ShopAgentic is betting ecommerce’s next platform is not a dashboard. It is a team of AI agents.

The next big ecommerce platform may not look like Shopify, Magento, BigCommerce, or Seller Central.

It may look more like a squad.

That is the bet behind ShopAgentic, a German startup that just raised €1.9 million to build what it calls a native agentic commerce system. Instead of giving merchants another dashboard, checklist, and maze of settings, ShopAgentic wants to give them a team of specialized AI agents that handle commerce jobs end to end.

One agent cleans product data. One makes decisions. One handles distribution. One helps you sell through AI channels. The merchant still approves the important stuff, but the machine does the heavy lifting.

That sounds cute until you understand the bigger shift.

For 30 years, ecommerce platforms were built around how humans buy. Desktop. Mobile. Social. Omnichannel. Voice. The interface changed, but the buyer was still a person clicking, scrolling, comparing, and checking out.

Agentic commerce changes the buyer.

Now the shopper may be ChatGPT. Or Gemini. Or Claude. Or Alexa or Sparky. Or some personal shopping agent sitting on a phone, laptop, browser, smart speaker, or car dashboard. The human still makes the final decision, at least for now. But the fetching, filtering, comparing, and recommending is increasingly done by software.

That means the game is no longer just, “Can I convince a customer?”

It is also, “Can I convince the customer’s agent?”

That is why ShopAgentic matters. It is not just another AI wrapper for ecommerce. It is a signal that the platform layer is being rebuilt for a world where agents talk to agents, query product data, compare inventory, trigger payments, and decide which products even make the recommendation set.

Its founders, Alexander Ringsdorff and Kai-Thomas Krause, are not random AI tourists who discovered ecommerce last Thursday. Their background runs through Magento, NewStore, CommerceTools, logistics, SaaS commerce, composable commerce, and the backend plumbing that most sellers never see but depend on every day.

That matters because agentic commerce is not a chatbot problem.

It is a plumbing problem.

The old systems were built for humans. Dashboards. Buttons. Tabs. Toggles buried eight menus deep. Pretty PDPs. Image carousels. Rich brand pages. A human can fight through that.

An agent does not want to.

Agents want structure.

They want clean product data. Plain answers. Availability. Pricing. Attributes. Materials. Dimensions. Compatibility. Reviews. Return policy. Fulfillment speed. Comparison logic. Verified claims. Schema. Feeds. APIs.

The boring stuff.

And here is where this connects to Amazon sellers.

The Economist just gave everyone a preview of what is coming. They are testing two versions of the internet. One for humans, full of glossy pages, feature art, and rich layouts. Another for AI agents, stripped down into clean structures, questions, answers, and plain text.

That is the move.

Not human pages or bot pages.

Both.

Amazon brands need the same mindset. Your listing still has to stop a human thumb. Your images still have to sell. Your A+ content still has to build trust. Your hero image still has to be clean, compliant, and conversion friendly.

But behind the pretty version, you need the machine version.

That means your content has to answer the questions a shopping agent would ask before recommending your product.

Most sellers still write listings like they are trying to impress a rushed human. The next layer is writing listings so an AI agent can retrieve, verify, compare, and explain the product without guessing.

That is the AEO game.

Answer Engine Optimization is not replacing SEO. It is adding another referee. SEO helps you rank where humans search. AEO helps you show up where agents answer.

Discoverability means the machine can find you.

Desirability means it has a reason to choose you.

ShopAgentic is aiming at the merchant side of that same problem. Their pitch is that brands should not have to rip out their entire stack and replatform just to survive the agentic shift. Instead, they can plug in agents where they need help: data, decisions, distribution, or selling through AI.

A brand launches a new product and wants it discoverable inside ChatGPT, Gemini, Rufus, or whatever agentic shopping layer wins. Instead of waiting for old channels to crawl, interpret, and maybe understand the product, an agentic system could structure the data, package the proof, push it into the right rails, and make it available for agent-led discovery.

The payment piece gets even weirder. ShopAgentic is also watching machine payments, MPP, x402, and stablecoin rails. That sounds nerdy, but it matters. When agents start buying from agents, they need ways to authenticate, transact, settle, and stay inside guardrails without a human typing a credit card number every time.

In other words, checkout may stop being a page.

It may become a protocol.

That should make every seller pay attention. Amazon is not standing still. Alexa is already teaching shoppers to ask instead of search. ChatGPT shopping is pushing product discovery into the conversation layer. Google, Shopify, OpenAI, Stripe, Meta, Apple, Anthropic, and Perplexity all want a role in the new buying path.

The seller mistake is waiting until the traffic shows up in analytics.

By then, you may already be invisible.

Make sure answers exist in clean text somewhere an AI can read. Bullets. Description. A+ modules. Brand Store. FAQ content. Off-Amazon landing pages. Comparison pages. Blog posts. Support docs. Schema markup. Product feeds. Wherever the agents are likely to look.

