Marketplace Briefing: Why Amazon is selling its AI-powered shopping tools to other retailers

The move takes a ​page out of ⁠Amazon’s cloud computing unit’s playbook. The company built AWS after developing cloud computing infrastructure for its own e-commerce operations, and it later grew into the world’s largest cloud computing provider. Amazon has since expanded that strategy into areas including supply chain services, cashier-less checkout technology and advertising.

Earlier this month, Amazon replaced its Rufus shopping chatbot with an assistant called Alexa for Shopping, which combines Rufus with the company’s newer Alexa+ technology. Amazon has increasingly pushed AI deeper into its shopping experience as tech companies race to turn chatbots into tools that can recommend products, compare items and complete purchases.

Amazon introduced Rufus in 2024 as an AI shopping assistant designed to answer product questions and recommend items. Over time, the company added more automated shopping features, including price tracking and purchasing tools. Amazon previously disclosed that Rufus was used by more than 300 million customers last year and helped drive nearly $12 billion in incremental annualized sales.

Now, Amazon is positioning that technology as a service for the broader retail industry.

Tapestry-owned luxury womenswear brand Kate Spade is already using the technology to power a new AI gifting assistant, according to Amazon. Other retailers are “currently in testing.” Amazon said the Kate Spade assistant was designed to help shoppers choose gifts by asking conversational questions about the recipient, occasion and style preferences.

In its announcement on Wednesday, Amazon said the new AWS offering was built using “the technology and learnings behind Amazon’s successful Alexa for Shopping agentic assistant.” The company said retailers can combine the tools with their own “business rules and brand voice” to create customized shopping assistants.

“With every meeting with customers we had, they would say, ‘We want Rufus,’” David Dorf, global head of retail solutions at AWS, told Modern Retail in an interview. “Everybody could benefit from this, so let’s productize it a little bit and figure out a repeatable way to deliver the same technology to multiple retailers.”

Dorf said Amazon believes retailers should build their own branded shopping assistants on their websites and apps, rather than rely solely on general-purpose AI chatbots like ChatGPT and Gemini to guide shoppers toward products. He added that broad AI chatbots are often limited to information scraped from retailer websites, while retailers themselves have access to much deeper customer and product data.

“Retailers know things like: This particular item is returned a lot. This particular item runs in the wrong size. People who buy this love to buy that,” Dorf said. “All that data can be brought to bear in a shopping assistant on their own site that can be more intelligent than a horizontal answer engine like ChatGPT or Gemini.”

Petra Schindler-Carter, general manager of retail and consumer packaged goods at AWS, said many retailers are increasingly worried about losing control over the customer relationship as consumers turn to AI agents and chatbots to discover products online.

“They’re worried about getting disintermediated,” she said.

Sky Canaves, principal retail and e-commerce analyst at eMarketer, said the move fits into Amazon’s broader strategy of supporting more commerce that happens outside its own marketplace while still finding ways to profit from it.

“Amazon wants to have a finger in the pie of all of e-commerce, whether [or not] it takes place on Amazon,” she said. “By improving e-commerce experiences everywhere, that ultimately benefits Amazon because it increases shopping activity.”

The AI shopping market has become heavily contested. OpenAI, Google and Perplexity have all introduced AI shopping tools that allow consumers to research products and, in some cases, make purchases through chat interfaces. Some of those efforts have underperformed, in part due to lagging adoption among both retailers and consumers.

In addition to inking deals with outside AI companies, retailers including Walmart and Target have also been developing their own AI-powered shopping features. Walmart said in February that customers who use its AI agent Sparky spend 35% more per order than those who don’t.

Amazon, meanwhile, is pitching its product as a faster alternative to building an AI shopping assistant from scratch. In its announcement, the company said retailers receive “architecture guidance, starter code, and support from AWS experts,” allowing them to launch AI shopping experiences in weeks “rather than the years it would take starting from scratch.”

AWS executives said speed and responsiveness have become some of the biggest priorities as retailers test AI shopping tools. Schindler-Carter said consumers have little patience for slow AI experiences, particularly in shopping. Indeed, OpenAI quietly scrapped its Shopping Research tool, originally launched in November, because the product was too slow, a company executive told The Aisle in an interview earlier this year.

“This can be a wonderful experience, but if you have to wait a long time, it’s not going to go anywhere,” she said.

Dorf said AWS paused an early rollout of the product after testing showed the experience was not responding quickly enough.

“We had to delay our production run to make sure that the customer experience was the way the customer wanted it,” Dorf said.

Amazon has largely resisted opening its marketplace to outside AI shopping agents. CEO Andy Jassy has said on multiple occasions that the company expects to eventually partner with third-party agents, though Amazon continues to block many bots from accessing its site. Meanwhile, Amazon has launched its own agentic shopping features, like “Buy for Me,” which can purchase products from other retailers’ websites on a customer’s behalf. Some retailers criticized the feature, saying they never opted into the program, Modern Retail previously reported.

Canaves said Amazon may also be trying to make its own AI shopping ecosystem more “interoperable” with outside retailers over time. If more brands adopt Amazon’s technology foundations, it could make it easier for Amazon’s own shopping agents to interact with those retailers in the future, she said.

While AWS executives declined to specify what brands are currently testing the new product, they said interest in the technology spans a broad range of retail categories. Dorf said many of the early retail partners share one common trait: They sell products where shoppers often want guidance or recommendations before making a purchase.

As he put it, “The thread is every one of these is a retailer where if I walked into their store, I’m likely to ask for help.”

What I’m reading

• Logistics firm Stord nabs $250 million to help brands take on Amazon. (Bloomberg)
• Temu owner PDD posts profit miss amid fierce competition in China. (The Wall Street Journal)
• OpenAI targets smaller advertisers with new ChatGPT ads. (The Information)

What we’ve covered

• How Olly is updating its product detail pages for the AI era
• Brands are getting creative as fuel costs raise shipping fees
• What recent retail earnings revealed about the state of consumer health