April call. “It’s about how fast do we roll them out.”
Add as Preferred Source
