The uplift figure and the delegation figure both come from parties with something to sell. Envive is an agentic commerce company whose business is making brands findable by AI shoppers, and its co-founder Aniket Deosthali is the source of the claim that shoppers spend 30% to 40% more per order when shopping with AI [4]; the almost-70% delegation number comes from Kristina Coughlin of the creator marketing agency Trevant [3], nine points above the National Retail Federation's 61% of shoppers who said they used generative AI tools to find prices and ask for recommendations [1], a relative gap of about 15% [14]. The two are not measuring the same act, and that is the difficulty: using a chatbot to compare calculators and letting it pick the calculator are separated by a conversion step nobody in this material has counted.
Take the arithmetic at face value anyway. If the average AI-assisted order runs 35% larger while the household is also timing purchases to capture the 20% to 30% that Alexander Lee described [10], then 1.35 divided by 0.75 puts roughly 80% more units in the basket [15], which is a vendor's cross-sectional average multiplied by one Brooklyn family's timing claim, so treat it as arithmetic and not as proof. Even a third of it would change what the shelf is for.
What a delegated list costs a retailer is the visit, not the pens. Lee named Target, Home Depot and Costco as the trips his family no longer makes, and said the assistant even adjusted the list [9]; Lee's account leaves open which retailer ultimately filled the order [17]. That is where the upstream-capture reading thins out. The decision moved, but whether the money moved with it is not something this account shows.
The cash tell is the spend Deosthali reports, a big spike in companies expanding their use of AI so they become more discoverable to shoppers [5], from a founder who previously built Walmart's generative AI shopping experiences [6] and now sells the fix. That budget buys placement inside someone else's window, which makes it a lease rather than an owned channel, and the landlord sets the renewal.
My read, with the counter in the same breath: the comparison layer is migrating toward whoever holds the intent, which is why Alphabet's Search-and-Gemini read on intent and Amazon's catalog, Prime purchase history and Alexa for Shopping [12] are the only listed ways to own any of it while OpenAI and Anthropic remain private with IPOs possibly before year-end [11]. The counter is selection. If AI-assisted shoppers were already the heaviest spenders before they opened a chatbot, then 30% to 40% measures who adopted first rather than what adoption did, and the whole thesis is a cohort artefact. A second reading has the uplift real but temporary, competed away once merchandising teams optimise for the same prompts. A third has it durable, in which case the 61% is an advertising story before it is a retail one. Worth noting while weighing the source's conclusion about who wins: CNBC calls both Alphabet and Amazon Club names [13].
