Thank You, Amazon

That was my answer yesterday when a good friend asked me, with genuine concern: “Now that Amazon has decided to license their AI-search engine to merchants, does that spell the end of 𝐬𝐡𝐨𝐩𝐩𝐞𝐫𝐆𝐏𝐓?”

And here is why I meant every word.

On Tuesday, AWS announced its Agentic Shopping Assistant. The pitch to retailers is direct: take the conversational shopping technology Amazon has been refining inside its own marketplace and bring it to your storefront. Build it on Bedrock, AgentCore, and OpenSearch. Launch in “as little as 60 days.” Tapestry’s Kate Spade is the marquee customer, deployed in partnership with the AWS Generative AI Innovation Center.

Two years ago, when we started building 𝐬𝐡𝐨𝐩𝐩𝐞𝐫𝐆𝐏𝐓, the hardest conversations were not technical. They were about the premise. Did merchants actually need AI-native discovery on their own site? Was conversational search a feature, or a category? Would the mid-market really pay for it?

The architecture was the easier call. Shopify app dev stack on top, Google Cloud underneath – Cloud Run, Cloud SQL, Vertex AI, Gemini. The same stack AWS is now selling, just a different cloud provider.

Amazon’s launch settles every one of those questions.

𝐓𝐡𝐞 𝐡𝐲𝐩𝐨𝐭𝐡𝐞𝐬𝐢𝐬 𝐢𝐬 𝐯𝐚𝐥𝐢𝐝𝐚𝐭𝐞𝐝

When a company with more product discovery data than any retailer in history decides the answer is to license its discovery stack to other retailers, the market has spoken. Every merchant should have AI search. Tapestry / Kate Spade is the marquee public customer. Amazon does not enter categories too small to move its needle, so the revenue opportunity is now also validated.

The premise question is closed. The revenue question is closed with it. Investors no longer need to be convinced this is a category, or that the category is large. Merchants no longer need to be convinced this is a 2026 priority.

AI search is no longer optional – it’s now table stakes if you want to play in the game.

So what’s left to debate?

The how.

𝐓𝐡𝐞 𝐛𝐮𝐢𝐥𝐝-𝐯𝐬-𝐛𝐮𝐲 𝐦𝐚𝐭𝐡 𝐦𝐞𝐫𝐜𝐡𝐚𝐧𝐭𝐬 𝐰𝐢𝐥𝐥 𝐚𝐬𝐤

Sophisticated merchants and investors will look at the AWS launch and ask the right question. If Amazon is selling the pieces, why pay 𝐯𝐢𝐬𝐮𝐚𝐥𝐀𝐈 to bundle them? Why not roll your own?

The build math, which Amazon’s own announcement makes for me, is the answer.

“As little as 60 days” with AWS Professional Services, and that is the floor, not the average. It assumes you already have an AI plus AWS team in place. An AWS account, IAM and security configuration, Bedrock model provisioning, AgentCore agent definition, OpenSearch ingestion pipeline, and storefront integration code. Pricing is a private offer on AWS Marketplace, structured as consumption against Bedrock plus AgentCore plus OpenSearch billing. Catalog ingestion is “out of the box,” but enrichment, attribute normalization, image description, and color taxonomy are not. The shopper interaction is text-only conversation. The discovery surface is inbound to the storefront. Distribution to ChatGPT, Gemini, and Perplexity is not part of the package.

For a mid-market merchant with a 5K to 50K SKU catalog and no AI or AWS-fluent platform team, the 60 days resets to “first hire your team, then start.” That is enterprise IT, not a Shopify app install. Even with the right team already on payroll, you own the infrastructure, the model bill, and the ongoing engineering load forever. And there’s close to 6M of them on the Shopify platform alone.

𝐯𝐢𝐬𝐮𝐚𝐥𝐀𝐈 exists for the ones that the math doesn’t work for.

𝐖𝐡𝐚𝐭 𝐰𝐞 𝐬𝐡𝐢𝐩, 𝐭𝐨𝐝𝐚𝐲

𝐬𝐡𝐨𝐩𝐩𝐞𝐫𝐆𝐏𝐓 is in the Shopify App Store. Click install and the pipeline takes over. We ingest. We clean. We enrich. We create AI embeddings for every product across text, image, and color. We auto-generate color names so a shopper can search “burnt orange” or select an exact color match to their wedding dress and the system finds the best matches. The shopper search bar accepts text, image, or color. Trimodal, not text-only. Live in production on Cassie’s Kurations’ 8,000-SKU catalog. 𝐏𝐥𝐚𝐧𝐬 𝐬𝐭𝐚𝐫𝐭 𝐚𝐭 $𝟑𝟗 𝐰𝐢𝐭𝐡 𝐚 𝐟𝐫𝐞𝐞 𝐭𝐫𝐢𝐚𝐥. All up front costs and risks are ours, not the merchant’s. And our A/B testing dashboard allows merchants to start at as little as 1% deployment, watch comparative results roll in and then dial up the percentage when the results prove the value.

