This week in AI commerce: Visa builds the on-ramp, Shopify hands over the keys
Visa launched a unified agentic payments platform, Shopify gave AI agents live store access, Juniper sized the market at $1.5T, and Meta shipped its first commerce model. Here's what happened April 7–11.
Every week, the line between “AI commerce is coming” and “AI commerce is here” gets harder to draw. Lumio’s weekly rundown covers the moves that matter — new protocols, platform launches, legal fights, and market signals — with context on what they mean for brands.
Here’s what happened the week of April 7–11, 2026.
Payments infrastructure: the rails are being laid
Visa launches Intelligent Commerce Connect
Visa unveiled Intelligent Commerce Connect on April 9 — a unified platform that lets brands accept AI agent-initiated payments across every major agentic protocol (ACP, UCP, Trusted Agent Protocol, Machine Payments Protocol) through a single integration on Visa’s acceptance network.
The key detail: it’s protocol-agnostic and network-agnostic. Brands can accept payments from AI agents using both Visa and non-Visa cards. Instead of integrating with each protocol separately, you plug into Visa’s platform once and get coverage across all of them.
Currently in pilot with select partners. Broader rollout later this year. For brands watching the protocol fragmentation problem (ACP, UCP, Agent Pay, APOP — four protocols and counting), this is the first serious attempt at an abstraction layer.
Mastercard completes live agentic transactions across APAC
Mastercard completed its first live agentic transaction in Hong Kong on March 27, working with HSBC and DBS Hong Kong. An AI agent booked a ride to Hong Kong International Airport through mobility provider Hoppa — end to end, no human in the loop for the payment.
The security model uses Mastercard Agent Pay: each agent gets a unique agentic token, consumer consent is explicitly captured, and purchase confirmation runs through Mastercard Payment Passkeys. The system is now live across eight APAC markets — Australia, New Zealand, Singapore, Malaysia, India, South Korea, Taiwan, and Hong Kong — with a new AI Center of Excellence in Singapore backing the expansion.
UnionPay enters with its own protocol
Not to be left out, UnionPay launched the Agentic Payment Open Protocol (APOP) on April 3 — the fourth major payment network to publish an agentic commerce standard. APOP covers agent identity lifecycle management, intent verification, single sign-on across brands and financial institutions, and payment authorization.
The first live transaction: a taxi booking in Hong Kong via AI assistant developer Evonet and ride-hailing platform Hoppa. If the Hong Kong taxi-to-airport use case sounds familiar, it’s because Mastercard’s first live agentic transaction was the same scenario with the same partner. Different rails, same ride.
The pattern
Four payment networks. Four protocols. Two of them demonstrated with the same taxi booking. The payments industry is moving fast but hasn’t converged on standards — which is exactly why Visa’s protocol-agnostic approach matters. Brands who wait for a single winner will wait a long time. The smart play is infrastructure that abstracts the protocol layer away.
The number: $1.5 trillion by 2030
Juniper Research published the first major market sizing report for agentic commerce on April 7, projecting $1.5 trillion in global spend by 2030 — up from “only pilot deployments” in 2025 and 2026.
The report ranks payment infrastructure readiness: Mastercard first, Visa second, Stripe third — based on agentic-specific capabilities and protocol participation across 14 providers.
Two findings worth anchoring on. First: trust is the number one barrier to adoption, ahead of technology and regulation. Second, and more sobering: agentic commerce “will not replace traditional eCommerce checkouts for the foreseeable future.” The $1.5T figure is incremental, not replacement. Anyone selling agentic commerce as the death of checkout pages is ahead of the data.
Platform moves: Shopify, Meta, and OroCommerce
Shopify AI Toolkit goes live
Shopify launched its AI Toolkit on April 9 — a free, open-source plugin that connects AI coding agents directly to the Shopify platform. Claude Code, Cursor, Gemini CLI, VS Code, and OpenAI Codex can now execute live store operations in plain English: update products, adjust inventory, change pricing, scaffold apps, generate validated GraphQL queries, and build UI extensions.
This is the first major ecommerce platform to give external AI agents full operational control of a live store. Agents get real-time API schemas and code validation instead of relying on stale training data. The capability is real — and so is the risk. Changes execute immediately on live stores. No draft mode. No undo. Shopify recommends testing on development stores first, which is good advice that will be universally ignored.
Meta ships Muse Spark
Meta launched Muse Spark this week — the first model from Meta Superintelligence Labs. It powers product recommendations, price tracking, and alternative surfacing directly within Instagram and Facebook feeds.
