# Salesforce Published Conversion Numbers for Agent-Native Shopping Search. They Cover One Thing. > Source: > Published: 2026-10-30 14:30:31+00:00 Salesforce has published merchant-side conversion numbers for [agentic commerce](https://forkast.news/glossary/agentic-commerce/) product discovery, marking a shift from theoretical agentic potential to reported performance metrics. According to the [Winter ’27 Release](https://www.salesforce.com/news/stories/winter-2027-product-release-announcement/), the company’s Agentic Commerce Search capability is now delivering a 13% conversion lift and a 17% increase in add-to-cart rates. These figures, framed by the vendor as “early results,” put a number on what agent-native discovery does to the funnel. ## The Mechanic of Discovery The underlying machinery driving these results relies on a dedicated small language model trained on each merchant’s specific catalog and historical shopper behavior. By interpreting natural-language queries, the system surfaces products in real time, bypassing traditional keyword-based search filters. This approach is designed to move beyond static indexing, attempting to match the intent of a shopper’s query directly to the inventory data provided by the merchant. However, the efficacy of this model is tethered to the quality of the input. While the search capability is now standard for all Shopper Agent customers, the broader market readiness remains uneven. Our [October 24 readiness analysis](https://forkast.news/67-of-retailers-say-they-plan-to-open-catalogs-to-ai-agents-fewer-than-a-handful-have-actually-done-it/) put the share of merchants holding the structured product data agents need at 15%. The search model trains on exactly that data. Without it, the performance gains Salesforce reports are not universally replicable across the enterprise landscape. ## What the Numbers Leave Open While the 13% conversion lift provides a headline figure, the release material leaves significant questions unanswered. The vendor has not published the methodology, sample size, or baseline metrics used to calculate these results. Consequently, it is difficult to determine how much of this lift is attributable to the agentic search itself versus other variables in the shopper journey. Furthermore, the Winter ’27 material names no loss placement anywhere and says nothing about who bears loss when the search leads an agent wrong. That silence is the finding: our [map of five products called agentic checkout](https://forkast.news/five-products-called-agentic-checkout-five-different-loss-placements/) asked the question, and the release does not answer it. Where liability would actually land is an inference from that silence, and it is ours, not the vendor’s. The Visa and BILL pilot, mapped in the same piece, states on its own record that approval there exists before the agent appears. ## The Default Move and the Meter Salesforce’s decision to make Agentic Commerce Search standard for all Shopper Agent customers mirrors the platform-scale default move [Shopify made](https://forkast.news/shopify-opens-the-checkout-it-called-human-only/) with its Fall ’26 Edition. Both platforms folded agent-native commerce into the default configuration rather than leaving it an opt-in – the door widening pattern, at platform scale. Yet this arrives alongside a looming shift in cost. Agentic usage is not metered today, but Salesforce has confirmed it will charge for every call a registered agent makes to its platform, with a November target and the multiplier still to be announced. As our [coverage of the meter](https://forkast.news/the-meter-arrives-in-november-salesforce-will-charge-for-every-call-your-agents-make-and-the-price-is-tba/) noted, the newest row on that rate card reads “TBA.” Merchants are being handed performance data with an upside attached while the cost of the calls producing it remains a blank space. The chargeable surface in this stack is the authorization, not the transaction – and now the meter sits one layer below the discovery numbers. ## What Moves the Read The gap between the promise of agentic efficiency and the reality of merchant readiness is where this number has to land. Salesforce is betting that standardizing the search will pull the necessary data structuring across its ecosystem. Until the methodology behind the lift is published and the liability question stops being unanswered, these figures are better read as a signal of direction than a guarantee of outcome. For enterprise merchants, the question is no longer whether agents will search the catalog. It is what the discovery costs when the multiplier lands, and who absorbs the miss when the search leads an agent wrong.