Salesforce Just Went All-In On Agentic Commerce. Here Is How It Compares To Shopify's Stack.

Salesforce Agentforce Commerce versus Shopify agentic commerce stack

On July 6, Salesforce shipped what it calls its biggest Agentforce Commerce release ever. Shopper Agent, Buyer Agent, and Merchant Agent are now generally available. The product catalog connects natively into ChatGPT today, with Google Search, AI Mode, and the Gemini app arriving over the summer.

Read past the launch language and something bigger is visible: the two largest commerce platforms have now converged on the same architecture for the agentic era. Owned agents on your own properties. Syndication into external AI apps. A portable commerce layer underneath. And agent skills shipped for the coding assistants that developers actually use.

If you build on Shopify, this release is worth studying closely. Not because you should switch. Because it tells you which parts of the agentic commerce playbook are now table stakes, and where the real differences between the platforms actually live.

What Salesforce announced

The GA release covers three named agents plus a wave of supporting infrastructure.

Shopper Agent runs conversational shopping on the merchant's own storefront: real inventory checks, carrier cutoffs, store pickup fallbacks, and checkout in one conversation, with customer history carried across sessions.

Buyer Agent targets B2B procurement inside WhatsApp and SMS. A buyer texts "40 cases of the 16-oz fasteners, same as the March order" and gets SKU confirmation, contract pricing, and a completed order without a portal or login.

Merchant Agent is the back-office half: catalog organization, boost and bury rules, and sort tuning in natural language. Salesforce claims early customers cut time to task by 88 percent.

Around the agents, four platform moves matter most:

Agentic Commerce Search, built from the Cimulate acquisition, is an AI-native discovery engine that runs anywhere the storefront lives, including non-Salesforce sites. GA in July.
Deep external channel integration: catalog sync into ChatGPT from Business Manager, with Google Search, AI Mode, and Gemini following. The merchant stays merchant of record, and orders land in the same platform that runs service and loyalty.
Storefront Next, now GA and included in every B2C Commerce SKU, promises a production-ready storefront in under 30 minutes and ships an agentic developer toolkit with more than 40 prebuilt skills for AI coding tools.
Fully headless B2B: two paths, one back end. Use Salesforce's templated front end, or build your own head in React, WhatsApp, an AI agent, or a partner system, all running on the same catalog, pricing, and accounts.

Salesforce also cites its holiday data to justify the urgency: AI influenced 20 percent of global online sales during the 2025 holiday season, worth 262 billion dollars, retailers running their own shopper agents grew sales 59 percent faster, and AI-referred traffic converts at eight times the rate of social. Those are vendor numbers from Salesforce's own shopper panel, so treat the precision with care. The direction, though, matches what everyone else is seeing.

The convergence is the real story

Put the Salesforce release next to what Shopify has shipped in 2026 and the overlap is striking.

Shopify's agentic stack now includes Agentic Storefronts live in ChatGPT, the Universal Commerce Protocol co-developed with Google, Storefront MCP, Catalog API access for developers, llms.txt shipped by default on storefronts, Sidekick moving toward propose-then-approve workflows, and a rebuilt Hydrogen that is framework-agnostic, runtime-agnostic, and built for agents, with versioned agent skills copied into every project.

Salesforce's stack: owned Shopper, Buyer, and Merchant Agents, Agentic Commerce Search, native ChatGPT and Google integrations, Storefront Next with 40-plus agent skills, and headless B2B with a portable back end.

Line them up and both platforms are making the same five bets:

Owned agents beat rented ones. Both platforms want the merchant's agent, grounded in the merchant's data, running on the merchant's properties. Salesforce's EVP Nitin Mangtani says it directly: discovery will be influenced by AI platforms, but commerce itself happens predominantly on owned and operated properties.
External AI apps are channels, not threats. Both sync catalogs into ChatGPT and Google rather than fighting the traffic shift. Both keep the merchant as merchant of record.
The commerce layer must be portable. Shopify moved Hydrogen's commerce primitives out of React Router so any framework can use them. Salesforce made B2B fully headless with "two paths, one back end." Same abstraction, different vocabulary.
Agent skills are the new SDK. Hydrogen ships versioned skills into the project's agent folder. Storefront Next ships more than 40 prebuilt skills for AI coding tools. Both companies concluded that the way to make coding agents produce correct commerce code is to ship the instructions with the platform.
UCP is becoming shared ground. Salesforce explicitly names UCP as part of its long-term syndicated commerce plan. A protocol co-developed by Shopify and Google, now cited in a Salesforce launch, is on its way to becoming infrastructure rather than a competitive edge.

When the two biggest platforms in commerce independently converge on the same architecture, that architecture stops being a bet. It is the roadmap.

Where the platforms actually differ

Convergent architecture does not mean identical products. The differences are structural, and they map to who each platform serves.

