Most board decks still file agentic commerce under “emerging,” somewhere between a pilot slide and a risk footnote. The premise underneath that filing is that the channel is too small and too unproven to fund at scale, so the responsible move is to watch it, learn cheaply, and revisit next year. That premise had a good run. In the week of August 19, 2026, it collided with two earnings calls.
Target and Walmart, between them roughly a fifth of US general merchandise and grocery spend, did something the wait-and-see posture was not built to survive: they put specific, audited numbers on AI-driven demand inside the same filings that carry their comparable sales. Once a growth rate shows up next to comps in a public earnings release, it stops being a lab experiment and becomes a line the market is pricing. That is the shift this essay is about, and it changes what a board should fund, govern, and stop.
The week the channel got a number
Start with what was actually said, because the specificity is the story. On its second-quarter 2026 call, Target reported guest traffic up 3.6% and comparable sales up 3.8%, and then CEO Michael Fiddelke added the sentence that matters: Target’s digital traffic sourced from external AI platforms is growing “more than three and a half times the industry” compared to a year ago, driven by agentic commerce partnerships with companies like OpenAI (PYMNTS). He hedged it as “still small in total today,” and that hedge is honest. It is also the point. A small base growing at 3.5 times the industry rate is exactly the profile of a channel that reorders share before most boards have approved a budget line for it.
The next day, Walmart supplied the demand-side mirror image. Global eCommerce grew 23% in the quarter, and CEO John Furner disclosed that shoppers using Sparky, Walmart’s AI assistant, spend 40% more per order, with total Sparky users up 70% year over year (PYMNTS). The same pattern surfaced across the sector inside that reporting. Amazon, after merging its tools into a single Alexa for Shopping assistant in May 2026, said more than 350 million shoppers had used it in the past year, active users nearly doubled year over year, and US customers who use it spend 40% more per order than those who do not. Albertsons reported average order value up 10% when customers use its conversational search and 26% when they use its fuller assistants. Shopify said AI-referred traffic and orders each tripled year over year.
Read those figures as one data set and the “experimental” label breaks. When Target, Walmart, and Amazon independently land on the same 40 percent lift and the same doubling-or-better user growth, that is a channel with a repeatable economic signature: it sources traffic faster than the channels it displaces, and the customers who arrive through it spend more per order.. A board that treats a channel with those properties as a watch-item is mispricing an asset that competitors are already compounding..
Why grocery moved first, and why that is not comfort
The honest counterweight in the earnings, and the one a good director should press on, is that nearly every hard number so far comes from repetitive, habitual shopping: groceries and household staples. Predicting what a customer needs again is a far easier task for an assistant than helping someone discover something they were not looking for. Albertsons’ 26% order-value lift sits in exactly that easy zone. The categories where AI is proving out are the categories where demand was already predictable.
The discovery side tells the harder story. TJX, parent of TJ Maxx, Marshalls, and HomeGoods, posted comparable sales growth of 4% in the same week and spent its call on merchandising rather than AI, which fits a business built almost entirely on in-store treasure-hunt browsing, the least scriptable behavior an assistant can replicate. Etsy shows the ceiling from the other direction: agentic traffic on the platform grew roughly 15 times year over year in the fourth quarter of 2025 and still sat below 1% of total traffic as of its most recent quarter. Fast rate, tiny share.
Here is why the grocery-first pattern should not put a board at ease. The mechanism that makes assistants win in staples, removing friction from a purchase the shopper already intended, is the same mechanism now spreading outward. Shopify’s data already shows AI reaching niche, non-repetitive purchases, with its share of sales outside major product categories holding steady since 2025. Consumer willingness runs ahead of the current use cases. PYMNTS Intelligence found that 49% of consumers interested in agentic AI would let an assistant complete both routine and larger, research-driven purchases on their own, and separately that 38% of AI users already compare prices across retailers with the technology while 36% use it to find deals (PYMNTS).
The strategic read is straightforward. Grocery and staples are the beachhead because they are the easiest wedge, and the wedge is widening into discovery categories on the same rails. A retailer that is present and machine-readable when the assistant is doing routine reordering owns the relationship when the same assistant starts handling considered purchases. Waiting for the discovery numbers to prove out before investing means arriving in that category after the assistants have already learned who to route to.
The visibility tax you are already paying
If the earnings calls describe the demand, Adobe’s traffic analysis describes the plumbing, and the plumbing is where most boards are quietly losing. Across more than one trillion visits to US retail sites, Adobe found that AI-referred traffic now converts 42% better than non-AI traffic as of March 2026, a record high and a full reversal from a year earlier when AI traffic converted 38% worse in March 2025 (Adobe). The behavior behind that reversal is measurable: shoppers arriving from an AI source engage 12% more, spend 48% longer on the site, and browse 13% more pages. Trust is rising with it, with 66% of surveyed respondents saying they believe AI tools deliver accurate results. AI traffic to US retail sites grew 393% year over year in the first quarter of 2026, on top of a 693% year-over-year surge during the Nov to Dec 2025 holiday window.
