What the new data actually says
CI&T published its Retail Tech Report: Agentic Commerce Edition on July 22, and the headline numbers are hard to wave away. The consultancy found that 74 percent of surveyed US consumers have already used an AI agent while shopping, and 90 percent have either used one or say they are open to it. Among the people who have not tried agents yet, 63 percent report willingness to start. We read these figures as a threshold moment. Agentic shopping has moved out of the early adopter fringe and into ordinary consumer behavior, which changes the planning assumptions every retail technology leader carried into 2026.
The behavioral detail matters more than the adoption count. CI&T found that 78 percent of respondents believe agents make them smarter shoppers, and 55 percent say agents lead them to shop more often. Consumers reach for agents to compare brands, hunt for the lowest price, and confirm where a specific item can be bought. Those are exactly the high intent moments that decide conversion and margin. When a machine intermediary sits in that gap, the retailer no longer owns the framing of the choice. We think that shift, more than any single feature, is what should focus executive attention this quarter.
The purchase sequence is reordering
Melissa Minkow, Global Director of Retail Strategy and Insights at CI&T, put the structural change plainly. "Agentic commerce is now a reality in modern shopping," she said, describing how the traditional path to purchase has evolved from discover, research, buy toward research, discovery, buy. That inversion sounds academic until you map it onto a funnel. Discovery used to happen on a retailer surface, a store aisle, a homepage, a category page. Now research often happens inside an agent before the shopper ever lands anywhere a brand controls. The retailer gets pulled into consideration only after the agent has already narrowed the field.
We see two consequences that CTOs should plan around. First, the top of funnel is migrating to surfaces the retailer does not own, which means visibility now depends on how well product data feeds those agents. Second, the window to influence a decision compresses, because the agent arrives with a shortlist rather than an open mind. Retailers that still optimize primarily for their own site traffic are measuring a stage of the journey that is shrinking. The reordering rewards teams that instrument the agent layer and understand how their catalog surfaces inside it.
Agents as a shared mall
CI&T uses a useful image. Agents have become a kind of mall, pulling every brand and product category under one roof where a shopper can optimize decisions in real time. The comparison is instructive because a mall flattens differentiation. Inside an agent conversation, a premium brand and a private label sit side by side as structured attributes, price points, and availability signals. The rich context a retailer builds on its own site, the photography, the editorial, the merchandising logic, gets stripped down to whatever the agent can parse and rank. That is a demanding environment for any brand that competes on experience rather than raw specification.
This is why product data quality moves from a back office concern to a board level one. If a catalog is thin, inconsistent, or poorly attributed, the agent will underrepresent it or rank it below better structured competitors. We have watched several retailers this year invest in attribute enrichment precisely because agents reward machine legibility. The uncomfortable truth is that the shopper never sees the gap. They see an agent that simply did not surface a product, and the retailer absorbs a silent loss it cannot easily measure. Winning the shared mall starts with feeding it clean, complete, current data.
Treating agents as a real channel
The report argues that retail and CPG organizations must fold agents into their omnichannel strategy because these surfaces are becoming powerful revenue drivers. We agree, and we would push the point harder. Most enterprises still treat agentic commerce as an experiment owned by an innovation team, funded on a pilot budget, and reviewed quarterly. That posture no longer matches the adoption data. A channel that 74 percent of consumers already touch deserves the same operational rigor as marketplace, mobile app, and store. It needs owners, service levels, measurement, and a roadmap rather than a proof of concept and a slide.
Practically, that means naming the team accountable for agent readiness and giving it authority over the data pipelines that feed external agents. It means deciding which protocols and platforms to support and building the integration once so the catalog flows everywhere agents shop. It also means new metrics, because site conversion will not capture what happens inside a third party agent. Retailers that stand up this capability now will compound an advantage, because the data feedback loop improves ranking, and better ranking drives more agent traffic. The organizations still debating whether agents matter are ceding that loop to faster rivals.
The relationship risk hiding in the numbers
There is a strategic tension buried in the adoption story. As agents become the first touchpoint, the retailer risks losing the direct relationship that first party data and loyalty programs are built to protect. When an agent brokers the comparison and the checkout, the retailer may receive an order without the behavioral context around it, the browsing, the hesitation, the alternatives considered. That context is the raw material of personalization. A retailer that becomes a fulfillment endpoint for someone else's agent keeps the revenue and loses the intelligence, which is a poor long term trade for any business that competes on knowing its customer.
We think the defensible response is to make the retailer's own surfaces genuinely useful to agents while investing in reasons for shoppers to engage directly. That includes richer loyalty value, service quality that agents can cite, and proprietary data that improves how the brand is represented downstream. CI&T frames agents as extremely powerful revenue drivers, and that is true, yet revenue without relationship is fragile. The retailers that thrive will supply agents generously and still give customers concrete reasons to come back through a door the brand controls. Balancing those two demands is the harder management problem the data exposes.
What technology leaders should do this quarter
The near term work is unglamorous and high leverage. Audit product data for completeness and attribute quality, because that determines how a catalog surfaces inside agents. Map which agent platforms your customers actually use and confirm your assortment appears accurately across them. Instrument what you can measure about agent driven traffic and orders, even if the picture is partial, so the channel stops being invisible in the numbers. None of this requires a moonshot. It requires treating a live consumer behavior with the seriousness the adoption data now warrants, and funding it accordingly rather than leaving it to a pilot team.
The longer arc is about positioning. Agentic commerce is compressing the distance between a shopper's intent and a completed purchase, and the retailers that feed that machinery best will capture a disproportionate share of it. CI&T's data is a clear signal that the behavior is here, broad, and growing across electronics, personal care, and apparel. We would rather see enterprise leaders act on a 74 percent adoption figure now than wait for it to reach ninety and discover their catalog was illegible the whole time. The window to build the muscle is open, and it will not stay open indefinitely.