Do not abandon the pretty version. Humans still buy with emotion. Agents may fetch the shortlist, but trust still comes from brand, proof, reviews, story, and reputation.

The winning brands will build for both internets.

One for the human scroll.

One for the machine read.

ShopAgentic is early. The protocols are still messy. The winners are not obvious. But the direction is clear.

Ecommerce is shifting from pages people browse to systems agents query.

Your next buyer may not have eyes.

But it will have standards.

TikTok Shop launched on the idea that content-driven discovery would let anyone win, bypassing the search-and-ad paywalls that define marketplaces like Amazon.

But early Marketplace Pulse data on nearly 100,000 U.S. sellers shows the opposite happened: the platform is even more top-heavy than the marketplaces it aimed to disrupt.

The key figures: the top 1% of sellers (fewer than 900) drive 60% of tracked U.S. GMV, and the top 0.1% (fewer than 90 sellers, each averaging $100M+ in lifetime sales) account for over a quarter of it.

The bottom half of sellers contribute just ~0.1% combined. By comparison, on Amazon it takes about 1.6% of sellers to generate half of third-party GMV. So TikTok's concentration runs deeper. The middle tier is thin.

A few caveats on the data: it's early and lifetime-based (estimated from units sold × average price), so it reflects who has captured demand to date rather than current run-rate. Long-tail coverage is still filling in, but the overall shape is expected to hold.

Platform curation reinforces the gap. Badges like Official Store and Gold Star steer visibility toward proven sellers. Official Store sellers move roughly 40x the volume of unbadged ones, Gold Star about 18x. Badging mostly follows scale rather than creating it, but it compounds the advantage.

The takeaway: removing the paywall on discovery didn't flatten outcomes. It sharpened them. Where Amazon's filter is operational sophistication and capital, TikTok Shop's filter is virality, which is its own form of scarcity.

Democratized discovery has produced a smaller winners' circle, not a wider one.

From Jon Derkits, BDSS Dream 100 member and author of the Best@Amazon newsletter. If you sell on Amazon and you're not subscribed, fix that.

Amazon waits until the last minute to drop the real Prime Day dates (it’s next week!). So by the time you know, most of your prep is already locked.

Inventory is inbounded to the FCs. Deals are set up. PPC budgets are mostly set.

People love to act like Prime Day is this massive, overwhelming sprint for sellers. It isn't. The work is spread out.

And more sellers are choosing to sit it out entirely. That's been me for three years.

But if you're betting the farm on Prime Day this year, what should you expect?

Jon went back and pulled post-event data from Prime Day 2025. That was the first time Amazon stretched it from two days to four.

No crystal ball here. But I'd put money on these patterns repeating.

Here's what the data showed:

Activity peaked on Day 1. It declined through the middle days. Then it partially rebounded on Day 4 as shoppers finalized their carts.

Days 1 and 2 were heavy on browsing and click-throughs. That's discovery and cross-marketplace price comparison.

Days 3 and 4 flipped to surging conversions and higher average order values as the pent-up demand finally hit. A lot of shoppers waited until the end.

The day before Prime Day (July 7 last year) actually generated more orders than official Day 1 in some datasets. About 4% more. Clicks during the build-up averaged 61% higher than during the event itself.

Day 1 had balanced volume with clear evening peaks on Eastern and Pacific time. Days 2 through 4 had slower mornings and sharp spikes around 6 PM PST.

So here's what he recommends you do as a seller or brand:

In Greek myth, Prometheus stole fire from the gods and handed it to humanity. Now it's the name of Jeff Bezos's new startup, and the fire he's chasing is manufacturing.

This is Bezos's first CEO job since he walked away from Amazon in 2021. The company came out of stealth last week with $12 billion in fresh funding and a $41 billion valuation. Backers include JPMorgan, BlackRock, Goldman Sachs, plus a heavy slug of Bezos's own cash.

The pitch is simple. Build for engineering and manufacturing what ChatGPT built for words. Bezos calls it an "artificial general engineer."

Instead of training on the internet, Prometheus trains on the laws of physics and real-world testing data pulled from manufacturers it won't name. The goal is to crush the time and cost of designing physical things, from skyscrapers to smartphones to jet engines to products sold on Amazon.

He's running it with Vik Bajaj, a former Google exec. About 150 employees, a big GPU cluster, and even Bezos admits compute is still hard to get.

The bigger play is quieter. He's reportedly trying to raise as much as $100 billion to buy industrial companies outright, run them on Prometheus models, and stack them into a Berkshire-style portfolio that makes physical goods cheaper and faster.

Manufacturing today means long lead times, expensive tooling, and slow iteration on new products. If Bezos pulls this off, the cost of getting a physical product to market drops and the speed goes up.

Cheaper prototypes. More SKUs. Faster launches. The same guy who reshaped how you sell is now aiming at how things get made.

✌🏼 See you again Thursday …

The answer to today’s STUMP BEZOS is
32 years to count to a billion, 31,500 years to a trillion