On June 24, 2026, Salesforce published the most consequential retail AI data point of the year. In its Agentforce Commerce announcement, the company revealed striking data from the 2025 holiday season. AI influenced 20 percent of global online sales, worth 262 billion dollars. Furthermore, retailers who ran their own shopper agents grew sales 59 percent faster than retailers on the sidelines. And AI-referred traffic converted at eight times the rate of social.
The announcement also included a timing signal that makes the data immediately relevant. The Agentforce Commerce integration with ChatGPT enters general availability this month. The integration with Google Search AI Mode and Gemini follows this summer. Specifically, the holiday season traffic driven by AI agents will run through infrastructure that retailers can deploy today. The retailers who build now will be operational before peak season. The ones who defer will be watching from the outside.
The 59 percent faster growth differential is not a feature gap. It is an operating model gap. The retailers who outperformed in the 2025 holiday season did not just appear in AI recommendations. They ran their own agents that managed the shopper journey from discovery through checkout with full customer and catalog context. That is a different deployment, built differently, with different outcomes.
What Salesforce Revealed This Week
The Agentforce Commerce release introduced three agent types that together cover the full commerce journey. The Shopper Agent handles conversational shopping on brand storefronts, managing discovery through checkout with full catalog and customer context. The Buyer Agent enables procurement via WhatsApp and SMS for B2B commerce. The Merchant Agent gives operations teams natural language control over catalog management, with early customers reporting 88 percent reductions in time-to-task.
PacSun was among the first retailers to deploy. Shirley Gao, Chief Digital and Information Officer at PacSun, described the OpenAI integration as an opportunity to extend PacSun products into AI platforms like ChatGPT. Her framing focused on transforming conversations into meaningful engagement with Gen Z and Gen Alpha customers. That is not a pilot framing. It is a channel strategy framing. And it reflects what the 59 percent differential actually represents.
The timing compounds the urgency. As I described in the AI-referred traffic conversion analysis, AI-referred shoppers convert at significantly higher rates than visitors from any other acquisition channel. The Salesforce data adds the competitive dimension: retailers running their own agents grow faster in aggregate because the agent manages the relationship from the first conversation to the repeat purchase.
Why 59 Percent Faster Is Not a Marketing Number
The 59 percent faster growth figure is not a feature comparison. It is an outcome of three structural differences between retailers who run their own shopper agents and those who rely on external AI platforms for discovery.
Gap 1: The Agent Has Full Customer Context
A retailer’s own shopper agent has access to the customer’s full purchase history, loyalty status, saved preferences, and real-time inventory. An external AI assistant recommending the retailer’s products has none of that. As a result, the retailer’s own agent can promise a real delivery date and honor a loyalty discount. It can also surface a product the customer bought before and complete the transaction with a saved payment method. The external agent can recommend and link. The difference in conversion outcome between those two interactions is structural, not marginal.
Gap 2: The Agent Closes the Loop at Every Channel
Retailers running their own shopper agents deploy them on their own storefronts and in their own apps. They also connect to AI platform integrations for ChatGPT, Google Search AI Mode, and Gemini. Consequently, when a shopper starts a conversation on ChatGPT and then returns to the retailer’s own site, the agent carries the context forward. The shopper does not start over. The recommendation the external agent made is reinforced by the retailer’s own agent with additional context. That continuity across the customer journey is what converts at eight times the rate of social. Social sends traffic. An agent carries a relationship.
Gap 3: The 2025 Holiday Season Already Proved It
The 59 percent growth differential is not a projection. It is a reported outcome from last season. The retailers on the right side of that differential already built and deployed their shopper agents before the 2025 peak. They spent the season operating while competitors were still evaluating. However, the baseline will be higher this year. More AI agents will be active, more shoppers will use them, and retailers who deployed in 2025 will have 12 months of agent learning and customer data compounding. The window to be in the first group rather than the second is now, not after the summer planning cycle.
The Adoption Layer: What Has to Change to Run Your Own Shopper Agent
The Salesforce platform exists and the integrations are available. What does not come with the platform is the organizational readiness to run an agent at production scale during peak season.
The Operating Model Has to Shift
A shopper agent running on a retailer’s storefront handles customer interactions that previously went to the search bar, the product page, the customer service chat, and in some cases the store associate. Each of those touchpoints had a different owner in the pre-agent organization. The e-commerce team owned search and product discovery, while customer service ran the chat and store operations managed the associate interaction. An agent that spans all of those simultaneously does not fit inside any single ownership structure. Therefore, retailers who deploy a shopper agent without redesigning the ownership model around it will discover the organizational friction during the first peak week, when the cost of fixing it is highest.
The KPIs Have to Change
The standard e-commerce KPIs measure what happens when a human navigates the retailer’s site. They do not measure what an agent does on behalf of a shopper. By contrast, the KPIs that matter for a shopper agent deployment are different. They include agent conversation completion rate, agent-initiated order value versus site average, repeat purchase rate for agent-acquired customers, and agent escalation rate to human support. Retailers who do not install those measurements before deployment will spend the post-season trying to understand whether the agent drove the growth or happened to run alongside it. The retailers who measure from day one will optimize with confidence.
