# Automated Commerce Moved Inside Albertsons' AI Assistant > Albertsons became the first retailer to sell sponsored product placements inside its AI shopping assistant, the same week Salesforce shipped brand-run shopper agents and new research showed a growing share of consumers, especially Gen Z, already shopping through AI. For consumer brands, the practical question is how to show up inside the assistant and keep control of spend, data, and the customer relationship as agents start to mediate the sale. Content type: article Source URL: https://www.agentpmt.com/articles/automated-commerce-moved-inside-albertsons-ai-assistant Markdown URL: https://www.agentpmt.com/articles/automated-commerce-moved-inside-albertsons-ai-assistant?format=agent-md Updated: 2026-06-28T16:56:04.349Z Author: Pancakes Tags: MCP, autonomous agents, AI Agents In Business, AI Agent Payment Systems, News, Agent Orchestration --- # Albertsons Put Ad Space Inside the AI Assistant, and the Shelf Moved Into the Conversation On June 22, Albertsons Media Collective and the ad-tech company Criteo began selling sponsored product placements inside Albertsons' AI-powered shopping assistant. Ask the assistant what to cook this week, and the conversational answer can now fold eligible sponsored products into its suggestions. Albertsons is the first retailer to put paid brand discovery directly inside an AI shopping assistant, and the move settles a question consumer brands have been circling for a year: when a customer talks to software instead of walking an aisle, where does a paid brand actually show up? The detail that makes this matter is hiding in how people use these assistants. Albertsons says more than 85% of the AI conversations begin with an open-ended question rather than a product name, something closer to "what should I make for dinner this week" than "pasta sauce, 24 ounces." That habit pulls brand discovery to the very start of the trip, before the shopper has named anything to buy. Old-school retail media let a brand bid on the search term after intent was already formed. Inside the assistant, the suggestion lands while the shopper is still describing the problem, and whatever the assistant proposes is what a busy customer is likely to drop in the basket. A conversational answer is also a tighter surface than a page of search results. It offers fewer slots and more surrounding context, and a suggestion that does not fit the request reads as a worse answer, not merely a worse ad. So the placement Albertsons is selling is not a banner or a ranked listing. It is a chance to be the product the assistant names in the moment it is being helpful, which is both more valuable and more fragile to win than a slot on a results page. Criteo framed the launch around control. "We are helping brands connect with shoppers in those moments of intent," said Sherry Smith, Criteo's president of retail media, "setting the foundation for how retailers will monetize AI-driven discovery while maintaining control of the shopper journey." Jill Pavlovich, who runs digital customer experience at Albertsons, described it as retail media that helps customers along the way and gives advertisers a path to engage closer to the moment of purchase. The phrase worth holding onto is "control of the shopper journey," because the same week made clear that control is about to get complicated. ## Bring your own agent Two days later, the supply side answered. On June 24, Salesforce moved three Agentforce Commerce agents into general availability. A Shopper Agent carries a consumer from browsing through checkout and post-sale support on a brand's own storefront, in the brand's voice. A Buyer Agent handles business purchasing over WhatsApp or text, so a wholesale customer can write "send 40 cases of the 16-ounce fasteners, same as the March order" and let the agent confirm the SKU, apply contract pricing, and place it. A Merchant Agent sits in the back office and lets a team organize catalogs and set merchandising rules in plain language instead of code. Salesforce also said brand catalogs will plug straight into ChatGPT, Google's AI Mode, and the Gemini app over the summer, with the retailer staying the merchant of record. That last term is doing quiet work. The merchant of record owns the transaction, the customer data, and the relationship, and brands fight to keep it because losing it means becoming a commodity supplier inside someone else's interface. Salesforce's pitch, set aside from its own growth figures, is blunt: a brand that does not run its own agent will be represented by one it does not control. Automated customer service, product recommendations, and checkout start flowing through an agent no matter what; the open choice is whether that agent speaks for your brand or for the platform that built it. Salesforce sells that capability as a packaged set of agents and customer service automation solutions tuned to its own commerce systems, which is a reasonable buy for companies already standardized on Salesforce. It is also a single-vendor answer to a question most brands will want to settle on their own terms, with their own models and their own systems, which is where the build-it-yourself path comes in. ## The shoppers already showed up None of this would matter if customers were not playing along. They are. Research from SAP Engagement Cloud's Emarsys unit, published the same June 24, found that 29% of consumers now use AI agents to help guide purchasing decisions, and 43% of Gen Z do. For a brand planning its next year, the distance between those two numbers is the tell: the behavior is not a novelty fading out, it is a habit working its way up through the customer base. A real slice of shoppers already let an agent shortlist for them, and the youngest cohort is the most comfortable handing over the choice. Brands are racing to meet that behavior with automation of their own. Also on June 24, the customer-engagement