In September 2025, OpenAI made the biggest agentic commerce retail announcement of the year. Instant Checkout launched inside ChatGPT with Etsy as the launch partner and a promise of over one million Shopify merchants to follow. The infrastructure was real. The partners were credible. The direction seemed clear: the agent would find the product, confirm the price, and complete the purchase without the shopper ever leaving the conversation.
By March 2026, OpenAI ended Instant Checkout in favor of dedicated retailer apps inside the chatbot. Forrester principal analyst Emily Pfeiffer put the actual count at roughly 30 Shopify sellers live at that point, not the million-plus promised. Walmart, evaluating its own AI commerce data, found conversion three times lower for in-chat purchase than for redirecting shoppers to its own site. The agent could complete the transaction. The shopper preferred to complete it somewhere else.
Year One of agentic commerce did not produce fully autonomous buying at scale. It produced something more specific and, for retailers who understood it, more useful. The dominant pattern across 2025 and 2026 is agent-assisted discovery followed by checkout on the retailer’s own site. That pattern has different infrastructure requirements, different loyalty implications, and a different competitive advantage than the fully autonomous buying scenario that most vendor pitches described.
What the Data From Year One Actually Shows
The volume numbers are real. Adobe Analytics recorded 693 percent year-over-year growth in AI-sourced retail traffic during the 2025 holiday season. Salesforce documented $67 billion in AI-influenced Cyber Week sales, roughly 20 percent of all digital orders. AI-sourced retail traffic surged over 1,200 percent for retailers year over year, while traditional search traffic declined 10 percent. ChatGPT processes 50 million shopping queries daily. These are not projections. They are measurements of behavior that is already happening.
However, the conversion data tells a more precise story. AI-referred visitors convert at higher rates than any other channel when they complete the purchase on the retailer’s own site. The friction that kills the conversion is not the AI recommendation. It is the in-chat checkout experience itself. When the shopper has to enter payment details, confirm a shipping address, and create or link an account inside a chat interface not built for commerce, they abandon. When the agent recommends the product and the shopper clicks through to a familiar checkout on the retailer’s own platform, they convert.
Consequently, the retailers winning in AI commerce right now are not the ones who built the most sophisticated in-chat checkout. They are the ones who made their products discoverable to the agent and their own checkout trustworthy enough to convert the click-through.
Why the Vendor Pitch Got Ahead of the Consumer Behavior
The Trust Gap That Research Confirmed
The consumer behavior data from Year One aligns precisely with the trust research I described in the AI shopping trust gap analysis. VoCoVo and Alchemer found that 48.5 percent of shoppers used AI to research purchases in the past year. Only 35.4 percent trusted the AI enough to act on it without additional confirmation. That is a 13-point trust gap between research behavior and purchase behavior. Shoppers use the agent to narrow the options. They use their own judgment, and often their own familiar checkout environment, to complete the transaction.
Additionally, IBM and NRF surveyed more than 18,000 consumers across 23 countries in January 2026. They found that 73 percent use AI somewhere in their buying journey. However, only 45 percent use it for at least part of the buying process. A significant portion uses AI only for the research and discovery phase. The agent is doing the work shoppers found tedious: comparison, research, narrowing. It is not yet replacing the moment of commitment that shoppers still want to control.
The Infrastructure That Was Not Ready
The Instant Checkout failure was not only a consumer behavior story. It was also an infrastructure story. OpenAI moved to dedicated retailer apps inside the chatbot. The alternative, a generic checkout layer across thousands of diverse merchant backends, produced too much friction and too many failure modes to scale. As I described in the agentic commerce inventory accuracy analysis, reliable agentic transactions require three things clean and connected in real time: product data, inventory signal, and checkout infrastructure. Most merchant backends are not built for that standard. The checkout failures were not random. The data architecture underneath them predicted them.
Furthermore, merchants with 95 percent or higher data fill rates on core product attributes saw dramatically better agent discovery outcomes. Below 80 percent, AI agents routinely skipped those products entirely. The technical standard for agentic commerce is higher than the standard for search-based commerce. Most retailers entered Year One without meeting it.
What the Retailers Who Got It Right Did Differently
They Built for Discovery First and Transaction Second
The clearest case study of the Year One playbook executed correctly is the one I described in the Sephora AI commerce analysis. Sephora launched in ChatGPT in March 2026 with full loyalty integration and no checkout. They built the discovery and recommendation layer first. They made the AI experience more personalized and more useful than a generic search by connecting 80 million Beauty Insider profiles to the ChatGPT interface. Then, 72 days later, they added Google Agentic Checkout for the shopper who was ready to complete the transaction. The sequencing was not a delay. It was a strategy built on an accurate read of where consumer trust was in Year One.