The shopping experience uses product carousels with brand information, pricing, website references, and bullet-point explanations for why each item fits the query. A new checkout flow built with Stripe and PayPal lets users complete purchases in a single tap. Meta also expanded creator affiliate partnerships to include Amazon, eBay, and Temu in the US, plus Mercado Libre and Shopee internationally.
The strategic play: Meta is embedding commerce into the social feed rather than building a separate shopping surface. If your products show up in Instagram’s AI recommendations, the shopper never leaves the app. That means your product data needs to be structured well enough for Meta’s model to reason about — the same challenge that ACP, UCP, and every other agentic surface presents.
OroCommerce 7.0: AI comes to B2B
OroCommerce shipped version 7.0 on April 9 with platform-native AI co-pilots that run on unified commerce data. The AI layer, called OroIQ, provides SmartAssistant and SmartInsights through a natural language interface — query data, execute tasks, run analytics, all from one conversational UI.
Notable: OroIQ includes MCP server support, meaning external AI tools can integrate while maintaining governance over how they interact with business data. There are no additional AI fees — it’s part of the standard platform. SmartOrder is already reporting 20%+ productivity gains with early adopters.
B2B has lagged behind D2C in AI commerce adoption, partly because the data is messier and the workflows are more complex. Oro shipping production-ready AI with a full governance layer — rather than bolting on a chatbot — is a signal that the B2B side is catching up.
Legal and M&A: the fights over control
Perplexity vs. Amazon heads to appeals court
Perplexity filed a federal appeal on April 1 seeking to overturn the injunction banning its AI shopping agent, Comet, from accessing Amazon. Amazon had sued in November 2025, claiming Comet accessed password-protected customer accounts while disguising itself as a Google Chrome browser.1
Perplexity’s core argument: “A Comet user accessing Amazon from her own computer is no more equivalent to Perplexity accessing Amazon than a Safari user accessing Amazon is equivalent to Apple accessing Amazon.” The company also points out Amazon presented no evidence of actual harm during the eight months Comet operated freely.
Amazon’s response is due April 22. This case will set precedent for whether third-party AI agents can operate on major platforms without explicit platform authorization — a question that affects every agentic commerce player, not just Perplexity.
Rezolve AI goes hostile on Commerce.com
Rezolve AI bypassed Commerce.com’s board on April 8, taking a takeover proposal directly to shareholders after negotiations collapsed. The offer: one Rezolve share for every two Commerce.com shares, pitched as creating a “$700M+ global agentic commerce powerhouse.”
Commerce.com’s board rejected the bid as undervalued. Both stocks dropped — Commerce.com by 7.3%, Rezolve by 6.2%. The market, apparently, is not yet convinced that “agentic commerce powerhouse” is a category worth paying a premium for.
Also on the radar
Google is folding Doppl into Search. The standalone virtual try-on app shuts down April 30, but the technology moves into Google Shopping and Search natively. Users will be able to tap any product result and try it on. Launch partners include L’Agence, Zalando, and Zara. Separately, Shopify integrated Genlook’s AI virtual try-on app into its platform, and retailers from ASOS to Zara are reporting early conversion gains from the technology.
Alibaba’s Accio crossed 10 million monthly active users in March, meaning roughly one in five Alibaba users now consults AI for product sourcing. The company also launched Accio Work, an enterprise AI agent fleet covering market analysis, design, sourcing, store optimization, and inventory monitoring. Mike McClary, an Illinois-based seller, used Accio to redesign his Guardian LTE flashlight and source a Ningbo manufacturer that cut per-unit costs from $17 to about $2.50 — an 85% reduction — relaunching the product within a month.
What it all means
Three themes from this week:
1. The payments layer is racing ahead of the commerce layer. Visa, Mastercard, UnionPay, and Stripe are all shipping live agentic payment infrastructure. Meanwhile, the commerce protocols those payments ride on (ACP, UCP) are still maturing, brand adoption is early, and accuracy problems persist. The rails are being built faster than the trains.
2. Platform control is the real battleground. Shopify handing AI agents the keys to live stores. Meta embedding commerce into social feeds. Amazon blocking external agents entirely and fighting Perplexity in court. OroCommerce shipping AI with governance rails. Each platform is making a bet about how much control to cede — and to whom.
3. The data quality gap is the constant. Every story this week — Visa’s protocol abstraction, Shopify’s agent toolkit, Meta’s product carousels, Juniper’s $1.5T projection — assumes that brands have structured, accurate, AI-readable product data. Most don’t. The brands who close that gap first capture the value from all of these developments simultaneously.
If you want to know where your product data stands across the surfaces that matter — ACP, UCP, Google Merchant Center, Meta, and the rest — get in touch. Lumio shows you exactly what to fix.
