Two competitors, one number
In the same week of Q2 2026 earnings calls, Walmart CEO John Furner and Amazon CEO Andy Jassy each cited a 40 percent increase in per-order spending among shoppers who use their respective AI assistants, an alignment neither company had any reason to coordinate on and every reason to state independently. Furner attributed the lift to Sparky, Walmart's in-app assistant, whose total user base is up 70 percent year over year and growing faster than the company's overall digital business. Jassy attributed the identical figure to Alexa for Shopping, now used by more than 350 million people in the past year, with active users nearly doubling over that same twelve-month period across the US customer base.
Two companies with entirely different AI architectures, different customer bases, and every competitive incentive to inflate their own numbers relative to each other landed on the same figure independently, in the same earnings week, without any apparent coordination between them. That is either a remarkable coincidence worth dismissing, or genuine evidence that AI-assisted shopping produces a fairly consistent basket effect once adoption crosses a certain threshold of scale, and having reviewed both companies' methodology as disclosed on their respective calls, we lean firmly toward the latter explanation.
What is actually driving the lift
Furner described the mechanism plainly on the call: Sparky builds a basket, and Walmart executes it through fast delivery, pickup, or in-store fulfillment, turning AI engagement into what he called immediate physical outcomes rather than abstract engagement metrics that never translate into a completed order. Walmart's president, David Guggina, went further still, describing the shift as evolving from traditional search toward intent-driven commerce, where the assistant infers what an occasion or context actually requires rather than waiting passively for a shopper to type a precise search query it can match against inventory.
That distinction, intent versus search, is the actual mechanism behind the basket lift, and it is worth sitting with. A search box returns exactly what a shopper explicitly asks for, no more and no less. An assistant that understands a customer is hosting a barbecue, say, surfaces the charcoal, the buns, and the drinks the customer never thought to search for individually in the first place. The 40 percent lift is best read as the AI successfully reminding shoppers of genuinely adjacent needs, not as some persuasive engine talking anyone into buying more than they actually wanted.
Target's number tells a different, earlier-stage story
Target's contribution to the same earnings week was structurally different from its two larger peers, and worth reading separately rather than lumping in with the other two figures. CEO Michael Fiddelke reported digital traffic from external AI platforms, meaning shoppers arriving via third-party AI agents and assistants rather than any tool Target itself built and controls, growing more than three and a half times faster than the company's overall digital baseline. AI-powered wish list creation jumped more than 50 percent heading into back-to-school season, a leading indicator rather than a lagging one.
That is a discovery-and-traffic story, not yet a proven basket-lift story on the order of what Walmart and Amazon reported, and the distinction matters more than it might first appear. Target is seeing the top of its funnel respond strongly to AI before it has a Sparky- or Alexa-scale number to report on what that incremental traffic actually converts into further down the line. Any CTO citing this trio of results side by side should be precise about which stage of the funnel each company is actually measuring and disclosing.
Why 40 percent should become a benchmark, carefully
For a retail CIO building an internal case for AI shopping investment, having two large, independent, competitively adversarial companies report the exact same order-value lift is more useful evidence than any single vendor's cherry-picked pilot results could ever be. It gives finance a concrete number to anchor a projection against, and it gives skeptical stakeholders a genuine reason to believe the figure is not simply manufactured for a favorable press release. We would treat 40 percent as a reasonable upper-bound target for a mature, well-adopted assistant operating at scale, not as a guaranteed first-year result for a newly launched program.
The caveat is attribution discipline, which is easy to state and hard to build. Both Walmart and Amazon can make this claim credibly because they can track a shopper's assistant usage against their subsequent order, at the individual transaction level, across a user base large enough to smooth out ordinary noise in the data. A retailer without that instrumentation already in place should build the underlying measurement pipeline first, before promising anything like this lift to its own board or investors.
The gap this leaves unresolved
None of the three companies disclosed what happens to margin alongside order value, which is a notable and telling omission from every one of these earnings calls. A 40 percent bigger basket is unambiguously good for top-line revenue, but if the AI is systematically upselling lower-margin categories, groceries and household staples in Walmart and Amazon's case specifically, the profit story could be far less clean than the headline number suggests on its own. That is the exact question we would put to any retailer citing this benchmark internally before celebrating it.
It is also worth remembering the earlier ceiling this industry has already found through independent research. Separate studies this year showed AI shopping agents remain far better at browsing and research than at completing a purchase end to end without human help. The 40 percent figure describes AI-assisted shopping within a retailer's own app, where the retailer still fully controls checkout, not autonomous agent-to-agent commerce happening outside that walled garden. Those are different problems on different timelines, and conflating them is the single most common mistake we see in board presentations right now.
