Loading
Prepare for Email to Become the Default Login ExperienceRead More
Ongoing maintenance for Salesforce HelpRead More
Salesforce Enforces New Security Requirements in Summer 2026Read More
Get Started with B2C Commerce
B2C Shopper Agent | Product Expert

You are here:

  1. Salesforce Help
  2. Docs
  3. B2C Commerce for Merchandisers and Administrators

B2C Shopper Agent | Product Expert

Handles all product-related requests, such as browsing, searching, recommendations, comparisons, and product questions. It searches the catalog and looks up detailed product information as needed.

Subagent Name Subagent Handles Actions
product_expert
  • Guided discovery—helps shoppers find the right product through natural conversation. It can narrow results based on use case, budget, or preference. Example, "Show me waterproof hiking boots under $150."
  • Product comparisons—compares two or more products and highlights key differences. It compares specifications, features, price, ratings, and trade-offs so shoppers make decisions with confidence. Example, "Compare these two running shoes on cushioning, weight, and price."
  • Multi-item kit builds—assemble complete sets, outfits, or routines by finding each component together. For example, a full poolside look or a starter kit.
  • Cross-selling opportunities—surfaces relevant complementary products alongside the main item to increase basket size and round out the purchase.
  • Answers to product questions—responds to specific feature questions, for example, "Which of these two laptops is better?" or "Do these shoes run true to size?"
  • GetB2CUserAccessToken—authenticates the shopper session (guest or logged-in) before API calls.
  • CommerceProductSearch—searches the catalog, and supports conversational context, facets, refinements, and custom properties.
  • GetB2CProductDetails—full details for a single product ID, including specifications, variants, pricing, inventory. The agent uses this information before rendering a product tile or comparison.

Response components: product_carousel, product_tile, product_comparison, suggestions, tools.

To customize this agent:

Product Expert drives the discovery experience so it exposes the widest set of tuning points. You can adjust this subagent without changing the underlying reasoning node:

  • Catalog context (retailer_description)—determines the search quality. Provide guidance on how to search the catalog. Include catalog structure, product and collection names, and category vocabulary. The more specific the information is, the more effective the search.
  • Brand voice (retailer_voice)—full tone specification: register, vocabulary, sentence length, what to say and an explicit never-say list, and how to shift tone by context. For example, a positive tone upon product discovery, or a neutral tone when the question is out of scope.
  • Tool-calling behavior (tool_calling_guidance)—rules for how and when to search. Run parallel searches for multi-item requests, expand weak keywords into synonym lists, and find natural cross-sell partners. It helps you recover from zero-result searches by refining the search. Each rule string can have a maximum of 255 characters.
  • Response formatting (response_guidance)—behavioral and formatting rules applied to every response. Specify how the response looks. You can instruct the subagent to include a carousel when mentioning products, bold product names, and make sure that the agent ends the conversation with a follow-up question. Optionally, specify whether to enforce language rules. Each rule string can have a maximum of 255 characters.
  • Search enrichment and source (customProperties)—show custom attributes like ratings or sale flags on result. Use of configTemplate, which identifies the search source. The configuration drives the results. Also, id_col, which holds the product ID.
  • Integration path—the Product Expert subagent communicates with your B2C Commerce instance to look up products and manage the cart. The integration path controls how it connects. Set the useSfraApi switch to true to route through SFRA controllers. Set it to false to call SCAPI directly. Select the option that suites your business logic. Specify the connections settings per session from the Messaging Session record. The settings include named credentials, apiVersion, siteId, orgId, locale, and currency.

These same variable-driven levers apply to the Cart and Order Confirmation subagents, with the scoped variants cart_tool_calling_guidance and order_tool_calling_guidance, per-subagent behavior.

 
Loading
Salesforce Help | Article