Amazon’s marketplace is broken—but Andy Jassy wants to give you an AI shopping assistant
On Thursday, Amazon CEO Andy Jassy released his third letter to shareholders since Jeff Bezos departed the company to go build an island compound in Miami. That annual memo gets studied inside and outside of tech as a sort of weather vane pointing not just at where Amazon is headed, but to where the e-commerce customer experience in general is moving. Since 2021, Jassy has stayed fairly circumspect about AI’s future at Amazon, but his latest letter does not: When people ask “what’s next” at Amazon, he writes that he now responds with: “Generative AI.” It has potential to be bigger than the Amazon Marketplace, Prime, and AWS, Jassy goes on to say. The letter outlines three specific examples. The first is by selling top-of-the-line chips that companies like Anthropic, Airbnb, and Snap can use to develop and train their versions of AI. The second is meant for a different suite of clients like Delta Air Lines, Intuit, and Pfizer that want to build and scale bespoke generative AI apps. The third finally gets to customer-facing apps, the only space where Amazon’s AI products intersect with end consumers directly. Examples include Rufus, the brand new AI-powered assistant that Amazon started touting two months ago that brings a more “conversational shopping experience” to the Marketplace. These may sound fine and well to many consumers, except that for a growing number, there is a looming fear that the major tech innovations that transformed life in recent decades, from Google internet search, to social networking on Facebook, to Amazon Prime free shipping, are all being slowly hijacked by AI. “The internet is in decay” is how New York Times columnist Ezra Klein put it in a recent podcast interview with Verge editor-in-chief Nilay Patel. Both agreed that Google Search now essentially blasts users with a firehose of junk sites built to game the algorithm. Facebook, YouTube, TikTok, and X, meanwhile, are filled with misinformation, bots, and hate speech galore. “Into this weakened internet came the flood of AI-generated junk,” Klein went on to say, causing Patel to add: “When you increase the supply of stuff onto those platforms to infinity . . . the business models of the internet break down completely.” Like others, they believe Amazon has a definite problem here, as a primary avenue through which people access their subpar internet. Yet the company hasn’t addressed it proactively enough for many consumers. Fraud has plagued the Marketplace for years, but AI exacerbates the issue by making it easier than ever to list counterfeit products. Reached for comment, an Amazon spokesperson said the company is “becoming more effective at identifying, seizing, and disposing of counterfeit products which is acting as a deterrent to other bad actors,” and added that “since 2020, while the number of products available for sale in our store has grown significantly, we have seen a more than 30% decrease in the total valid notices of infringement submitted by brands.” More worryingly, in what sounds like it should be part of the same problem but somehow isn’t, the Marketplace is also drowning in a sea of questionable brands that Amazon is either unable or unwilling to remove, which often carry patently absurd names like Bizzliz and SHTSFD. For instance, Fast Company did a quick Amazon search for “pens” before this story was published, just for fun. The sponsored result up top was for three different 50-packs of Deegtran No Bleed Click Pens. Deegtran doesn’t have a digital presence outside of Amazon. Mysteriously, all three of its pen packs showed the same 4.5-star rating and the same number of product reviews (461). We also tried one more random search, for paperclips. The top hits? A 600-count box of “Ootday Sanbanqiaos.” The next options were for paperclips by Fudao Family, the Luxurecourt Store, and SOTMALTK, followed by a box of Upstoues by Vinaco, which seems to be a brand specializing in light bulbs. The sole recognizable brand—Officemate—was also the cheapest of these, but it appeared below all of them. To the extent Amazon has tried to control AI’s negative impact, attempts arguably lay bare the futility of fighting back. Last September, a rule was enacted to limit the number of books that authors can self-publish on the Kindle Direct Publishing platform to three per day. Who on Earth is capable of writing three new books every day? A book-writing bot. Amazon argued the change would “help protect against abuse,” but people could call that a generous description: If even a mere 100 bots maxed out that new daily cap, by next fall book shoppers will be navigating past an additional 109,500 AI-generated volumes that, if history’s any clue, are going to carry names like When the Three Attacks, Apricot Bar Code Architecture, and Ma La Er Snorted Scornfully. Making matters worse, human authors increasingly complain to Amazon about the latest trick by scammers: Kindle e-books where AI simply s
Business Economic Syndicated News 
On Thursday, Amazon CEO Andy Jassy released his third letter to shareholders since Jeff Bezos departed the company to go build an island compound in Miami. That annual memo gets studied inside and outside of tech as a sort of weather vane pointing not just at where Amazon is headed, but to where the e-commerce customer experience in general is moving.
Since 2021, Jassy has stayed fairly circumspect about AI’s future at Amazon, but his latest letter does not: When people ask “what’s next” at Amazon, he writes that he now responds with: “Generative AI.”
