Consumer Goods AI: Agentic Commerce Standards Take Shape

The consumer goods AI cycle ran wider than the two launches that drew the most attention. While Albertsons and Salesforce put the spotlight on paid placement and brand-run agents, the topic of our feature this week, the rest of the cycle was spent settling how agents will actually check out, and counting how many retailers have already committed. Here is what else moved.

Google Built One Cart That Follows Shoppers Across Its Whole Footprint

At Google I/O on May 19, Google introduced Universal Cart, a single shopping cart that travels with a shopper across Search, Gemini, YouTube, and even Gmail. Add an item from any of those surfaces and the same cart picks it up, then works in the background to track price history, flag price drops, and ping the shopper when something out of stock returns.

For a consumer brand, that is a different kind of entry point than a search result or a product page. The cart is tied to a person rather than a session, so the contest to be the product it suggests, or the deal it surfaces, plays out across a shopper's whole day on Google instead of one visit to one site.

Google did not ship the cart alone. It launched with Nike, Sephora, Target, Ulta Beauty, Walmart, Wayfair, and Shopify merchants including Fenty and Steve Madden, and Google said its agent-assisted checkout will expand to Canada and Australia in the coming months, then the U.K., with YouTube, hotel booking, and local food delivery next. McKinsey, cited in the coverage, put the agentic commerce opportunity at roughly $5 trillion by 2030.

The takeaway for retail and DTC operators: the assistant is not the only new surface worth a brand's attention. The cart itself is becoming a place where products get compared, discounted, and chosen, often without the shopper ever loading the brand's own store.

Source: Google (reported by TechCrunch and The Next Web)

Two Open Standards Are Racing to Wire Up Agentic Checkout

Beneath the product launches, a quieter contest is deciding how an agent actually completes a purchase. Two open standards now anchor it. Google co-developed the Universal Commerce Protocol (UCP) with a group of retailers as a shared language for agentic checkout, paired with the Agent Payments Protocol (AP2), which lets a shopper set guardrails, specific brands, specific products, a spending limit, before an agent is allowed to pay.

On the other side, OpenAI and Stripe open-sourced the Agentic Commerce Protocol (ACP), the standard behind ChatGPT's Instant Checkout, which began letting U.S. shoppers buy from Etsy sellers and a growing set of Shopify merchants such as Glossier, SKIMS, and Vuori directly in chat. PayPal is standing up its own ACP server to bring its vast base of small businesses onto the same standard this year.

The standards are still being shaped in public, and the early going has been messy: rival agent-payment rails that still do not interoperate are a reminder that betting a roadmap on one camp too early carries its own risk. OpenAI recently said it is moving Instant Checkout into its Apps surface, where it expects purchases to flow more naturally, a sign that the winning shape of agent checkout is being worked out by trial rather than declared up front. For brands, that argues for supporting the open, interoperable standards rather than locking into any single vendor's interface.

This is also where the build-or-buy question gets concrete. A brand that wants its catalog, pricing, and order systems reachable by agents, and its own customer service automation answering across these surfaces, does not have to adopt one company's full system to get there. Model- and platform-agnostic infrastructure that runs on open standards like MCP, AgentPMT included, lets a brand make its commerce APIs agent-callable once and plug into whichever assistant a shopper happens to use, then wire in the spending limits and human approval steps it wants around what those agents can do.

Source: OpenAI and Stripe

NVIDIA's Survey Shows Retail and CPG Already Committed

The demand side of all this is not a forecast. NVIDIA's State of AI in Retail and CPG survey, released earlier this year, found 91% of retail and consumer-goods organizations already using or assessing AI, with nearly all of them planning to raise their AI budgets in the year ahead. The overwhelming majority credited AI with lifting annual revenue, and almost everyone surveyed said it had cut costs.

Agentic AI is the part moving fastest. Nearly half, 47%, are using or assessing agents, with a sizable share already running them in production and a comparable group expecting to within a year. The top goals are speed and efficiency, customer experience and personalization, and real-time decision-making, and a large block of respondents said AI had already improved customer service, the area where automated customer service tends to land first.

The survey was not all tailwind. A majority reported worse supply-chain pressure year over year, from geopolitical instability to labor constraints to tightening regulation, and most respondents framed open-source models as important to scaling without lock-in. As Chris Walton of Omni Talk put it in the report, retailers who succeed "start with boring use cases solving specific P&L problems, prove value, then scale."

Read against this week's launches, the picture explains the rush. Vendors shipped agentic-commerce products into a market that had already opened its budgets and started its pilots, which is why a single week produced this many moves at once.

Source: NVIDIA

Sources

• Google Shopping introduces Universal Cart, Google
• Google's new Universal Cart wants to follow your entire shopping journey across the internet, TechCrunch
• Google launches Universal Cart and updates AP2 at I/O 2026, The Next Web
• Buy it in ChatGPT: Instant Checkout and the Agentic Commerce Protocol, OpenAI
• Stripe powers Instant Checkout in ChatGPT and releases the Agentic Commerce Protocol, Stripe
• State of AI in Retail and CPG 2026 Survey Report, NVIDIA