Amazon’s AI Project PI looks for product defects before shipping

The principle of operation is that products on their way to customers pass through a tunnel that scans them. An image recognition computer program – a type of artificial intelligence that analyzes images and understands what is in them – checks for damage. If it finds something, that item is isolated and the system evaluates the defect and determines if there is a problem with similar items to track down the root cause.

According to Amazon, Project PI operates in “several” warehouses in North America and will be added to more locations throughout the year. Last year, Amazon implemented a different system that flags frequently returned products to highlight those that tend to have problems before customers actually press the order button. This all underscores the mode of avoidance a potentially “nightmarish” return process is good for customers, Amazon and the environment in terms of reducing greenhouse gas emissions.

The company says Amazon employees review items marked by Project PI to decide whether they will be sold at a discounted price on Amazon Second Chance Sales Site or communicated in another way.

Amazon is also working to introduce a multimodal, multilingual model to investigate why customers are dissatisfied with the products they receive. The AI ​​tool looks at what customers are saying in their reviews, then scans images from Project PI and other data sources to find out where something went wrong. Amazon says this technology could be helpful for other sellers to know if they have accidentally mislabeled products.