WooCommerce Product Vendors links each product custom post to a vendor through the wcpv_product_vendors taxonomy, and stores commission rows in wp_wcpv_commissions (commission id, vendor id, product id, order id, total commission amount, status, paid status). Vendor managers live as users with roles like wc_product_vendors_manager_vendor. SleekAI exposes the taxonomy term, the related commission rows, and the per-order vendor breakdown as named variables on the bot's system message.

That gives shoppers and vendors completely different but accurate answers from the same bot. A shopper asks who makes a specific candle, the bot reads the assigned wcpv_product_vendors term and returns the vendor name, location, and contact policy. A vendor manager logged into the storefront asks about their last 30 days of payouts, and the bot queries wp_wcpv_commissions for rows where vendor_id matches, returning total earned, total paid, and what is still pending.

Display conditions split the experience cleanly: a public bot scoped to product pages handles shopper questions, while a vendor-only bot scoped to the Vendor Dashboard URL and the wc_product_vendors_manager_vendor role handles commission questions. Conversation logs capture each ask with model name and token usage, which is useful when marketplace operators want to see what each side is asking the bot.

Multi-vendor marketplaces have two completely different support populations and one inbox. Shoppers want to know who actually made the candle, where the vendor ships from, and whether the marketplace or the maker handles a return. Vendors want to know how much they earned, when they get paid, and whether a specific commission row has cleared.

The operator in the middle gets both streams and has to context-switch on every reply. Product Vendors already structures the data. The taxonomy ties each product to a vendor, the commissions table tracks earnings per order, and the user roles separate shoppers from vendor managers.

A chatbot that reads the same fields lets each side ask its own questions and get the right answer without bothering the operator. The shopper bot stays scoped to public pages and only exposes vendor-name data. The vendor bot is locked behind the dashboard URL and the manager role, so commission numbers never leak across vendors.

Generic chatbots cannot do this. They have no taxonomy lookup, no commission table to query, and no role-aware display conditions. They will either tell shoppers the platform makes everything (which the curated marketplaces precisely want to avoid) or refuse to engage with vendor payout questions because they have no source of truth.

The variable mapping between the bot's system message and the actual Product Vendors data is what makes both flows work. It is also what makes multibot useful: operators can ship a friendly public bot and a precise vendor bot from the same install, with two separate instructions and two separate logs, instead of stitching together two unrelated services.