Then comes the part that should end the “we will get to it” posture. Adobe’s AI Content Visibility Checker, which scores how much of a page a large language model can actually read, found the average US retail homepage scoring 75%, category pages 74%, and individual product pages only 66%. Roughly a third of the content on the pages that actually convert, the product pages, is invisible to the models now routing the highest-converting traffic in the channel. The spread between the best and worst performers is wide: 82.5% for the leaders and 54.2% for the laggards.
That gap is a tax, and it is being paid right now. A retailer whose product data is machine-unreadable simply stays absent from the assistant’s recommendation, or appears with thin, stale, or missing attributes, and loses the exact visit that Adobe shows would have converted 42% better than average.. The cost is invisible on the P&L because it never becomes a transaction. This is the crucial point for governance: the loss from staying idle on AI readiness shows up as demand that quietly routes to the competitor whose product feed the model could read..
What “wait and see” actually costs
Put the demand data and the plumbing data together and the cost of doing nothing stops being a forecast. It becomes arithmetic a board can reason about without inventing a single number.
Consider the compounding. AI-referred traffic is growing at triple-digit rates and converting better than any channel it displaces. The models sit between the shopper and the shelf, and they learn. Each quarter a retailer is machine-readable and price-competitive, it accumulates conversions, satisfied completions, and the ranking signal that comes with them. Each quarter a competitor’s product pages sit at 66% readability, that competitor forfeits those same signals. The advantage is not linear. The retailer that shows up cleanly this quarter is more likely to be selected next quarter, which produces more selection after that. Target framing external-AI traffic growth at 3.5 times the industry rate is a picture of that compounding already underway at one operator while others debate the pilot.
Now the discovery timing. Boards taking comfort in “it is only grocery so far” are betting that the discovery categories stay off-limits to assistants long enough to invest later. The data argues the window is a matter of quarters. Shopify already reaches non-repetitive purchases, Etsy’s agentic traffic grew 15 times year over year, and half of interested consumers say they would delegate research-driven purchases. When the assistant starts handling a considered category, it will route to the catalogs it already trusts from the staples it has been reordering for a year. Readiness bought in December 2025 is worth more than the same readiness bought in December 2026, because the first buyer trains the router.
That is the reframing the thesis demands. The Cost of Doing Nothing on agent readiness is no longer a slide in a strategy deck. It is disclosed, quarter over quarter, in the filings of the operators who moved. Target and Walmart just told the market what forward motion looks like in basis points and multiples. Every competitor’s silence on the same metrics is now a statement too.
The operating model that fixes it
The fix is an operating model with four parts, and each of them requires only a number the market has already validated..
First, make your catalog machine-readable as a standing obligation, not a project. Adobe’s product-page score of 66% is the single most fixable gap in this entire picture. Structured, complete, current product data (attributes, availability, price, specifications, returns terms) is what determines whether an assistant can represent you at all. Treat AI content visibility as a monitored operational metric with an owner, a baseline, and a target above the 82.5% the sector leaders already reach. This is the highest-return work available, because it converts traffic that is otherwise routing to someone else.
Second, instrument the channel so it appears in your own reporting the way it appeared in Target’s and Walmart’s. If you cannot state your AI-sourced traffic growth rate and the order-value difference for AI-assisted shoppers, you cannot govern the channel and you cannot tell your board whether you are gaining or losing it. The metrics exist because these companies chose to measure them. Adopt the same measures: AI-referred traffic, conversion versus non-AI channels, and per-order spend for assisted versus unassisted customers.
Third, put an accountable owner over it. Target appointed Chandu Nair as its first chief AI officer, stepping in August 24, 2026, with a mandate to coordinate AI across guest experience, inventory, and decision-making. The title is less important than the accountability. Someone needs to own catalog readiness, the assistant partnerships, and the internal AI that supports them, such as Target’s Proxima digital twin of its middle-mile supply chain, as one coordinated portfolio rather than scattered pilots. Fragmented ownership is how the 66% product-page gap persists while everyone assumes someone else has it.
Fourth, decide the partnership and rails question deliberately. The assistants that drive this traffic sit on external platforms, and being present on them is a choice with terms. A retailer needs a stated stance on which agentic platforms it integrates with, what data it exposes, and how it protects margin and the customer relationship when a third-party assistant intermediates the sale. That stance belongs in front of the board now, while the terms are still being written, rather than after competitors have set the defaults.
The through-line is that every element of this model is defensive against a loss you cannot see and offensive toward demand that is already the best-converting in retail. Boards spent the last two years asking whether agentic commerce was real. Target and Walmart answered that in their August 2026 filings. The question for the next board meeting is narrower and more urgent: when the discovery categories cross over in the coming quarters, will the assistants be able to read your catalog, and will you be able to prove to your own directors whether you won the visit or gave it away. The retailers that can answer both are the ones treating agent readiness as an operating discipline today, not an experiment to revisit next year.