Fighting packaging waste with artificial intelligence: Amazon’s “Package Decision Engine” uses deep learning, natural language processing, and computer vision to determine the ideal packaging for every product. According to the company, the AI can predict when a more durable item – like a wool blanket – requires no extra protection, or when a fragile product –such as glassware – needs a sturdier box. “With this technology, we’re able to determine the most efficient packaging type for every item we handle,” Amazon states. “This helps us reduce the amount of cardboard envelopes and boxes, paper padding, tape, and paper bags we use when shipping orders to customers.”

Amazon’s “Package Decision Engine” Drives Sustainability

The multimodal AI model, powered by Amazon Web Services, factors in product dimensions, material properties, and customer feedback. Before implementing the AI solution, Amazon employees conducted physical tests to identify appropriate packaging. However, relying solely on manual processes proved unsustainable at scale. With the “Package Decision Engine,” the process has been automated and scaled, enabling the efficient handling of a product catalog with hundreds of millions of items.

Global Rollout and Future Outlook

The AI solution is already deployed in numerous fulfillment centers across North America and Europe and is set to expand to additional sites in India, Australia, and Japan. Amazon plans to continue refining the model to enable even more efficient packaging decisions. Since 2015, the company reports having saved more than three million tons of packaging material globally.