There is nothing worse than receiving an Amazon package only to find that it is broken, not working correctly, or in an imperfect condition. This is especially troublesome when it’s an essential product — such as pet food — that you need ASAP. Now Amazon has developed an AI solution to tackle this problem.
On Wednesday, Amazon unveiled Project P.I. (private investigator), which uses generative AI and computer vision to detect product defects before they reach the customer. In addition to checking for product damage, Project P.I. can double-check that the color and size match the customer’s order, avoiding misshipments.
Project P.I. is found in fulfillment centers across North America, expanding to additional sites throughout 2024. At these sites, millions of products are scanned in imaging tunnels each month. If a defect is found, such as a bent book cover, Amazon isolates the product to ensure it is not sent to the customer, and investigates whether the issue affects similar items.
Then, Amazon associates review the products flagged by Project P.I. to determine whether they have another use, such as being donated or resold at a discounted price on Amazon’s Second Chance site, where Amazon sells open-box and certified refurbished products.
In addition to enhancing manual inspections at fulfillment centers and ensuring users get their products in ideal condition, Amazon notes that this initiative helps create a more sustainable experience by preventing returns that lead to wasted packing materials and unnecessary carbon emissions.
“By leveraging AI and product imaging within our operations facilities, we can efficiently detect potentially damaged products and address more of those issues before they ever reach a customer, which is a win for the customer, our selling partners, and the environment,” said Dharmesh Mehta, VP of Worldwide Selling Partner Services at Amazon.
Additionally, Amazon is using a multi-modal large language model (LLM) to investigate negative customer experiences. The LLM reviews customer feedback and analyzes images taken from Project P.I. and other sources to confirm the cause of the problem.
