Agentforce Commerce is Salesforce’s biggest commerce release in years: Shopper, Buyer, and Merchant Agents are generally available, the ChatGPT catalog integration reaches GA in July 2026, and Google Search and Gemini follow this summer. The winning move is to run both plays at once: syndicate your catalog to AI channels for discovery, and put your own Shopper Agent on your storefront for conversion. The brands that prepare their product data and unified profiles now will own the 2026 shopping season.
In This Article
A shopper opens ChatGPT and types: “I need waterproof trail running shoes under $120 that ship by Friday.” She gets three options, picks one, and buys it. She never opens a browser tab or sees a search results page. Your storefront never gets the visit. And yet, if your catalog is connected through Agentforce Commerce, that order lands in your Salesforce org with your brand as the merchant of record.
That transaction is now real. Salesforce announced its biggest Agentforce Commerce release yet in late June 2026, and the headline feature, native ChatGPT commerce, goes generally available this month.
The timing is deliberate. Traffic from AI assistants grew 119% year over year in the first half of 2025, and it kept climbing. Over the last holiday season, AI influenced 20% of global online sales, worth $262 billion. Salesforce projected that agents would drive 22% of Cyber Week orders, and the season went on to break records, with Agentforce Marketing and Commerce delivering 68 billion AI-powered product recommendations.
Year-over-year growth in traffic from AI assistants in the first half of 2025
AI-powered product recommendations delivered during the record Cyber Week
NPS lift Pandora reported after automating queries like “where is my order?”
Here is the part I find most interesting as an architect. The competitive question has quietly shifted away from “which LLM powers your agent?” toward “can your agent check live inventory, honor a contract price, and confirm a carrier cutoff?” Conversation is cheap. Transactions require plumbing. And plumbing is exactly where Salesforce has spent twenty-five years.
Agentforce Commerce ships a workforce of ready-to-deploy agents that your business owns and your data grounds. Each one is wired natively into your catalog, inventory, and order data from day one. That native wiring is the whole point. A bolt-on chatbot can describe a product. These agents can sell it.
Carries the customer from first question to purchase and on into service, in your brand voice, on your storefront. It checks live inventory, confirms carrier cutoffs, offers store pickup, and closes the sale inside a single conversation. Context carries into any later service interaction, so the customer never repeats themselves.
Meets B2B buyers inside WhatsApp and SMS, where reorders actually happen. Contract pricing, account context, and order history come along for the ride. Your distributor can restock from a text thread instead of logging into a portal.
Runs the back office in plain language. Your merchandising team can organize catalogs, sort products, and respond to demand trends conversationally, without waiting on a developer ticket for every change.
Guided Shopping, already generally available, is worth calling out separately. It turns the agent-to-shopper exchange into a rich experience with product page content, interactive image carousels, and one-click checkout. Pandora used this class of capability to automate simple order-status queries and improve storefront recommendations, and reported a 10% NPS lift. Summer updates extend the same continuity into post-purchase territory with guest order tracking and intelligent FAQs on product pages.
The design decision I respect most here: these agents act on the same catalog, order, and customer objects your teams already govern. There is no shadow commerce database to reconcile. When the Shopper Agent promises a delivery date, it promised it from your order management data, which means your existing data governance investment directly improves agent accuracy.
The mechanics are simpler than you might expect. Through the OpenAI integration, your product catalog connects directly to ChatGPT, synced straight from Business Manager. No extra software. No third-party middleware. A shopper discovers your product inside a ChatGPT conversation, and the purchase completes with your brand as the merchant of record.
That last phrase deserves a second read. Merchant of record means the order, the customer relationship, and the data all land on your core Salesforce platform, available for service, loyalty, and marketing. The AI channel is the front door. The customer relationship stays in your house.
Shopper, Buyer, and Merchant Agents are generally available, alongside Guided Shopping and Storefront Next. Brands can deploy owned agents ahead of peak season.
Product catalogs sync from Business Manager into ChatGPT. Discovery and purchase happen inside the assistant, with the brand as merchant of record.
Syndication extends to Google Search, including AI Mode, and the Gemini app. One catalog feeds multiple AI discovery surfaces.
The Universal Commerce Protocol drives commerce directly into external channels over time, while ACP integrations with partners like Stripe and AP2 payment support handle the transaction layer.
