Amazon ships 20 million packages a day to 19 countries, requiring reams of paper, cardboard, and plastic packaging, and figuring out how to package all that stuff efficiently is good for both the environment and its bottom line.
So in 2019, Amazon launched its own proprietary AI model to reduce packaging waste, and five years later, it’s helping save at least 500,000 tons of packaging annually, which is roughly the weight of 7,750 Boeing 737 aircraft, according to company data.
To handle the massive volume of packages, Amazon researchers built an artificial intelligence model called a Packaging Decision Engine to predict the most efficient packaging choices, for example, ensuring that a set of dinner plates gets a sturdy box while a blanket doesn’t.
The model uses natural language processing and text-based data on every item in the online store, including basic information like product name and description. It also collects feedback from returns and product reviews and integrates information about damaged products received. It combines all this data with photos taken when items arrive at Amazon warehouses using special computer vision channels. These photos provide the company with detailed information on the exact dimensions of each object and capture images from multiple angles to help it determine the best way to pack it.
Over time, the model added more nuance in identifying specific items; for example, personal items like adult diapers wouldn’t be shipped without packaging to save materials, while products with strong magnets would be protected enough so that they could be shipped without packaging.
In a 2021 paper, Amazon researchers Prasanth Meiyappan and Matthew Bales wrote that combining visual and textual data can improve a model’s performance by as much as 30%.
Euihark Lee, an assistant professor at Michigan State University's College of Packaging, said using AI to make packaging decisions is unusual. "I think Amazon is at the forefront of AI applications for packaging because they have so much data," he said. "Other companies don't have that much data."
Amazon said its packaging model helped it reduce more than 2 million tons of packaging between 2015 and 2022. That's an increase of 500,000 tons from the retail giant's previous packaging reduction announcement in 2021, when it reported savings of more than 1.5 million tons since 2015. Amazon doesn't disclose how many tons of packaging materials it uses annually, making it difficult to determine how much of an impact the AI model has had on its overall packaging usage reduction.
Experts told Forbes that any reduction is positive. "I think it's significant!" said David Feber, a senior partner at McKinsey who specializes in packaging. Rafael Auras, professor of packaging sustainability at Michigan State University's College of Packaging, said the reductions are impressive for a single company, but the problem is so widespread that it will have little impact. "Using artificial intelligence to reduce packaging waste is a huge opportunity not only for Amazon but for the entire industry," he said.
Source: Forbes (April 16, 2024) How AI Is Helping Amazon Save Half a Million Tons of Packaging Per Year
