On June 24, 2026, Salesforce announced general availability of Agentforce Commerce. Three agents — Shopper Agent, Buyer Agent, and Merchant Agent — are now activatable by Salesforce Commerce customers, subject to regional and agreement terms.
General availability in Salesforce's announcement means the feature is available to enable. It does not mean the agents have been tested against your data.
What Salesforce Released
Salesforce described each agent as connected to catalog, inventory, and order data from the start. That architecture makes data readiness part of the activation decision, not a follow-up task.
Shopper Agent operates on a merchant storefront. It can use inventory and order context during customer conversations. Stale inventory or outdated fulfillment data could produce a commitment the warehouse cannot keep.
Buyer Agent supports B2B procurement through channels including WhatsApp and SMS. The announcement does not detail what account-specific data the agent draws on in those conversations. If contract pricing lives only in your ERP and is not synchronized to Salesforce, the agent may surface incorrect pricing before the order is placed.
Merchant Agent gives operations teams natural-language tools for catalog organization, boost and bury rules, and sort order changes. If those changes execute directly against your live catalog without a review step, a natural-language instruction could propagate an existing catalog error at scale.
Salesforce noted that Agentforce Commerce uses the same underlying commerce data and business logic across customer and operational workflows. A data problem could surface across all three agents instead of remaining in a single channel.
Where the Risk Concentrates
If your storefront uses staging environments, preview modes, or manual approvals before changes reach customers, note that the Salesforce announcement does not describe an equivalent review step between a natural-language request and its execution. Salesforce described the agents as connected to catalog, inventory, and order data; the announcement does not mention a staging layer.
The gap is widest when data is split across systems. If your inventory is mastered in a WMS syncing every few minutes, that latency flows directly into agent responses. If contract pricing lives in an ERP with no scheduled sync to Salesforce, the Buyer Agent may surface incorrect figures. If some fulfillment channels do not write order history back to Salesforce, that creates gaps in the data the agents can access.
Each gap is a potential agent error in front of a customer.
What to Audit Before You Enable Any Agent
Catalog completeness. What percentage of active SKUs have complete attributes and current pricing in Salesforce? Incomplete records may produce incomplete agent responses.
Inventory feed latency. What is the real sync frequency between your WMS or ERP and Salesforce? Identify which SKUs or locations are excluded from the feed.
Contract pricing location. Does account-specific pricing live in Salesforce, or only in your ERP? Verify how the Buyer Agent resolves a contract price and what happens on a failed lookup before enabling it for B2B conversations.
Order data coverage. Do all active fulfillment channels write orders back to Salesforce? Channels that do not reconcile create gaps in data the agents can access.
Merchant Agent governance. Who owns boost and bury rules and sort configurations? Apply the same change-control standards to natural-language catalog changes as to any other production configuration.
What to Watch
Salesforce cautioned that availability can vary by region and customer agreement. Confirm your agreement covers the agents you intend to enable before scheduling a rollout.
Buyer Agent conversations over WhatsApp and SMS open a customer path your team may not yet monitor. If you plan to enable those channels, consider how you will handle pricing access controls, order logging, and escalation. The announcement does not detail what additional integration work those channels may require on your side.