The two words carrying the weight in the Cyber Week estimate are "influence" and "touching" [1]. Neither is the same as an agent completing a purchase against its own criteria, and Yu, who runs an agentic-AI vendor and so holds a position in this market, effectively concedes the point when he calls the volumes small against total retail sales [5][10]. An order an assistant helped a person find and an order a protocol placed against a machine-readable feed are different operational problems, and the argument he builds rests on the second [7].
The Shopify figure is the more useful one, mostly because it shows how low the base sat. Orders originating from AI-powered search arriving at nearly 13 times the year-earlier volume means the year-earlier number was under 8 percent of the current one, since 1 divided by 12.9 is 0.078 [2][13]. That is the profile of a channel young enough that nobody's product data was built for it, which is also the profile of a channel where one quarter of unglamorous cleanup can still buy a position.
The survey line gives a floor worth writing down. If a third of consumers expect at least 10 percent of their purchases to be AI-driven within a year, the implied floor for that group is roughly 3.3 percent of their spend, being 0.33 multiplied by 0.10 [4][14]. A few percent of spend is not a crisis on its own, though a gate with published criteria behaves differently from a soft preference: one you can lose without ever appearing in the comparison.
A skeptic reads the same column and says this is a vendor telling brand teams to buy infrastructure, and that an order "touched" by AI is a referral statistic in costume. The first half is fair, and worth stating plainly, because the numbers arrive from a seller of agentic services [10]. The answer to the second half is that the failure mode is asymmetric. Being outranked is recoverable inside a quarter with price or spend; being absent from the candidate set because a spec field sits empty, a SKU number doesn't match the listing, or the return policy hasn't been updated in months is a gap that campaign creative cannot close [8][9].
The tradeoff to name here is ownership rather than budget size. Feed accuracy and price consistency cost little next to a campaign, and review depth costs even less, but they are somebody's headcount and somebody's on-call, and Yu's argument moves them out of operations and into the brand team's remit [11]. That is the decision genuinely available this quarter: who is accountable when the feed goes stale, and whose review it fails. What belongs in the decade column is which rails prevail, and the record here does not say, since ACP and UCP are both described as standardized ways for agents to discover, evaluate and transact, with no adoption figures attached [6]. Nor does the source tell us how agents weight aggregated review quality against price, which is the parameter that determines whether brand equity re-enters the calculation at all or simply sits outside it [7][12].