company MoEngage acquired Aampe, a startup that builds per-user "agentic decisioning," software that decides what to send each customer, when, on which channel, and how often. MoEngage plans to put that capability in front of more than 1,350 consumer brands. Consumer intelligence companies are now converging on the same idea from both directions: agents that help the shopper choose, and agents that help the brand decide how to reach that shopper. Demand and supply arrived in the same quarter. ## Showing up is a build-and-govern job For an operator, the week reduces to two concrete tasks, and neither is "have an AI strategy." The first is making your products callable by agents at all. Your catalog, inventory, pricing, and order systems have to be reachable as tools an agent can actually use, whether that agent is yours or a partner platform's. The second is deciding what those agents are allowed to do once they can transact on your behalf: how much they can spend, which tools and vendors they may touch, and which actions need a human to sign off. This is the part of agentic commerce that AgentPMT works on, and where "control of the shopper journey" stops being a slogan and turns into a set of switches. AgentPMT is a no-code platform for building and running AI agents, and its [API-to-MCP feature](https://www.agentpmt.com/docs/agent-connections/api-to-mcp-tool-conversion) turns an existing commerce or catalog API into an agent-callable tool in a few minutes, without an engineering sprint. That is the practical mechanism behind participating: a brand can stand up its own shopper or automated service agent on whatever model it already runs, rather than adopting one vendor's full system just to get in the door. The broader fight over who owns these shopping agents, and the money riding on it, is the subject of a separate piece on [the battle for AI shopping agents](https://www.agentpmt.com/articles/ai-agents-are-becoming-shoppers-the-15-trillion-battle-for-their-loyalty-just-started). Control is the other half, and it is more concrete than the word suggests. Once an agent can spend, a brand wants a [budget cap per agent](https://www.agentpmt.com/docs/core-concepts/setting-spending-caps), restrictions on which tools and vendors it can reach, an [encrypted vault](https://www.agentpmt.com/secure-ai-credential-management) so credentials are injected on the server and the agent never sees a key, and a running audit feed of every request and response. A human approval step can be built into the workflow for anything sensitive, a checkout above a threshold, a new vendor, a price override, so a person clears it before the agent acts. Worth being exact here: that checkpoint is something the brand designs into its own workflow, not a promise that any single tool keeps on its own. Handled as budgets, logs, and approvals rather than as a posture, "keeping control" becomes something a team can actually configure. We have written before about why [per-agent micro-budgets beat traditional purchase orders](https://www.agentpmt.com/articles/procurement-for-agents-micro-budgets-beat-purchase-orders) for exactly this kind of spending. This already runs in production, which is the useful part for anyone weighing whether to build. AgentPMT operates a live [Grocery Shopping agent for Kroger](https://www.agentpmt.com/marketplace/grocery-shopping-kroger) that turns a shopping list into a real grocery order, a working example of an agent taking an ordinary consumer-goods action from start to finish. Automated commerce of this kind is no longer a demo reel; it is something a brand can assemble from its own systems, on its own terms, and watch run. ## The trap on the way in There is an honest caution worth keeping in view. Forrester's analysts warned this week that most marketing organizations are aiming AI at cost-cutting, treating it as a way to do the same work cheaper rather than to build an experience a competitor cannot match. For consumer brands entering agentic commerce, that is the wrong instinct. The win inside an AI assistant has little to do with deflecting a support ticket or trimming a headcount line. It comes from being the brand whose agent answers a vague dinner question better than a rival's bare catalog can, because the experience was designed rather than budgeted down. Where e-commerce and AI meet, the brands that treat the assistant as a capability to build will out-show the ones that treat it as an expense to cut. The assistant is becoming a place brands compete to be present, useful, and buyable, and the three moves of this single week, paid placement, brand-run agents, and the integration work that makes products callable, are pieces of the same shift. The takeaway is small and immediate: decide how your brand shows up before an agent decides for you. The tools to do it, to make your products agent-callable and to set the guardrails around what those agents can spend and touch, already exist. Brands that wire them up now get to shape how they appear in the conversation. The ones that wait will be described by an assistant they were never part of. * * * ## Sources - Albertsons integrates branded product placement into AI-search tool, Grocery Dive - Albertsons Media Collective Brings Sponsored Products to AI-Powered Search, The Shelby Report - Albertsons integrates branded product placement into AI-search tool, Marketing Dive - Salesforce releases AI agents among B2B ecommerce updates, Digital Commerce 360 - Salesforce adds AI agents to Agentforce Commerce with direct catalog integration, Shopifreaks - Real-time engagement will define how brands win in 2026, SAP Engagement Cloud / Emarsys - MoEngage acquires Aampe to bring 1:1 agentic decisioning to B2C marketing teams, PR Newswire - Nine in 10 US marketing agencies use AI to cut costs at the expense of creativity, Forrester - Yesterday's Marketing Technology and AI News, June 25, 2026, The Agile Brand Guide