Originally published on The Searchless Journal
One Week That Killed the Shopping Cart
Between April 13 and April 17, 2026, the retail industry shipped more agentic commerce functionality than the previous twelve months combined. Walmart announced a partnership with OpenAI. Puma deployed an AI concierge in its Las Vegas flagship. Amazon Rufus gained the ability to buy from third-party retailers on your behalf. David's Bridal launched shopping inside ChatGPT and Microsoft Copilot. Shopware baked agentic features into its core platform. American Express shipped a developer kit for agent-driven payments.
Six major moves. Five days. One theme: the shopping cart is being replaced by an AI agent that discovers, compares, negotiates, and buys on your behalf.
This is agentic commerce, and it is no longer a slide in a venture deck. It is live infrastructure at scale.
What Happened
Walmart x OpenAI: From Question to Cart
On April 17, Walmart formally announced its partnership with OpenAI to bring AI-first shopping to consumers via ChatGPT. The centerpiece is "Sparky," a shopping agent that takes a customer from an open-ended question ("What do I need for a weekend camping trip under $200?") directly to a curated, purchasable cart.
Sparky does not redirect you to walmart.com. It completes the transaction inside the conversational interface. The agent pulls from Walmart's product catalog, applies available discounts, checks inventory at nearby stores, and presents a single "confirm purchase" prompt. The customer never visits a product page, never browses a category, never uses a search bar.
This is the model that kills the funnel. There is no top, middle, or bottom. There is a question and a purchase.
Amazon Rufus: Buying From Competitors
On April 16, Amazon's AI assistant Rufus received its most aggressive update yet: "Buy For Me." The feature allows Rufus to purchase products from third-party retailers outside Amazon's marketplace, completing the transaction without the customer ever leaving the Amazon app.
The implications are significant. Amazon is positioning itself not just as a store but as the agent layer on top of all commerce. If Rufus can buy from any retailer, Amazon becomes the default shopping interface regardless of where the product lives. For retailers, this is both an opportunity (new demand channel) and a threat (the customer relationship belongs to Amazon's agent, not to you).
Puma's AI Concierge "Dylan"
Puma took a different route. On April 17, the brand debuted "Dylan," an AI-powered digital human concierge at its Las Vegas flagship store. Dylan is not a chatbot. It is a full-screen interactive presence that greets customers, understands natural language requests about products and sizing, and guides them through the store experience.
Dylan represents the physical retail counterpoint to digital agentic commerce: AI agents are not just online. They are becoming the interface between brands and customers in physical spaces too. The in-store experience is being rebuilt around conversational AI, not the traditional sales associate model.
David's Bridal: Shopping Inside ChatGPT
On April 14, David's Bridal launched shopping directly inside ChatGPT and Microsoft Copilot. Customers describe their event, budget, style preferences, and body measurements, and the AI agent curates a selection of dresses available for immediate purchase.
Wedding dresses are high-consideration purchases with long research cycles. If agentic commerce works here, it works everywhere. David's Bridal is betting that customers will trust an AI agent to narrow a field of hundreds of dresses to three or four options, then buy one without ever visiting the website.
The bet is less wild than it sounds. Bridal shoppers already rely heavily on recommendations from friends, family, and consultants. An AI agent is a natural extension of that behavior, scaled to the entire inventory.
Shopware 6.7.9: Agentic Commerce Infrastructure
On April 17, Shopware released version 6.7.9 with built-in agentic commerce features. This matters because Shopware powers thousands of mid-market online stores across Europe and North America. They are not building a single agent; they are giving every merchant on their platform the tools to become agent-discoverable.
Shopware's approach includes structured product feeds designed for AI consumption, agent-friendly APIs for inventory and pricing, and integration hooks for major AI platforms. It is infrastructure play, not a consumer feature, and it signals that agentic commerce is becoming a platform-level concern, not a bespoke integration.
American Express: The Payment Layer
On April 14, American Express launched a developer kit for agentic commerce. The kit gives AI agents the ability to initiate, authenticate, and complete Amex payments on behalf of cardholders, with consent frameworks and fraud protections built in.
Amex is solving the hardest problem in agentic commerce: the payment. An agent can recommend products all day, but if it cannot securely complete a transaction, the experience collapses at the finish line. By shipping a developer kit, Amex is inviting the entire ecosystem to build on its payment rails. Expect Visa and Mastercard to follow with their own agent payment protocols.
The Numbers Behind the Shift
Two data points frame the gap between where consumers are and where the infrastructure is headed.
Commercetools reports that 73% of consumers now use AI in some part of their shopping journey. That includes product discovery, price comparison, review summarization, and sizing recommendations. AI is already woven into how people shop.
But only 13% have completed a purchase led entirely by an AI agent. The intent is there. The infrastructure is catching up. The week of April 13-17 was the week the infrastructure arrived.
Forrester adds a B2B dimension: they predict that by the end of 2026, 20% of B2B sellers will face agent-led negotiations. Procurement agents that compare specs, negotiate pricing, and execute purchase orders autonomously. The same pattern playing out in consumer retail is hitting wholesale and enterprise procurement on a parallel track.
The Protocol Question
With six major players launching agentic features in one week, the obvious question is: which protocol wins?
Amazon wants Rufus to be the agent layer. OpenAI wants ChatGPT to be the agent layer. Google is building the Universal Commerce Protocol with Walmart, Target, Shopify, and Etsy. Amex wants to own the payment layer regardless of which agent wins. Shopware wants every merchant to be agent-ready regardless of which agent wins.