It has potential to be bigger than the Amazon Marketplace, Prime, and AWS, Jassy goes on to say. The letter outlines three specific examples. The first is by selling top-of-the-line chips that companies like Anthropic, Airbnb, and Snap can use to develop and train their versions of AI. The second is meant for a different suite of clients like Delta Air Lines, Intuit, and Pfizer that want to build and scale bespoke generative AI apps. The third finally gets to customer-facing apps, the only space where Amazon’s AI products intersect with end consumers directly. Examples include Rufus, the brand new AI-powered assistant that Amazon started touting two months ago that brings a more “conversational shopping experience” to the Marketplace.
These may sound fine and well to many consumers, except that for a growing number, there is a looming fear that the major tech innovations that transformed life in recent decades, from Google internet search, to social networking on Facebook, to Amazon Prime free shipping, are all being slowly hijacked by AI.
“The internet is in decay” is how New York Times columnist Ezra Klein put it in a recent podcast interview with Verge editor-in-chief Nilay Patel. Both agreed that Google Search now essentially blasts users with a firehose of junk sites built to game the algorithm. Facebook, YouTube, TikTok, and X, meanwhile, are filled with misinformation, bots, and hate speech galore.
“Into this weakened internet came the flood of AI-generated junk,” Klein went on to say, causing Patel to add: “When you increase the supply of stuff onto those platforms to infinity . . . the business models of the internet break down completely.”
Like others, they believe Amazon has a definite problem here, as a primary avenue through which people access their subpar internet. Yet the company hasn’t addressed it proactively enough for many consumers. Fraud has plagued the Marketplace for years, but AI exacerbates the issue by making it easier than ever to list counterfeit products.
Reached for comment, an Amazon spokesperson said the company is “becoming more effective at identifying, seizing, and disposing of counterfeit products which is acting as a deterrent to other bad actors,” and added that “since 2020, while the number of products available for sale in our store has grown significantly, we have seen a more than 30% decrease in the total valid notices of infringement submitted by brands.”
More worryingly, in what sounds like it should be part of the same problem but somehow isn’t, the Marketplace is also drowning in a sea of questionable brands that Amazon is either unable or unwilling to remove, which often carry patently absurd names like Bizzliz and SHTSFD.
For instance, Fast Company did a quick Amazon search for “pens” before this story was published, just for fun. The sponsored result up top was for three different 50-packs of Deegtran No Bleed Click Pens. Deegtran doesn’t have a digital presence outside of Amazon. Mysteriously, all three of its pen packs showed the same 4.5-star rating and the same number of product reviews (461).
We also tried one more random search, for paperclips. The top hits? A 600-count box of “Ootday Sanbanqiaos.” The next options were for paperclips by Fudao Family, the Luxurecourt Store, and SOTMALTK, followed by a box of Upstoues by Vinaco, which seems to be a brand specializing in light bulbs. The sole recognizable brand—Officemate—was also the cheapest of these, but it appeared below all of them.
To the extent Amazon has tried to control AI’s negative impact, attempts arguably lay bare the futility of fighting back. Last September, a rule was enacted to limit the number of books that authors can self-publish on the Kindle Direct Publishing platform to three per day. Who on Earth is capable of writing three new books every day? A book-writing bot. Amazon argued the change would “help protect against abuse,” but people could call that a generous description: If even a mere 100 bots maxed out that new daily cap, by next fall book shoppers will be navigating past an additional 109,500 AI-generated volumes that, if history’s any clue, are going to carry names like When the Three Attacks, Apricot Bar Code Architecture, and Ma La Er Snorted Scornfully.
Making matters worse, human authors increasingly complain to Amazon about the latest trick by scammers: Kindle e-books where AI simply summarized or reworded their copyrighted work. It’s hard to thwart them, at least not without Amazon proactively stopping them outright, because unless the e-books have committed clear-cut plagiarism, the authors have learned they have little legal recourse—similar in principle to how they wouldn’t be able to sue Cliffs Notes.
Yet Amazon doesn’t appear terribly concerned by this—or not in the new shareholder letter, anyway. The term “generative AI” appears 22 times in it, and Jassy writes that of all the “customer experience improvements across our businesses,” one he’s extremely pleased with is that Amazon continues to have “the broadest retail selection, with hundreds of millions of products available, tens of millions added last year alone.”
Jassy appeared on CNBC’s Squawk Box Thursday for an exclusive interview on Amazon’s exciting pivot to AI. Host Andrew Ross Sorkin noted that Jassy’s whole shareholder letter seemed to revolve around it. “Yeah,” Jassy said, “AI is going to transform every customer experience that we know.”
A customer experience they didn’t get around to discussing was how AI may already be transforming shopping on the Amazon Marketplace. But Jassy did note their big plans to deliver orders much faster through Prime Air. “When you think about what we’re doing with drones,” he explained, “I think in several years we’ll be able to get items to customers in less than an hour.”
