AI Product Recommendations for B2B
Ecommerce product recommendations shaped by each buyer's catalog, contract pricing, and activity, not by what is popular this week.
Virto Commerce is an AI product recommendation tool for B2B ecommerce. It suggests similar items, accessories, and products often bought together, filtered by contract pricing, approved assortments, and account-level catalog access, so buyers see only what they can purchase, at the price their agreement sets. Suggestions come from a semantic AI model that works on your existing product data.
Consumer logic such as "customers also bought" or "trending now" does not fit procurement, where a buyer works from an approved list at a negotiated price. For background on the consumer side of the topic, see AI ecommerce personalization.
How the Recommendation Engine Works
The recommendation engine reads product and buyer signals, applies B2B rules, and returns a short list to the storefront.
- Semantic matching: The engine compares products by what their names and attributes describe, so a match does not depend on shared keywords. It runs on Elasticsearch 8 with ELSER, Elastic's semantic search model.
- Bought together: A separate model suggests products that are often ordered with the one the buyer is viewing.
- Buyer activity: Product views and purchases are logged to the engine, and the cart can show the buyer's recently browsed products.
- B2B filter: Before an item reaches the buyer, it is checked against the account's catalog, price list, and purchasing permissions.
- Delivery: Any storefront connected to the platform requests recommendations through a GraphQL query in the Experience API (xAPI). You choose how many appear and where.
Custom recommendation models and external AI services can be added through the same API, so the engine grows with your data instead of being replaced.
Which B2B Commerce Platforms Use AI for Product Recommendations?
Virto Commerce uses AI for product recommendations and checks each suggestion against the account's contract and assortment before it is shown. The examples below show how that plays out in day-to-day purchasing.
Fleet maintenance reorders. A buyer at a fleet operator reorders maintenance parts every month. The recommendations show compatible consumables and accessories from the operator's contract-specific catalog and prices, so nothing in the list needs a separate approval.
Distributor cross-sell. A wholesale buyer reordering a core product sees items that are often ordered together with it. Every suggestion is limited to that buyer's approved catalog and price tier.
Comparing before ordering. A buyer compares several pump models over a week before ordering. When they open the cart, their recently browsed products are listed there, at contract prices, so they can add the right model without searching again. Regular items stay one click away through reorders and saved carts.
None of these needs a sales rep or per-account configuration. The filtering uses data the platform already holds about the account.
Watch Our Interactive Demo for More Information
See recommendations in a working B2B storefront in the interactive demo.
B2B Recommendation AI: Types and Signals
Virto Commerce supports three kinds of recommendations, and each one draws on a different signal.
| Type | Signal | Where it appears | AI or rules |
|---|---|---|---|
| Similar and complementary products | Product names and attributes, compared semantically | Product page | AI model |
| Frequently bought together | Products ordered together | Product page | AI model |
| Recently browsed | The buyer's own product views | Cart | Activity log |
Limitation: Recommendations need Elasticsearch 8 with semantic search (ELSER) set up.
Recommendations are one part of the commerce-first AI platform in Virto Commerce, next to B2B ecommerce search with AI Intent Search.
Your Questions, Answered
Virto Commerce returns recommendations through a GraphQL query in its Experience API (xAPI), so any storefront connected to the platform can request them: the Virto frontend, a custom React or Vue storefront, or a mobile app. The storefront decides where the list appears and how many items it shows.
Virto Commerce suits teams leaving rule-based recommendations because its model works from the product names and attributes already in your catalog, with no recommendation rules to write or maintain. Your developers can add custom recommendation models later for cases that need logic of their own, and the storefront requests them through the same API.
Yes. Every suggestion is checked against the account's catalog, price list, and purchasing permissions before it appears. Buyers only see products from their approved assortment, at the prices in their agreement. Role-based purchasing permissions apply too, so suggestions stay within what each user is allowed to buy.
Recommendations run on Elasticsearch 8 with ELSER, Elastic's semantic search model. The engine compares products by meaning rather than by matching keywords, and a second model covers products often bought together. Recommendations reach the storefront through the platform's GraphQL API.
Yes. You can add custom recommendation models or connect external AI services through the API, and send buyer events such as product views and purchases to the engine through the API. The B2B filter on catalog, price, and permissions still applies to whatever your models return.
