Salesforce Reports Rapid Growth in AI-Led Shopping Discovery

Salesforce said on 30 September 2026 that the use of agentic search as the first step in a shopping journey had grown 200% year on year in its research. The announcement accompanied the fourth edition of its State of Commerce report.

The Singapore release describes surveys of commerce professionals and consumers, supplemented by behavioural data from global shoppers. It reports changing discovery patterns alongside organisational challenges involving fragmented customer information and disconnected channels.

The findings are Salesforce's research, with the populations and definitions set out in its report. The growth figure should not be interpreted as the proportion of all shopping that now begins in AI, or as a sales-growth forecast for an individual retailer.

Discovery can change before checkout does

Frontier's interpretation is that brands need to investigate the earlier shortlist decision as well as their measured website visits. A shopper may ask an assistant to compare suitable products before choosing a retailer or using a familiar checkout.

That creates an information task. Can the brand explain the product, its limitations and local availability in a form that supports an accurate recommendation? Is third-party information consistent with the current offer?

It also creates a measurement task. Answer exposure, identifiable referrals and orders are different observations. The report provides a reason to examine the relationship, not a basis for assigning all unexplained demand to AI.

The operational response is more specific than “adopt AI”

A brand can begin with a small set of commercially important questions and inspect which products and sources appear. Material errors can then be traced to product records, retailer listings or missing explanations.

The recommendation-quality audit distinguishes being named from being suitably recommended. That distinction helps make a discovery programme useful when broad adoption numbers attract attention.

For Australian, New Zealand and China-market teams, platform and market conditions still need separate examination. Evidence from a global report cannot substitute for checking the relevant local experience.

The news strengthens the case for treating AI discovery as a researchable customer touchpoint. The next step is a declared baseline and an accountable improvement, rather than an assumed share of future revenue.