Salesforce rolls out Agentforce Commerce, positioning AI agents to carry deals from chat to checkout

Salesforce is betting that the next phase of e-commerce will be agent-driven, not merely search-driven. In a major product update called Agentforce Commerce, the company announced that three AI agents are now generally available and that they are designed to do more than talk to customers. Salesforce’s core claim is that these agents are connected to a business’s commerce data and operations so they can check inventory, apply contracts and pricing, and carry transactions through order management and service.

At the center of the rollout are three roles Salesforce says businesses can deploy as part of their own agent workforce. The Shopper Agent is aimed at B2C shopping journeys, carrying conversations from discovery through checkout and into service. The Buyer Agent is built for B2B procurement in channels buyers already use, including WhatsApp and SMS. The Merchant Agent is designed for internal commerce operations, letting merchandisers and operators organize catalogs, sort products, and respond to trends using plain language instead of menus.

Salesforce also said the agents are moving into the AI discovery layer where shoppers are increasingly starting. It described native integrations that place Shopper Agent into existing commerce channels tied to AI assistants. The company said Shopper Agent is generally available now with integration into OpenAI’s ChatGPT, Google Search including AI Mode, and that the Gemini app integration is expected to arrive this summer.

Salesforce anchored the push with data it attributed to its 2025 holiday shopping analysis. It said AI influenced 20% of global online sales, worth $262 billion, and that peak-season performance showed that retailers running their own shopper agents grew sales 59% faster than retailers that were not. Salesforce also cited that AI-referred traffic converts at eight times the rate of social, pointing to stronger downstream outcomes from AI-assisted discovery.

Beyond conversational interfaces, the company emphasized that action is the differentiator. It contrasted generative AI that “holds a conversation” with agentic AI that, in Salesforce’s framing, can take operational steps such as checking inventory and confirming shipping cutoffs before closing a sale. The company’s argument is that these abilities depend on connectivity to the underlying commerce system, including inventory visibility and order management, rather than functioning as a standalone chatbot bolted onto a website.

To support that connectivity, Salesforce said it has integrated Cimulate, which it described as context-aware, AI-native search. The resulting capability, labeled Agentic Commerce Search, is positioned as a discovery engine meant to surface relevant products based on intent rather than keywords. Salesforce said early customers are reporting double-digit improvements in both conversion and add-to-cart rates, and it added that the search can run anywhere a storefront lives, including non-Salesforce sites.

The release also expands how commerce can be linked across owned and external properties, and it is explicit about the merchant-of-record model. Salesforce said merchants can connect their catalog directly to ChatGPT through an OpenAI integration, synced from Business Manager without extra software or third-party tools. For broader discovery, Salesforce said an integration with Google Search including AI Mode and the Gemini app is available this summer. The company said orders land on the same platform that runs service, loyalty, and marketing, rather than being placed into a separate administration environment.

In B2B, Salesforce described the Buyer Agent as working without portals or logins, using SMS and WhatsApp workflows that translate a buyer’s message into a procurement process. It said a buyer can request quantities and see product images and exact SKUs, along with current contract pricing, then complete an order, all within the messaging channel. Salesforce also highlighted features it called Multi-Cart and Round-Trip Quoting for handling multiple purchases in flight, plus Mobile Publisher for shipping branded iOS and Android buyer apps.

Salesforce further described architectural changes it claims make this agentic approach workable across different storefront fronts. It said B2B commerce can run headless as part of a broader “Headless 360” vision, using either a templated front end on Lightning Web Runtime and Experience Builder or custom surfaces such as React storefronts, partner systems, and AI agents. It said the back end, including catalog, pricing, accounts, and Buyer Agent, stays consistent, allowing businesses to build different “surfaces” while keeping the operational foundation unified.

Salesforce did not provide pricing details or disclose which customer contracts will be required for every component of the rollout. It also did not specify rollout coverage by region beyond noting that availability may vary by customer agreement. What to watch next is whether early adopters validate Salesforce’s claims about conversion and operational lift, and whether the generative-to-agent shift reduces friction in fulfillment and service, areas where Salesforce says the connected system matters most. In the nearer term, the company said it will run an Agentic Commerce Summer Tour in 25 cities to demonstrate a live purchase flow inside ChatGPT and a B2B ordering experience over WhatsApp, which should offer the clearest preview of how the agents behave in practice.

Salesforce is positioning Agentforce Commerce as a platform bet rather than a set of point features, and it is also leaning on its partner ecosystem to drive deployment speed and reach. It said Storefront Next can stand up a production-ready storefront in under 30 minutes and that it ships in every B2C Commerce SKU. For development, Salesforce said it is introducing a Commerce Apps Framework that lets developers build apps discoverable and installable from Business Manager, alongside a list of launch partners it named. The company’s central message is that businesses need AI agents connected to inventory, order management, and customer data across B2C, B2B, point of sale, and service, if they expect AI to do the shopping rather than merely assist it.