They Maintained the Customer Relationship Through the Agent Layer
The retailers who built durable AI commerce strategies in Year One understood the data ownership question before they integrated. As I described in the AI retail discovery and transaction analysis, the critical question in AI commerce is not whether you can sell inside a third-party platform. It is whether you retain the customer relationship when you do. Loyalty integration, first-party data connectivity, and checkout architecture that brings the shopper back to the retailer’s own environment are not defensive moves. They are the structure that makes AI-driven discovery commercially valuable rather than a new source of traffic that converts on someone else’s platform.
They Measured the Right Things
The retailers reporting strong Year One AI commerce results measured AI-referred conversion rates separately from other channel conversion rates. Adobe documented that AI-referred visitors generate 38 percent higher revenue per visit than visitors from other channels. That signal is only visible if the analytics capture it. Retailers who measured total traffic without attribution could not see the AI-referred conversion advantage and could not optimize for it. Moreover, retailers who measured in-chat checkout conversion without measuring redirect conversion missed the pattern that Walmart found. The redirect was outperforming the in-chat experience by 3x. You cannot fix what you cannot see.
What This Means for the Next Twelve Months
The Year One data suggests a specific architecture for the AI commerce strategy that will produce returns in Year Two and beyond. It has three components that have to work together. Product data quality at 95 percent or higher fill rate on core attributes, so AI agents can discover, describe, and recommend accurately. Loyalty integration into the AI discovery layer, so the customer relationship stays with the retailer even as discovery moves to third-party AI interfaces. And checkout architecture that converts the AI-referred click-through on the retailer’s own platform, where trust is higher and conversion follows.
In other words, fully autonomous agent buying is not the near-term opportunity. Before Groceryshop 2026 opens, one question determines whether your AI commerce strategy is built on Year One evidence or on the 2025 vendor pitch. The near-term opportunity is the 1,200 percent growth in AI-sourced retail traffic. Those shoppers arrive with higher purchase intent than any other channel. The checkout experience that meets them either converts them or loses them. Both companies keynoting at Groceryshop are describing where agentic commerce is going. The Year One data describes where it is now.
What This Means for LatAm Retailers
The Year One pattern of agent-assisted discovery followed by checkout on the retailer’s own site is the correct starting architecture for most LatAm retailers. Fully autonomous in-chat checkout requires payment infrastructure, regulatory compliance, and consumer trust levels still developing across many LatAm markets. By contrast, the discovery layer is already active. WhatsApp AI, Google AI Mode, and TikTok Shop are all operating in LatAm markets with meaningful engagement. A LatAm retailer who makes products discoverable in those interfaces and builds a clean redirect to their own checkout is executing the Year One playbook correctly.
Furthermore, the product data quality standard matters as much in LatAm as in the United States. AI agents routinely skip products from merchants below 80 percent data fill rates on core attributes regardless of which market they operate in. That is an infrastructure investment with a clear return: be visible to the agent or be invisible to the fastest-growing discovery channel in retail.
The Honest Question About Your AI Commerce Strategy
What the Year One Data Tells You to Check
The vendor pitch for agentic commerce in 2025 was about autonomous buying. The Year One data is about discovery quality, data fill rates, loyalty integration, and redirect conversion. Those are different conversations. They require different infrastructure investments and different measurement frameworks.
Is your AI commerce investment designed to make your products more discoverable and your own checkout more convertible for AI-referred shoppers? Or is it designed to complete autonomous transactions in an environment where Year One data shows shoppers are not yet ready to do that at scale? The answer determines whether the next grocery shop announcement becomes an action item or another aspiration.
The agent is getting better every quarter. Consumer trust in autonomous AI buying will grow. The Year One data does not say that fully autonomous commerce is wrong. It says it is not yet the primary opportunity. The primary opportunity right now is the 1,200 percent growth in AI-sourced retail traffic. Those shoppers arrive with higher purchase intent than those from any other channel. The checkout experience that meets them either converts them or loses them. That opportunity exists today. Most retailers have not yet built for it.
If you are evaluating your AI commerce strategy or building the product data, loyalty, and checkout infrastructure that Year One showed actually drives returns, connect with me here or reach me on LinkedIn. I am happy to walk through the framework we use across the U.S. and Latin America.
Adriana Rivas is a retail technology executive and AI strategist. She is the recipient of the Gold Stevie® Award, Thought Leader of the Year 2026, and recognized by Thinkers360 as the #7 Global Thought Leader in Retail. She is the author of How to Implement Self-Service Without Failing (Revised and Expanded Edition, September 21, 2026)