Amazon, the leading e-commerce giant, is set to revolutionise its warehousing operations with the deployment of artificial intelligence (AI) across a dozen of its largest warehouses, as reported by the Wall Street Journal. This move aims to screen items for damage before they are shipped to customers, resulting in a significant reduction in damaged goods, expedited picking and packing processes, and ultimately paving the way for increased automation within Amazon’s fulfilment centres.
Currently, Amazon warehouse workers are responsible for meticulously inspecting goods for any signs of wear and tear while meeting the company’s productivity targets, which measure the number of orders handled per hour. However, this task can be time-consuming and mentally demanding, as most items are typically in excellent condition, as stated by Jeremy Wyatt, Director of Applied Science at Amazon Robotics.
With estimates suggesting that fewer than one in 1,000 items handled by Amazon are damaged, the overall volume of damaged goods remains significant given the company’s staggering annual package count of approximately 8 billion. Recognising the need for improved efficiency, Amazon is embracing AI technology to enhance its warehouse operations, especially in the inspection and quality control processes.
This strategic decision aligns with the broader industry trend of incorporating AI into logistics operations, as retailers, supply-chain operators, and software developers seek to streamline workflows and simplify decision-making throughout the supply chain. Amazon’s pursuit of increased warehouse automation arises from its ongoing efforts to address labour shortages and relieve human workers of physically demanding and repetitive tasks, with the objective of assigning such responsibilities to robots.
Deploying AI in logistics necessitates the development of technology that can effectively replace tasks typically performed by humans, such as item selection, packing, and damage assessment, according to Rueben Scriven, Research Manager for the warehouse automation sector at Interact Analysis. While these tasks may be straightforward for human workers, automation requires advanced systems capable of not only performing the tasks but also identifying damaged items accurately.
For Amazon, minimising the number of damaged goods sent to customers is of paramount importance as it directly impacts the overall customer experience. Recognising this, Amazon has already implemented AI technology at two fulfilment centres and plans to extend its use to ten additional sites across North America and Europe. Christoph Schwerdtfeger, a Software Development Manager at Amazon, highlights that the AI system is three times more effective than a human warehouse worker in identifying damaged items.
The AI inspection process occurs during the picking and packing stages. As goods are selected for individual orders and placed into bins, they pass through an imaging station where they undergo scrutiny to verify the accuracy of the selection. With the introduction of AI, this imaging station now includes an evaluation of the items for any damage. If an item is flagged as damaged, the bin is diverted to a human worker for further inspection. Conversely, if the item appears undamaged, the order proceeds to the packing stage and subsequently to the customer for delivery.
To train the AI, Amazon utilised a dataset of photos depicting both undamaged and damaged items. By comparing these images, the AI system was trained to recognise the distinctions between pristine and imperfect products, enabling it to flag items that fall short of perfection during the inspection process, as explained by Christoph Schwerdtfeger to WSJ.