Brazil’s 84 percent no-AI figure — the one Linx published and everyone has since cited — is not wrong. It is narrow. It answers a different question than its readers think they have asked.

The survey drew on 328 Linx clients and found, among the 16 percent already using AI, that the leading applications were chatbot service, automated reports, and automated messaging. All three sit on the dashboard tier of a retailer’s stack. The warehouse, the planogram, and the replenishment cycle were not in the questionnaire. The figure measures whether a retailer has bolted a chat window to its checkout; it does not ask whether the checkout itself is routed by an agent.

The operational tier runs on different arithmetic. Magazine Luiza has wired its Lu assistant into WhatsApp and now converts those conversations at three times the rate of search inside its own app. That gap is large enough to reframe the product: the agent is not a chatbot; it is the closer.

Lower in the stack, SAP Brazil is shipping Retail Intelligence into its S/4HANA installed base. Rui Botelho, president of the regional unit, says AI is no longer optional — a position worth weighing from a vendor whose ERP infrastructure underpins operations the Linx questionnaire was never designed to reach. The 328 retailers in the Linx survey were Linx clients; the SAP installed base carries different names and runs different infrastructure.

Logistics Reply, meanwhile, is selling its GaliLEA agent builder into warehouse management systems worldwide. GaliLEA handles dynamic task allocation and agent coordination inside the warehouse — the kind of operational layer that no retailer would describe on a survey as “using AI for chatbots.” None of these deployments ask the retailer to tick a box on a Linx form.

The 84 percent measures admission, not adoption.

The cleanest evidence is Linx itself. The tiinside report notes the company began testing its AI agent in 2023, the same year its survey established the 84 percent baseline. Linx shipped the agent for general deployment in 2025. By the logic of its own methodology, it has since moved retailers from “does not use AI” to something else without those retailers lifting a finger.

A competing survey drawn from across the industry, rather than from a single vendor’s client list, already puts adoption at 47 percent. The divergence is methodological, not cultural.

Foreign capital reads the 84 and decides Brazil is greenfield. The diagnosis shapes Gap’s leasing strategy, as we noted yesterday, as much as it flatters it. Brazilian retailers are not waiting. They are buying from SAP, from Reply, from the integrators that rewrote their warehouse stacks last quarter. They are simply not answering a survey designed to catch them at it.

The question is not whether Brazilian retail uses AI. It is who has been asked, and by whom.