By words Director of Applied Science at Amazon Robotics Jeremy Wyatt, employees of the company now independently inspect goods for damage during their sorting and packaging. However, they often fail to spot defective products as they are rare—less than 1 in 1,000 items are damaged, Amazon claims. On the other hand, Amazon processes billions of packages a year, so the number of defective items is quite high.

The AI-based software works by scanning objects placed in crates to look for signs of wear and tear at special stations. If an item is marked as damaged, an Amazon employee will check it again and decide if it can be packaged and shipped.

The technology was introduced in two warehouses. AI is 3 times more likely to detect damaged goods than a human. The artificial intelligence was trained to identify the typical visual characteristics of defective products by comparing images of products without damage to those with defects.

Amazon is interested in using more technology to automate routine and monotonous work. Amazon previously introduced robots capable of lifting and moving heavy objects, scanning barcodes, or placing goods in carts. Machines cannot fulfill all roles and work together with people.

Amazon believes that robots are the future, which is not surprising given that the company has come under fire for its relatively high injury rate in the workplace, complaints of union harassment and violations of labor laws. In addition, it is difficult for companies to retain employees.