While most of the tech world is focused on shiny new generative AI tools, Amazon has been chipping away at an ongoing challenge posed by modern consumerism: the proliferation of shipping materials.
For several years now, the e-commerce giant has been developing what it describes as a “multimodal AI model” called the Package Decision Engine.
The PDE’s job is to do a smarter job of selecting the right box, bag, or wrapper for each of the millions of unique items sold through the company’s warehouses.
Amazon’s earlier packing strategy — chosen by humans and less intelligent computers — was frequently a source of confusion for customers and ridicule toward the company. Frequent Amazon shoppers have almost certainly received a shipment with a single small item in a comically large box.
Now, instead of humans doing physical tests, the company says products are sent through a computer vision tunnel that gathers dimensions and particular features (like whether it has fragile parts or already resides in a box).
Those images are then matched with a natural language processing of text-based description of the product, plus other quantitative data to match the item with its ideal shipping solution.
While there are some unexpected reasons for why larger packaging is in fact a smart choice, Amazon says it is committed to reducing the amount of cardboard it uses as part of its sustainability pledges.
It makes business sense too: when sending billions of parcels, taking even a little bit from each one can add up to some staggeringly large numbers.
The company estimates using correctly sized boxes, switching to softer mailers, or skipping packaging altogether now saves 60,000 tons of cardboard per year in North America alone.