This is a replay of the mobile payments wars of the early 2010s. Apple Pay, Google Wallet, Samsung Pay, and a dozen others all competed for the same tap-to-pay moment. Eventually, the market settled on coexistence: multiple wallets, one NFC standard. Agentic commerce is heading the same direction. Multiple agents, multiple protocols, but a shared expectation that every retailer will expose a clean, machine-readable product and transaction layer.
Retailers that wait for a single winner will be too late. The right move is to become agent-discoverable now, across multiple protocols, and treat each agent as a new demand channel the way you treat Google Shopping, Amazon Marketplace, and social commerce today.
The protocol question also has a measurement dimension that most retailers are ignoring. Traditional analytics track sessions, pageviews, and cart abandonment. Agentic commerce generates none of those events. An AI agent queries your product catalog, evaluates options, and either completes a purchase or moves on. There is no session. There is no pageview. There is a transaction (or the absence of one) and a recommendation event that you may never see. The retailers that build agent-specific analytics early will have a six-month data advantage over those that discover the gap after the volume scales.
What Retailers Must Do
1. Build Agent-Readable Product Feeds
AI agents cannot browse your beautiful category pages. They need structured data: product attributes, availability, pricing, specifications, and shipping options in machine-readable formats. Schema.org markup is table stakes. Agent-specific APIs are the next step.
2. Optimize for Agent Discovery
When a customer asks ChatGPT "best running shoes under $100," the agent surfaces products based on structured data, reviews, and brand authority. This is not traditional SEO. This is GEO, Generative Engine Optimization, and it requires a fundamentally different approach to content, structure, and authority.
3. Prepare for Agent-to-Agent Negotiation
Forrester's B2B prediction applies to consumer retail sooner than most expect. Price comparison agents will negotiate with retailer agents in real time. Dynamic pricing strategies need to account for autonomous buyers that have perfect information and zero brand loyalty.
This changes pricing strategy fundamentally. Today, most retailers set prices based on competitive intelligence gathered weekly or daily. Agent-to-agent negotiation happens in milliseconds. A customer's agent queries ten retailers simultaneously, evaluates shipping times, return policies, loyalty programs, and price, then recommends the best option. If your product is not structured for this kind of instant evaluation, you lose before the customer even knows you exist.
The retailers that invest in real-time pricing APIs, transparent inventory feeds, and machine-readable policy documents will be the ones that show up in agent recommendations. Everyone else becomes invisible.
4. Own the Consent Layer
The brands that win in agentic commerce will be the ones that earn direct consent from the customer, not just the agent. If your only relationship is through Amazon's agent, Amazon owns the customer. If the customer authorizes your brand's agent directly, you own the relationship.
The Bigger Picture: Google's Zero-Click Enclosure
Agentic commerce does not exist in isolation. As Google's AI Mode reshapes search into a zero-click experience, retailers face a double squeeze: Google intercepts discovery, and AI agents intercept the transaction. The traditional click-to-visit-to-cart pipeline is being dismantled from both ends.
We analyzed this pattern in detail in our breakdown of Google AI Mode's split-screen zero-click enclosure. The retailers that survive will be the ones that show up inside AI outputs, not just in search results.
Who Moves Next
The week of April 13 established the pattern. Expect the following moves in the next 90 days:
- Shopify will announce agentic commerce tools for its merchant base. The platform has too much to lose by letting Shopware own this category.
- Google will expand the Universal Commerce Protocol beyond its founding partners, likely adding major fashion and electronics retailers.
- Payment networks beyond Amex (Visa, Mastercard, Stripe) will ship competing agent payment kits.
- At least one major luxury brand will launch an AI agent for personalized shopping, following Puma's lead in physical retail.
The pace is accelerating because the infrastructure is finally ready. Large language models can handle complex multi-step transactions. Payment APIs can authenticate agents. Product catalogs can be structured for machine consumption. The building blocks existed in 2025. In April 2026, someone finally assembled them.
What makes this moment different from previous commerce shifts (mobile, social, marketplace) is the speed of adoption. Mobile commerce took five years to go from novelty to majority. Social commerce is still climbing after seven years. Agentic commerce has the advantage of building on all three: it uses the mobile device as the interface, the social graph as context, and marketplace infrastructure as the transaction layer. The adoption curve will be steeper because the foundation is already in place.
The Cart Is Dead. The Agent Is the Interface.
The shopping cart was a metaphor for a particular kind of commerce: browse, select, add to cart, check out. Each step was a page, each page was a conversion opportunity, each conversion opportunity was a chance to upsell or cross-sell.
Agentic commerce collapses all of that into a single conversational exchange. The customer describes intent. The agent handles everything else. The cart never loads because the cart is never needed.
For retailers, this is the biggest shift since mobile. The window to become agent-discoverable is open now. It will close as protocols standardize and incumbents lock in their positions.
Is your brand discoverable by AI agents? Run a free AI visibility audit at audit.searchless.ai to see how your products appear (or don't) in ChatGPT, Google AI Mode, Perplexity, and other AI platforms.
Sources
- Walmart + OpenAI partnership announcement, April 17, 2026
- Amazon Rufus "Buy For Me" feature launch, April 16, 2026
- Puma AI concierge "Dylan" debut, Modern Retail, April 17, 2026
- David's Bridal ChatGPT and Copilot shopping integration, Retail Dive, April 14, 2026
- Shopware 6.7.9 release notes, April 17, 2026
- American Express agentic commerce developer kit press release, April 14, 2026
- commercetools: "AI Trends in Agentic Commerce" consumer survey (73% AI usage, 13% agent-led purchases)
- Forrester: "Agentic Payments in B2C and B2B Commerce" prediction brief (20% B2B agent-led negotiations by end 2026)
Learn how to make your brand visible in AI-generated answers at searchless.ai/ai-visibility.
Top comments (0)