Salesforce's edge is the unified enterprise back end. Its whole pitch is that the agent is only as good as what it is connected to: native order management, unified inventory across 7,000 locations, service history, loyalty, and B2B contract pricing on one data foundation. For a large retailer already running Salesforce for service and marketing, an agent that can promise a real delivery date and resolve a fulfillment issue is a genuinely hard thing to replicate with a bolt-on.

Shopify's edge is the developer surface and the merchant long tail. Shopify's agentic stack is protocol-first and developer-first: UCP and Catalog API are open for registration, Storefront MCP is documented, Hydrogen's primitives install into any JavaScript framework with one setup command, and the entire theme ecosystem gets standard storefront events and actions. Salesforce sells a platform to enterprises through account teams and a 25-city tour. Shopify ships primitives that any developer can adopt this afternoon.

Time to storefront is now a marketing metric on both sides. Storefront Next claims under 30 minutes to a production-ready storefront. Shopify and Vercel claim a deploy button that ships the Hydrogen Next.js starter in a few clicks. Both numbers are true in the demo sense and misleading in the operating sense, which brings us to the part both launches underplay.

The gap both platforms still have: who operates the storefront after the agents build it

Every claim in both stacks is about creation and transaction speed. Agents generate the storefront. Agents answer the shopper. Agents reorganize the catalog. Agents close the sale.

Almost nothing in either launch is about the day after: the marketing team that needs a campaign landing page by Friday, the merchandiser who wants to restructure a PDP without filing a ticket, the brand lead who needs to know which agent changed what and whether a human approved it.

This is the same conclusion we reached watching Shopify's Sidekick move toward propose-then-approve editing: the winning pattern is agents propose, humans approve, and the system keeps an audit trail. Salesforce's Merchant Agent gestures at this with natural-language merchandising, and its Visual Merchandising Workspace is a real nod to operator workflows. But an agent that rewrites your boost and bury rules from a sentence is exactly the kind of power that needs governance before it needs speed.

The faster agents make storefront creation, the more the bottleneck moves to safe, governed, merchant-operable iteration. That is true on Salesforce. It is true on Shopify. It is the reason Weaverse exists as the editing and governance layer above Hydrogen's commerce primitives, and it is the layer both of these launches assume someone else will provide.

What this means for merchants choosing a platform

If you are an enterprise already deep in Salesforce for service, marketing, and order management, Agentforce Commerce is a credible reason to consolidate rather than bolt agents onto a separate commerce stack. The unified data foundation argument is real at that scale.

If you are anywhere from SMB to upper mid-market, or you value developer velocity and an open protocol surface, Shopify's agentic stack remains the stronger position. UCP, Storefront MCP, Catalog API, and portable Hydrogen primitives are open and self-serve in a way Salesforce's account-team-led rollout is not.

And if you are choosing today, notice what neither platform will say out loud: the agentic layer is converging so fast that the durable differences are the back end you already run, the team you already have, and the operating model you can actually sustain. Choose for those. The agents will keep arriving on both platforms either way.

What this means for Hydrogen and headless teams

Three practical takeaways.

First, the "portable commerce layer plus agent skills" pattern is now industry-standard, not a Shopify experiment. Learning to work with Hydrogen's framework-agnostic primitives and shipped agent skills is a transferable skill, because the same pattern now exists on the other major platform.

Second, expect RFPs to change. Enterprise buyers evaluating replatforms will now ask both vendors the same agentic questions: owned agents, ChatGPT and Google syndication, agent-readable catalogs, natural-language merchandising. If you sell Hydrogen builds, be ready to answer the Agentforce comparison directly rather than dismissing it.

Third, governance is the differentiator nobody has fully shipped. Both platforms can generate and transact. Neither launch answers who approves agent changes to a live storefront, how merchants edit safely alongside agents, or how a brand keeps visual and content control when three different agents can touch the experience. Teams that solve that operating model, with clear boundaries between developer-owned architecture and merchant-owned iteration, will win the builds that matter.

If you are evaluating Hydrogen against Agentforce, or trying to figure out what an agent-ready storefront operating model looks like for your brand, the Weaverse team pressure-tests these decisions with production Hydrogen storefronts and the visual editing layer on top of them. Talk to us →

The Bottom Line

Salesforce's Agentforce Commerce GA is the strongest confirmation yet that the agentic commerce architecture has stabilized: owned agents on owned properties, syndication into ChatGPT and Google, a portable commerce layer, and agent skills shipped with the platform. Shopify got there first on the developer and protocol side. Salesforce is arriving with a deeper enterprise back end. The architecture war is effectively over because both sides built the same thing.

What is not over is the operating model war. Agents can now build storefronts and close sales on both platforms. Neither platform has fully answered who governs the storefront after the agents start changing it. That is the layer merchants will feel every day, and it is where the next real differentiation happens.