Discovery is migrating to AI chats faster than most commerce organizations can respond — and the data infrastructure underneath is already strained.
The opening move of the shopping journey is shifting away from the properties brands control. According to the fourth edition of Salesforce’s State of Commerce report, released July 28, use of agentic search as the first step in a purchase journey grew 200% year over year — a signal that consumers are increasingly beginning with a question posed to an AI assistant rather than a query typed into a search engine or a visit to a retailer’s homepage.
The finding sits alongside a broader set of data points that describe a widening gap between how quickly consumers have adopted AI-mediated discovery and how quickly commerce organizations have been able to restructure around it.
Referral traffic from AI chats is growing at multiples of overall traffic
The behavioral data is the more consequential part of the report. Drawing on platform activity from more than 1.5 billion global shoppers across 37 countries between Q1 2024 and Q1 2026, Salesforce found that traffic referred from AI chats grew between 150% and 428% year over year in every quarter measured. Overall site traffic, by comparison, grew in the single to low double digits.
Consumer survey responses point the same direction. Between August 2025 and May 2026, the rate at which shoppers reported discovering products through brand-owned properties declined 7%, and discovery through traditional search declined 15%. Over the same period, the rate of consumers turning to newer channels — AI assistants, AI features embedded in social platforms, and delivery apps — rose 38%.
Commerce leaders appear to have internalized the trajectory: 86% agree that large language models will be essential to product discovery within the next year. The most common responses so far are content-layer adjustments — improving product content quality, optimizing copy for conversational queries, submitting structured data feeds to AI search platforms, and rewriting product descriptions in natural language so they surface when a shopper asks an assistant for, say, a waterproof hiking boot under a specific price.
Caila Schwartz, head of agentic commerce shopper insights at Salesforce, framed the stakes in terms of visibility: brands absent from that first AI conversation are, in her words, <cite index=”1-1″>”invisible before the journey even begins.”</cite> She noted that by the time a shopper arrives on a brand’s site, an opinion has often already been formed elsewhere.
Worth noting for planning purposes: the shift is concentrated in discovery. Fully autonomous purchasing — an agent completing a transaction end to end — remains early by Salesforce’s own characterization.
Adoption trails behavior, but adopters are past the pilot stage
Just 28% of commerce organizations currently use agentic AI. Another 44% say they plan to adopt within the next six months, and 71% of respondents agree that scaling across channels is not viable without AI.
Among organizations already deployed, experimentation has largely given way to expansion. Only 4% describe themselves as primarily piloting use cases, while the largest single group — 35% — say their priority is scaling AI across functions and teams, spanning service, IT, and merchandising. Reported improvements are led by customer satisfaction, followed by personalization, development velocity, operational efficiency, and employee productivity.
That momentum comes with a measurement caveat that enterprise leaders should register: only 32% of organizations report having fully defined AI success metrics and KPIs. Organizations are, in effect, reporting wins before establishing a shared definition of what a win is — which complicates decisions about where to reinvest.
The infrastructure underneath is a constraint
The report’s less flattering findings concern the systems AI is being layered onto. Seventy-eight percent of commerce leaders say their vendor count has grown over the past two years. Only 27% say their customer data is fully unified across sales, service, marketing, and commerce.
Among organizations operating with fragmented customer data views, 40% acknowledge slow or ineffective responses to customer issues, 39% struggle to measure the impact of commerce investments, and 39% cite the cost of maintaining disconnected systems.
Omnichannel execution shows similar strain. Only 2% of multichannel organizations report no significant failure points. The most frequently cited breakdowns are inconsistent pricing and promotions (41%) and inventory that is not synchronized in real time (40%).
Pressure is compounding on both sides of the equation: 86% of commerce leaders say AI is raising customer expectations, while 61% say meeting those expectations is harder than ever. Implementing or expanding AI ranks as both the top priority and the top anticipated challenge for the year ahead.
The store is not a separate channel
The physical/digital distinction continues to erode. Physical retail remains the top holiday shopping destination for 77% of consumers, but the in-store journey is now largely digital: 79% of shoppers use their phones while in the aisle, and 12% ask an AI assistant for purchasing advice while standing in the store. Eighty-eight percent of B2C respondents agree that customers expect the same personalization in store as they receive online.
What it means for enterprise marketing and commerce leaders
Three implications stand out for teams setting 2027 roadmaps.
Discovery optimization now has a second surface. Traditional SEO investment does not automatically translate into visibility within LLM-generated recommendations. Product data quality, feed syndication to AI platforms, and natural-language product copy are becoming distinct workstreams rather than extensions of existing search programs.
Measurement discipline is the near-term gap. With fewer than a third of organizations reporting defined AI KPIs, the risk is not underinvestment but unattributable investment. Establishing baselines — for AI-referred traffic, conversion from AI-sourced sessions, and agent-assisted service outcomes — is a low-cost step available now.
Data unification is the gating factor, not the AI itself. Salesforce’s own framing is that organizations pulling ahead are using AI adoption as the occasion to unify data and define success criteria. Among organizations that have moved toward unification, the most commonly reported benefits are improved alignment across sales, marketing, and commerce; better AI and automation outcomes; and improved retention and loyalty.