The False Success Mode
The most predictable failure pattern is straightforward. A retailer deploys a shopper agent, measures session volume and first-week conversation counts, and declares the deployment successful. Then they never connect the agent to real-time inventory, live pricing, and the customer data platform. The agent will have conversations and make recommendations. However, those recommendations rely on stale catalog data and disconnected customer context. The result is a customer experience less useful than the one the retailer’s own search bar already delivered. Deploying the agent is not running the agent. Conversation volume is not conversion. A live agent without live data is not a shopper agent. It is a chatbot with a better interface.
Three Architecture Decisions Before the 2026 Holiday Season
The Customer Data Foundation
A shopper agent that delivers the 59 percent growth outcome has access to the customer’s full profile in real time: purchase history, loyalty tier, saved preferences, and payment methods. That access requires a customer data infrastructure the agent can query at the moment of the conversation, not one reconstructed from a batch sync that ran overnight. As I described in the Google Universal Commerce Protocol analysis, the customer data layer is a prerequisite for any agentic commerce deployment that delivers on its promise. Retailers whose customer data lives in disconnected systems without real-time API access cannot give a shopper agent the context it needs to outperform the search bar.
Catalog Quality for AI-Native Commerce
The Merchant Agent’s 88 percent reduction in time-to-task assumes a catalog structured for natural language queries. Consider a shopper asking the agent for a gift for a ten-year-old who likes science. The agent must read product descriptions, understand age relevance, and surface options that match that intent. That is not possible with product titles that are internal SKU codes or descriptions that are supplier sheet bullet points. Catalog quality for AI-native commerce is a different standard than catalog quality for keyword search. As I described in the agentic commerce infrastructure analysis, the retailers who invested in catalog quality before deploying agents are the ones seeing the growth differential. The ones who deploy the agent on top of a search-era catalog will see the agent underperform and blame the technology.
The Transaction Layer the Agent Can Complete
The conversion advantage of a shopper agent collapses at the checkout step when the transaction layer does not support agent completion. The agent needs to access saved payment methods, apply loyalty points, confirm delivery dates against real inventory, and complete the order without routing the shopper through a traditional checkout flow. Retailers whose checkout infrastructure does not expose those capabilities via API can get the customer to the cart. The agent cannot go further than that. It is not a shopper agent. The retailers who will capture the 2026 holiday season growth differential are the ones whose transaction layer can close when the agent asks it to.
What This Means for LatAm Retailers
The Salesforce data covers global commerce, but the application to Latin America requires a specific read. The Buyer Agent enables B2B procurement via WhatsApp and SMS, which is directly relevant to LatAm retailers whose wholesale and B2B customer relationships already run through WhatsApp. The infrastructure is the same channel their buyers already use. The agent adds structure, data capture, and transaction completion to conversations that currently end in an informal order that someone manually enters into a system.
For LatAm grocers, department stores, specialty chains, and distributors, the most immediate Agentforce Commerce application is the B2B Buyer Agent on WhatsApp, not the consumer Shopper Agent. That deployment does not require rebuilding the consumer storefront. It requires connecting the agent to the product catalog, the customer pricing rules, and the order management system. Most LatAm retailers already have all three. The barrier is organizational readiness, not technology availability.
Three Questions Worth Taking to the Next Leadership Meeting
Do you know which retailers in your category ran shopper agents during the 2025 holiday season? Not which ones announced AI initiatives. Specifically, which ones had a deployed agent managing customer conversations from discovery through checkout. If you cannot name them, you do not know what you are competing against in Q4 2026.
Can your customer data platform respond to a real-time query from an agent at the moment of a customer conversation? Not batch. Not overnight. At the moment the shopper asks the question. If the answer is no, the shopper agent you deploy will not have the context it needs to outperform your search bar, let alone to drive the 59 percent growth differential the data documents.
Who in your organization owns the decision to deploy a shopper agent? Moreover, do they have the authority to connect the agent to your customer data, your catalog, and your transaction infrastructure? If that decision requires sign-off from three separate VPs who have never worked on a shared project, the agent will not be operational before peak season. Organizational structure is the bottleneck most retailers do not identify until they have already missed the window.
The retailers who can answer all three with specifics are the ones building. The ones who cannot are the ones who will read the 2026 holiday season data and start the conversation about building for 2027.
Retailers who run their own shopper agents will own the 2026 holiday season. The ones who do not will compete for the traffic those agents send their way and wonder why conversion is lower than the year before.
If you are evaluating a shopper agent deployment or an agentic commerce architecture for your retail network, connect with me here or reach me on LinkedIn. I am happy to walk through the deployment framework we use across the U.S. and Latin America.
Adriana Rivas is a retail technology executive and AI strategist, and the founder of a U.S.-based hardware company specializing in self-service kiosks, POS systems, electronic shelf labels, and digital signage deployed across the United States and Latin America. She is the award-winning author of How to Implement Self-Service Without Failing (Amazon #1 Hot New Release, Silver Nonfiction Book Award 2025) and recipient of the Gold Stevie® Award, Thought Leader of the Year 2026. She is also recognized by Thinkers360 as a Top 10 Thought Leader in Retail and a Certified Master Expert in Retail.
Pingback: AI Shopping Assistant Consumer Trust: What You Need to Know