Amazon's Project PI combines AI and computer vision to detect damaged items before shipping. This new tech operates in several North American warehouses and aims to expand. Human reviewers decide if flagged items are resold or donated.
Amazon's Project PI, or "Private Investigator," combines generative AI and computer vision to detect product defects before shipping. This technology scans items as they pass through a tunnel, identifying damages or incorrect specifications such as color and size. If a defect is found, the item is isolated for further evaluation. Amazon plans to roll out Project PI to more warehouses throughout the year, currently active in several North American sites.
Project PI aims to enhance customer satisfaction by preventing the nightmare of returning damaged items. This initiative is not just beneficial for customers but also for Amazon and the environment. By reducing returns, Amazon cuts down on carbon emissions, contributing to a more sustainable operation. Last year, Amazon implemented another system that flags frequently returned items to preemptively address potential issues.
Even with advanced AI, human oversight remains crucial. Amazon employees review the items flagged by Project PI to decide if they should be sold at a discount on Amazon's Second Chance site or donated. This ensures that only quality products reach the customers, while defective ones are appropriately managed, adding another layer of quality control to the process.
Looking ahead, Amazon plans to introduce a multimodal large language model to further investigate customer dissatisfaction. This AI will analyze feedback and scan images from Project PI to determine the root cause of issues, helping sellers avoid mislabeling and other mistakes. Amazon's continuous innovation in AI and automation underscores its commitment to improving the customer experience and operational efficiency.
How do you feel about AI checking your Amazon purchases?
Each week we select most important sector news and statistic
so that you can be up to speed