Salesforce’s own framing of the channel strategy is candid: AI platforms and social apps will heavily influence referral traffic, but most commerce will still happen on surfaces the brand controls, from the storefront and mobile app to the messaging channels it already runs. So the architecture supports both plays at once. Syndicate for reach. Own the agent for conversion and loyalty. Nitin Mangtani, who leads Agentforce Commerce, put a deadline on it: winners will have a Shopper Agent live on their own properties before this shopping season.
Treat the AI channels as a demand-generation surface with a transactional tail, not as a replacement storefront. Your pricing rules, inventory truth, and fulfillment promises still originate from your Salesforce platform. That single source of commercial truth is what keeps a ChatGPT order and a storefront order from becoming two different customers with two different experiences.
Everything in sections two and three depends on infrastructure that shipped quietly alongside the agents. This is the part most coverage skips and the part architects should study first.
Enterprise storefront capability with out-of-the-box simplicity, plus the new Commerce Apps framework, a plug-and-play packaging model for extending the storefront. Iceland Foods’ ecommerce technology lead described it as exactly where they want to be. Implementations that once meant long timelines now start from a productized base.
Automatically transforms registered digital accounts into unified Person Accounts across the platform, and embeds the CRM Contact ID directly into the storefront authentication token. The Shopper Agent gains secure access to full customer context, which unlocks autonomous actions like processing returns or sharing real-time order data. This replaces the custom integration code many teams have maintained for years.
Built on technology from the Cimulate acquisition, this engine moves product discovery beyond keyword matching. It reads shopper context in real time across browsing and buying signals, so “something for a rainy trail race” resolves to actual products instead of a keyword miss.
Digital commerce, point of sale, and order management on a single platform, with connected data from Sales, Service, Marketing, and Revenue operations. Order Routing for Order Management and Data 360 Zero Copy Access for Commerce both reach GA in Winter 2026, extending the same foundation.
B2C Commerce development now works with AI coding agents like Agentforce Vibes, Claude Code, Cursor, and GitHub Copilot: spinning up sandboxes, deploying cartridges, and creating components through conversational prompts, with MCP integration for autonomous debugging and log monitoring.
Step back and the acquisition pattern comes into focus. Cimulate supplied the intent-aware search engine, Informatica the data quality muscle, Doti the enterprise search layer, and Spindle AI the reasoning depth. Each purchase strengthens a specific layer under the agents. Salesforce is building the commerce stack the agentic era requires, layer by deliberate layer, and this release is where those investments become product.
The 2026 shopping season is the first one where AI channels carry real order volume, and it arrives in months, not years. Here is how I would sequence the next 60 days for a team on Salesforce Commerce today.
A release this large rewards teams that plan deliberately. Four things belong on your radar.
- Shopper, Buyer, and Merchant Agents, Guided Shopping, and Storefront Next are GA today: build on these now
- ChatGPT lands in July, Google Search and Gemini this summer: plan channel launches accordingly
- Order Routing and Data 360 Zero Copy Access for Commerce arrive in Winter 2026: design for them, deploy when GA
- Phase your rollout to the availability calendar and every step lands on solid ground
- Agents recommend from structured attributes, so enrich size, material, use case, and compatibility fields
- Inventory accuracy becomes customer-facing the moment an agent promises availability
- Identity resolution quality decides whether the agent greets one customer or three fragments of her
- Every hour spent on catalog and profile quality pays out across every channel at once
Third: conversational commerce needs conversational skills. Your merchandisers will tune agent behavior the way they once tuned search synonyms, and your analysts will read conversation transcripts the way they once read funnel reports. Budget enablement time for both. The teams that treat this as a new craft, rather than a feature toggle, get better agents.
Fourth, and this one is genuinely exciting: the demographic wave is on your side. Younger shoppers already start product discovery in AI assistants. PacSun is connecting its catalog to ChatGPT precisely to meet Gen Z and Gen Alpha customers where their conversations already happen. Meeting the next generation of buyers in their native channel is the kind of opportunity that shows up once a decade.
My honest read after a week with this release: the checkout button moving into ChatGPT is the headline, but the durable advantage is the unified platform behind it. Discovery surfaces will keep multiplying. Gemini today, whatever comes next tomorrow. The brands that thrive will be the ones whose catalog, inventory, orders, and customer identity live on one governed foundation that can syndicate anywhere.
So here is the question worth asking your team this week: if a shopper found your product inside ChatGPT tonight, would the conversation end in your order management system? If the answer is yes, you are ahead of most of the market. If not, you now have the playbook.