Amazon is scale down The company's artificial intelligence (AI) system allows customers to leave the store without paying, and experts say there are lessons to be learned from the decision.
In the wake of the retail giant's recent move to suspend its Just Walk Out (JWO) technology, a system designed to automate the shopping experience by eliminating the need for traditional checkout. A dialogue between experts has begun about the current state and future possibilities of AI in retail. . This move highlights the challenges faced when integrating AI technology into retail environments.
“Amazon designed a 'sensor fusion' system, but cost and accuracy challenges arose.” Will GlaserFounder and CEO of Grabangotold PYMNTS, a rival provider of checkout-free technology for grocery and convenience stores. “AI in retail is as viable as ever. For JWO, the limitation of AI is that it is prohibitively expensive to install and operate.”
not fully automatic
Amazon's AI technology may have hit a practical roadblock.according to informationJWO wasn't completely autonomous, but relied on about 1,000 employees in India to help customers pick up items, put them back, and eventually store them equipped with Amazon's Just Walk Out technology. Amazon disputed this.
An Amazon spokesperson told PYMNTS that the company has more than 130 third-party Just Walk Out technology locations in the U.S., U.K., Australia and Canada, with new locations opening every month.
“The misconception that the Just Walk Out technology relies on human judges monitoring shoppers from India live is misleading and inaccurate,” an Amazon spokesperson said. “As with many AI systems, the underlying machine learning models are continually improved by generating synthetic data and annotating real video data. We verify a small portion of shopping visits by reviewing clips to ensure that our systems are operating to a high standard of accuracy. This is made possible through continuous improvements to both our algorithms. and uses human input to correct it.”
Glaser said JWO relies on shelf sensors and is not suitable for regular grocery stores of any size. Shelving must be completely replaced or retrofitted. “Even in small stores, these shelf sensors create thousands of single points of failure and require weekly outages for maintenance,” he added.
More importantly, systems like JWO can drive up operating costs, Glaser said. “For example, store teams must constantly ensure each item is in the right place for the system to work,” he added. “Also, shelves must be reprogrammed every time a new product is added. The additional labor hours required to support JWO outweigh the potential cost savings.”
jackie walkerRetail Experience Strategy Lead, North America Publicis Sapienta digital business transformation company, told PYMNTS that JWO's hardware doesn't make sense from a cost perspective.
“Unfortunately, in the United States, we always face significant challenges of scale. In anything that requires measurements of the physical environment, scale means cost, and scale is the enemy,” Walker added.
Amazon's JWO isn't the only retail AI system experiencing problems. For example, Walker said: Presto — speech recognition software Fast food drive-thrus are augmenting AI by having offshore staff take orders synchronously, but there hasn't been enough meaningful progress on this either. She pointed out that Just Walk Out also lacks features that provide value to consumers, such as finding deals at nearby stores and viewing cumulative receipts.
“If we are going to raise the perception that we are violating customer privacy by using cameras to track customers, we need commensurate value to the customer so that they can see a mutual exchange of value,” Walker said. he added. “While avoiding checkout lines is certainly worth it, it's only part of the equation and can come with additional trade-offs, such as checking your receipt at Sam's Club after using Scan and Go technology. It happens often.”
lessons learned
Observers say the setback at JWO may be less about commentary on AI itself and more about how the technology needs to evolve. It's important to remember that JWO was founded long before the generative AI boom of the past 18 months. michael zakurfounder of 5 new digitaltold retail and digital commerce consultancy PYMNTS.
“The AI that was being deployed was more traditional algorithmic and data-driven back-end technology that worked well in e-commerce, marketing, and data analytics, but was not fully applicable to brick-and-mortar stores. There was no,” he added.
Zakkour said this isn't the end of leveraging AI to improve the store experience and improving AI for self-checkout.
“But it is a wake-up call that there is a lot of work to do to make this a viable and scalable model for the retail industry,” he added.
Although JWO didn't go completely as planned, Denis BaranoffHead of Retail Business data artpredicted to PYMNTS that AI will eventually become more useful in retail.
“For example, we can read customer sentiment based on phrase recognition, or we can quickly find products based on photos and display them in-store,” he added. “There are many ways to progress and realize what was previously just a dream.”
