The Playbook Amazon Used to Build AWS Is Back.

Amazon has done this before. It built a cloud computing infrastructure to run its own retail operations, proved it worked at the world's most demanding commercial scale, and then sold it to every company on the planet, including its direct competitors. Amazon Web Services became the most profitable division in the company's history.

On May 27, 2026, Amazon did it again. AWS announced the Agentic Shopping Assistant on AWS, a packaged solution that takes the architecture, starter code, and learnings from Alexa for Shopping, the AI shopping assistant that drove nearly $12 billion in incremental sales for Amazon last year, and makes them available to retailers outside Amazon for the first time.

The pattern is now two decades old. Amazon builds something internally to solve a problem at a scale no other company faces.

It proves the solution works through billions of real-world interactions. Then it packages that solution and sells it to the industry, including the competitors it built it to beat.

The Technology That Proves the Case

The commercial evidence behind the Agentic Shopping Assistant on AWS is more specific than most retail AI announcements.

Amazon's AI shopping assistant served more than 300 million customers last year. It drove nearly $12 billion in incremental sales. Conversational shopping sessions built on the same architecture convert at 3.5 times the rate of traditional keyword search

Amazon serves as what AWS calls Customer Zero for its own retail technology, meaning every component of the Agentic Shopping Assistant has been tested in one of the world's most demanding retail environments before being offered to anyone else.

The architecture is built on Amazon Bedrock, AgentCore, and OpenSearch, validated through billions of real shopping interactions on Amazon.com. Retailers are not buying a pilot. They are buying a proven production system.

AWS offers retailers the technical foundation like architecture guidance, starter code, and hands-on support from the AWS Generative AI Innovation Center, with deployments achievable in approximately 60 days.

Each deployment is then customized to match the individual retailer's catalog, customer base, brand voice, and domain expertise. The retailer brings the knowledge. AWS brings the infrastructure.

Kate Spade as the First Case Study

The first production deployment demonstrates what the Agentic Shopping Assistant looks like in practice. Kate Spade launched the Kate Spade AI Gift Concierge on April 13, 2026, the first production-ready retail AI assistant built with Amazon Bedrock AgentCore.

The tool addresses a specific and measurable consumer pain point: 53% of shoppers report stress during gift purchases.

Rather than building a general shopping assistant, Kate Spade used the solution to create a conversational gift concierge that engages shoppers in natural dialogue about occasion, recipient, and style.

The experience was grounded in real shopper behavior and insights drawn from the questions customers asked Alexa for Shopping.

Built on Anthropic's Haiku 4.5 model with Amazon Bedrock providing observability, authentication, and evaluations, the Kate Spade team completed 2.5 months of rigorous testing before the tool became customer-facing.

"AWS brought the recipe, but together we built the customization our consumers needed," said Yang Lu, Chief Information and Digital Officer at Tapestry. Additional retailers are currently in testing, unnamed as of the announcement date.

The Strategic Argument Amazon Is Making

The announcement is notable not just for what it offers but for the argument Amazon is making to the retail industry alongside it. AWS told retailers explicitly that they face a critical choice, build their own AI presence or risk becoming dependent on general-purpose answer engines that do not serve their brand or their customers.

That argument has a clear subtext. Amazon is not the general-purpose answer engine it is warning retailers about. Amazon is the company offering the infrastructure to build something better, a proprietary AI shopping experience grounded in each retailer's own product knowledge, customer relationships, and brand identity.

The retailers that take the AWS offer maintain direct customer relationships. The retailers that rely on external AI agents surrender the discovery layer to whoever operates them.

Amazon's own retail strategy reinforces the framing. It has walled off Amazon.com from being scraped by external AI agents, protecting its own discovery layer while simultaneously offering retailers the tools to protect theirs.

It has also built Buy for Me, a feature that completes purchases on other retailers' websites on behalf of Amazon customers, a product that benefits from retailers not having their own AI shopping presence to intercept that journey.

The Agentic Shopping Assistant on AWS is Amazon's offer to the retail industry: build your own AI shopping infrastructure before someone else builds it for you.

The conversion rate data, 3.5 times that of traditional keyword search, is the evidence that the infrastructure is worth building. The $12 billion in incremental sales Amazon generated from the same technology last year is the evidence that it works.