And 𝐬𝐡𝐨𝐩𝐩𝐞𝐫𝐆𝐏𝐓 is only one of our discovery products. The full discovery platform that’s built around the patent-pending engine is what closes the door on anything a merchant could assemble themselves.

𝐜𝐥𝐞𝐚𝐧𝐞𝐫𝐆𝐏𝐓 sits underneath 𝐬𝐡𝐨𝐩𝐩𝐞𝐫𝐆𝐏𝐓 and rewrites bad titles, normalizes attributes, strips broken markup, and produces the structured catalog every AI surface needs. AWS’s Agentic Shopping Assistant ingests what you give it. We give you a better catalog before ingestion happens.

𝐜𝐚𝐭𝐚𝐥𝐨𝐠𝐆𝐏𝐓 makes the same clean catalog visible to ChatGPT, Gemini, and Perplexity through UCP, ACP, and Schema.org feeds. AWS’s Agentic Shopping Assistant is an inbound assistant on your storefront. 𝐜𝐚𝐭𝐚𝐥𝐨𝐠𝐆𝐏𝐓 is the outbound surface that gets you found before the shopper ever lands.

𝐬𝐭𝐮𝐝𝐢𝐨𝐆𝐏𝐓 is our virtual try-on layer, on track for Shopify App Store delivery in July. Trimodal discovery into a product page, then a digital-twin try-on of the garment the shopper just landed on. Discovery plus try-on, on one platform, for the merchants the Walmart and Amazon proprietary stacks were never going to serve.

𝐝𝐞𝐯𝐞𝐥𝐨𝐩𝐞𝐫𝐀𝐏𝐈 extends the engine to headless storefronts, agencies, and ERPs in the second half of the year. REST endpoints, an MCP server, and distribution into ChatGPT Apps, Claude, and any MCP-compliant host.

One engine. Five products. Shopify first.

𝐓𝐡𝐞 𝐩𝐨𝐬𝐢𝐭𝐢𝐨𝐧𝐢𝐧𝐠 𝐰𝐞 𝐝𝐢𝐝𝐧’𝐭 𝐡𝐚𝐯𝐞 𝐲𝐞𝐬𝐭𝐞𝐫𝐝𝐚𝐲

Before this week, the 𝐯𝐢𝐬𝐮𝐚𝐥𝐀𝐈 elevator pitch was “the Shopify-first AI discovery platform for the merchants the cloud giants like AWS won’t serve.” That was a story I had to explain.

After this week, the elevator pitch is “Amazon just licensed its version of what we already ship. Ours installs in a Shopify store in minutes, includes color and image search alongside text, and offers additional options to clean your catalog on the way in, make you visible to ChatGPT and Gemini, and add VTO in July. Amazon’s takes at least 60 days, only if you already have an AI plus AWS team, and only does the search piece.”

The story tells itself.

𝐀 𝐭𝐡𝐚𝐧𝐤 𝐲𝐨𝐮, 𝐧𝐨𝐭 𝐚 𝐰𝐚𝐫𝐧𝐢𝐧𝐠

I am not writing this as a competitive shot. AWS shipping its Agentic Shopping Assistant is the best thing that has happened to the AI-commerce category since Shopify added the App Store. Every merchant who reads about Kate Spade and asks “could we do that on our site” is a 𝐬𝐡𝐨𝐩𝐩𝐞𝐫𝐆𝐏𝐓 conversation. Every investor who looks at the AWS launch and asks “is this category real” now has their answer.

A prospective investor said to me about 6 months ago “I really don’t think this is ever going to be a thing”. Well… it’s officially a thing now.

One giant may have validated the category. But the mid-market is going to deploy it. And that’s where 𝐯𝐢𝐬𝐮𝐚𝐥𝐀𝐈 lives.

So, yes, thank you Amazon. The validation is appreciated. The at-best 60-day timeline, with an AI plus AWS team already in place, is also appreciated. We will be here when the merchant decides the 1-click install was the more interesting route to travel.

#𝐀𝐠𝐞𝐧𝐭𝐢𝐜𝐂𝐨𝐦𝐦𝐞𝐫𝐜𝐞 #𝐀𝐈𝐒𝐞𝐚𝐫𝐜𝐡 #𝐒𝐡𝐨𝐩𝐢𝐟𝐲 #𝐁𝐮𝐢𝐥𝐝𝐕𝐬𝐁𝐮𝐲 #𝐀𝐈𝐂𝐨𝐦𝐦𝐞𝐫𝐜𝐞