A GA Stamp Timed for the Holidays
On June 24, Salesforce declared its biggest Agentforce Commerce release yet, moving its Shopper, Buyer and Merchant agents to general availability. This is the moment where agentic commerce stops being a keynote slide and becomes something a retailer can switch on. The three agents now connect natively to inventory, order management and customer data, which means they can answer questions, build baskets and complete purchases without the usual handoffs between a chatbot and the checkout flow. Salesforce is also shipping Storefront Next, a template that it claims can stand up a production storefront in under thirty minutes.
The timing is not accidental. Salesforce is pushing this out before the 2026 peak season, when retailers make their technology bets for the fourth quarter. Nitin Mangtani, executive vice president and general manager of Agentforce Commerce, framed the season as a deadline: brands that win, he argued, will have a Shopper Agent live on their properties for the 2026 holidays. That is a sales pitch, but it is also a reasonable read of where the market is heading. Once a competitor's storefront can hold a conversation and close a sale, a static product grid starts to look like a liability.
Three Agents, Three Jobs
The Shopper Agent is the consumer-facing piece. It lives on the storefront and runs discovery through checkout and post-purchase service, which is why Salesforce keeps pointing to add-to-cart and conversion lifts. The Buyer Agent is the quieter but arguably more interesting release: it handles B2B procurement over WhatsApp and SMS, supports multiple carts and round-trip quoting, and lets a purchasing manager reorder without ever logging into a portal. For enterprises whose B2B ecommerce still runs on clunky punchout catalogs, moving the transaction into a messaging thread is a genuine change in workflow.
The Merchant Agent works in the back office, managing catalog and promotions and reacting to demand trends. Taken together, the three agents describe a view of the world in which the merchandiser, the shopper and the wholesale buyer each get their own autonomous assistant, all drawing on the same commerce data layer. That coherence is the real product. Point solutions have flooded the retail AI market for two years, and the pitch here is not any single agent but the fact that they share context, so a promotion the Merchant Agent creates is immediately something the Shopper Agent can surface.
The ChatGPT and Google On-Ramp
The headline integration for most executives will be ChatGPT, which reaches general availability in July 2026. Salesforce says retailers can sync a product catalog straight to ChatGPT from Business Manager, without third-party middleware or custom development. That is the part that matters. The friction in agentic commerce has never been the AI conversation, it has been getting a live, accurate catalog with inventory and pricing into someone else's assistant. A one-click catalog sync from the commerce platform that already runs the store removes an integration project that would otherwise take a quarter.
Google is next. Discovery and purchase across Google Search, including AI Mode, and the Gemini app are slated for general availability later in the summer, alongside an Agentic Commerce Search capability that Salesforce also dates to July. The strategic logic is that a retailer should not have to choose which assistant to bet on. By making the catalog portable across ChatGPT and Google, Salesforce is positioning itself as the neutral commerce backbone underneath whichever agent a shopper happens to use. That is a defensible place to stand, and it is a direct answer to retailers worried about building homegrown connectors for every AI platform.
The Numbers Doing the Persuading
Salesforce is arming its sales teams with data that is hard to ignore. During the 2025 holidays, the company says AI influenced twenty percent of global online sales, a market it sizes at 262 billion dollars. More pointedly, it claims retailers running shopper agents grew sales fifty-nine percent faster than those that had not yet deployed them, and that AI-referred traffic converts at eight times the rate of traffic from social media. Whatever the precise methodology, the direction of travel is consistent with what Amazon and Walmart have reported from their own assistants.
We would treat the eight-times conversion figure with some care, because AI-referred shoppers are, almost by definition, high-intent by the time they arrive. A customer who asks an assistant to find a specific product and is handed a buy button is not the same as a cold social impression. That caveat does not make the number wrong, but it does mean the lift is partly a selection effect. The honest read is that agentic traffic is small today and converts well, and the open question is what happens to conversion once agents become the default rather than the curious edge case.
Early Customers as Proof Points
Salesforce is leaning on named customers to make the abstraction concrete. Shirley Gao, chief digital officer at PacSun, tied the OpenAI integration to reaching Gen Z and Gen Alpha shoppers where they already spend time, turning conversations into engagement rather than forcing a jump to a separate app. For a retailer whose core customer treats a chat window as a natural surface, meeting them inside ChatGPT is less a novelty than table stakes. Luke Barber, head of ecommerce technology at Iceland Foods, pointed to Storefront Next as the enabler for what he called pioneering work on performance and AI development.
These are early and carefully chosen references, and neither executive is publishing an audited return on investment. But the choice of a US teen-apparel brand and a UK grocer is telling. It signals that Salesforce wants to show the agents working across very different baskets, price points and buying rhythms. Groceries are an especially useful test, because they involve large, repeat, low-margin baskets where a conversational assistant has to be fast and accurate to add value rather than annoyance. If agentic commerce holds up in the weekly food shop, it holds up almost anywhere.
Our Take for the Enterprise Buyer
For CTOs and CIOs, the practical significance is that the build-versus-buy calculation just shifted. A native catalog sync to ChatGPT and Google, plus a storefront that provisions in under an hour, undercuts the case for standing up bespoke connectors and agent frameworks in-house. The risk is lock-in: routing your discovery and checkout through one vendor's agents and one vendor's catalog sync makes that vendor very hard to leave. Retailers should press Salesforce on data portability and on how the agents behave when an assistant, not a human, is the primary customer.
The larger point is that the storefront is being redefined. For two decades the web store was a destination you sent traffic to. Agentic commerce inverts that, distributing the store into whatever assistant the shopper already trusts, and leaving the retailer to supply an accurate, transactable catalog on the back end. Salesforce is betting that most retailers would rather rent that plumbing than build it, and that the ones who wait past this holiday season will spend 2027 catching up. On current evidence, that is not a reckless bet.
