About seven years ago, Amazon introduced ‘Just Walk Out,’ a grab-and-go checkout system that allows customers to walk into a store, grab an item, and walk out without making a checkout line – with payment happening electronically after the visit.
Now, The Information is reporting ‘Just Walk Out’ got the ol’ boot. And it turns out the tech relied pretty heavily on human oversight.
Humans: One. AI: Zero.
Behind the Scenes of Amazon’s Just Walk Out Tech
This tech had a pretty cool concept.
“Just Walk Out” used computer vision, machine learning and sensors to detect when customers picked up items, what they grabbed and when they put something back on the shelves.
According to The Information, its success was largely driven by over 1,000 human workers who had to review hours of computer footage to correctly document transactions. These workers, based in India, also had to label images from the videos to train the tech’s AI model.
Amazon refuted the accuracy of The Information’s report regarding human reviewers but did not provide any specifics.
Given the immense manual labor involved, it could take hours for customers to get a receipt and know exactly how much they spent.
In mid-2022, the tech required 700 human reviews per 1,000 sales. Amazon’s team attempted to cut human reviews down by 90% but consistently missed goal.
In addition to Amazon Go stores, Amazon also pitched the tech to retailers like Target and Walmart, who had to decide if it was worth the expensive installation and maintenance costs.
One grocery store owner, Kortney Lee, was satisfied with the tech for two key reasons: It allowed employees to focus on customer interactions and the inventory sensors made restocking easier.
However, there are some things to consider:
- The installation is costly and time-consuming – Lee had to close his store for seven months and dish out $240,000 to build new walls and move outlets.
- Shelves require constant monitoring to ensure customers are charged for the right item.
- The tech isn’t budget-friendly, as it makes it hard for shoppers to track their spending while in store.
Retail stores weren’t the only ones dishing out funds – running “Just Walk Out” cost Amazon millions in just annual model training.
Where they saw the most profit is by selling the tech to stores in unique venues, like airports, sports stadiums, and college campuses.
Using AI Isn’t Quite So Simple
Although AI has been the buzzword of the last two years, it’s still out of reach to many companies. Namely for one reason: Cost.
As one MIT study revealed last year, the current cost of implementing AI exceeds its value in most cases.
In this case, while a grab-and-go checkout system creates a seamless shopping experience, it’s a costly investment for a store owner. Not only on the front end to install it but also the maintenance it requires.
For certain use cases, integrating AI can be both quick and cost effective – like adding an AI chatbot to your site similar to Klarna’s AI assistant.
Companies have to figure out when, where and how AI makes sense for their business and customers, rather than jumping head first.