In the October 5 edition of the Wall Street Journal's CFO Journal newsletter, Jennifer Williams of the WSJ Leadership Institute reported a set of numbers that will get forwarded around a lot of strategy offsites this quarter. Shopify, which processes sales for merchants including Barnes & Noble, Claire's and e.l.f. Beauty, has pushed AI tools into production across pricing, product discovery, cart building and checkout. According to the report, AI-assisted shoppers spend more, and AI-driven orders and traffic to Shopify stores tripled year-over-year in the latest quarter.
Chief Financial Officer Jeff Hoffmeister connected that directly to the P&L. "The largest portion of our revenue is our payments business," he told the Journal. "So if a merchant does more sales, and they're using Shopify Payments, obviously some of that translates back to our financials." On agentic commerce broadly, he was more guarded: "I do think there'll be an inflection. I just don't know the pace of the change."
Here is the argument worth taking from this, and it is not the one the headline suggests. Shopify's results are a strong signal about discovery — how products get found, described and recommended by machines — and a much weaker signal about agent-completed purchase, the part of agentic commerce that gets the most airtime. Leaders reading this story should fund the first and stay patient on the second. The two require different investments, carry different risk profiles, and pay back on completely different timelines.
What Shopify Disclosed, and What It Left Out
A tripling is impressive in percentage terms and uninformative in absolute terms. Shopify did not, in the Journal's account, give a base: no share of total GMV attributable to AI-driven orders, no dollar figure, no breakdown between "a shopper used a conversational tool to find this product" and "an agent transacted autonomously." Those are very different behaviors with very different implications, and they appear to be bundled into one growth statistic.
This is also a vendor reporting on its own product. Shopify has a direct commercial interest in merchants believing that AI-driven demand is arriving fast, because merchants who believe that adopt Shopify's AI tooling, sell more through Shopify Payments, and feed the take rate Hoffmeister described. None of that makes the number false. It does mean the appropriate read is directional, not decisive.
What makes the disclosure more credible than most AI-adoption claims is that the mechanism is boring and traceable. Shopify is not claiming productivity magic or headcount avoidance. It is claiming that more transactions flowed through a payments business it already monetizes. That is the kind of AI result a CFO can actually audit, and it sets a useful bar: if your own AI program can't name the revenue line it touches, it isn't at the stage Shopify's is.
Outside Data Says the Direction Is Right
Shopify is not an outlier, which is the strongest reason to take its figures seriously. Adobe Digital Insights has been tracking AI-referred traffic to retail sites, and the trend line is steep across its readings. Adobe found that AI-referred traffic to retail sites doubled year over year. Earlier in 2026, Adobe data showed AI traffic to U.S. retailers up 393% in the first quarter, with agentic shoppers outspending human shoppers.
Looking forward, Adobe projects U.S. online holiday sales of $275.1 billion for 2026, with AI-sourced traffic up 130% over the same period — meaning Adobe expects AI shopping referrals to more than double this holiday season. Multiple Adobe readings, one platform's internal data and a forward projection all pointing the same way is about as much corroboration as a leader gets before having to make a call. It is worth noting that those outside readings come from a single data provider, so they confirm a direction rather than triangulate a magnitude.
Notice what all of these figures measure, though. Traffic. Referrals. Visits. They measure arrival, not autonomous purchase. The evidence base for AI reshaping how customers find products is substantially stronger than the evidence base for AI replacing the act of buying.
The Checkout Layer Is Still Contested
Shopify has gone further than most on the transaction side. The company allows AI agents to complete purchases on behalf of shoppers, including through a partnership announced in September with Meta Platforms that lets Meta's Muse agent check out on behalf of customers of Shopify stores. That is a real capability shipped with a real partner, and it is more than most platforms can say.
But the broader market has been wobblier here than the hype cycle admits. Digital Commerce 360 reported in March that OpenAI shifted its checkout plans within its agentic commerce strategy — a meaningful wobble, given that OpenAI's Instant Checkout and the Agentic Commerce Protocol have been treated as the de facto standard-setting effort. A separate analysis from Digital Applied argues outright that AI checkout has stalled, and that the pattern actually winning is "discover in AI, buy on site."
That counter-thesis deserves weight, because it is consistent with the Adobe numbers rather than contradicted by them. Traffic referred from AI surfaces is exploding; agent-completed transactions are a thinner, newer slice. Hoffmeister's own hedge — he knows an inflection is coming but not its pace — reads less like executive caution and more like an accurate description of where the technology sits. Questions about authentication, liability and dispute handling sit behind any move to let software spend a customer's money, and nothing in Shopify's disclosure addresses them.
The Long Tail Is the Real Tell
The most interesting detail in the Journal's report is also the least quantified. Hoffmeister said specification-heavy goods — electronics, furniture — are seeing a boost from AI-driven shopping, and so are items outside Shopify's top 100 product categories, which he reads as a sign that AI tools are surfacing more niche inventory. Retailers, he added, are rewriting product descriptions with AI in mind: noting that a suit is made with breathable fabric, or is suited to warmer temperatures.
This is the structural change, and it is larger than any one quarter's order growth. Conventional e-commerce discovery is bestseller-driven: ranking algorithms reward volume, reviews and ad spend, which concentrates demand on a narrow head of the catalog. A conversational agent that can parse specifications works differently. It matches a stated constraint — breathable, under $300, ships by Friday, fits a 32-inch alcove — against structured attributes. Products that lose on popularity can win on fit.
For merchants with deep, differentiated catalogs and weak brand recognition, that is the most favorable shift in discovery economics in a decade. For brands whose position rests on being the obvious default, it is a quiet erosion. Either way, the asset that determines outcomes is not ad budget. It is the completeness and machine-readability of product data.
Adobe has flagged exactly this gap: AI traffic is surging while many retail sites are not machine-readable, limiting their visibility in AI-driven search. Shopify's merchants are already closing that gap — one of the company's tools exists specifically to make merchant products readable to AI agents generating consumer recommendations. Most companies outside a modern commerce platform are not.
Who Captures the Value Is a Separate Question
It is worth being clear-eyed about why Shopify is enthusiastic. The company monetizes flow. More merchant sales through Shopify Payments means more revenue, regardless of whether an individual merchant's margin improved. The merchant does the work of restructuring product data; the platform collects on the resulting transactions. That is not a criticism of Shopify's strategy — it is a well-constructed one — but it should temper how directly a merchant reads Shopify's results as their own.
Amazon is playing the same game from the other end. PYMNTS reported that Amazon positions its Rufus assistant as its edge in the agentic commerce race, pairing conversational discovery with agentic buying features. Circulating estimates of Rufus's sales contribution range from roughly $10 billion to $12 billion depending on the source, and those figures come from secondary commentary rather than Amazon's own disclosures — treat them as unconfirmed. The directional point stands regardless: the largest platforms are racing to own the layer where product discovery happens, because whoever owns that layer sets the terms for everyone selling through it.
The strategic risk for a brand is not that agents fail. It is that agents succeed on infrastructure the brand doesn't control, mediated by recommendation logic it can't inspect, with the brand reduced to a well-tagged row in someone else's catalog.
What Leaders Should Fund This Quarter
The practical split follows directly from the evidence. Discovery is proven and underfunded. Agent checkout is promising and unsettled. Budget accordingly.
- Audit machine-readability first. Run your own product and service catalog against the question Adobe raises: can a machine parse your attributes without rendering your page? Structured data, specification completeness, and attribute coverage are now discovery infrastructure, not an SEO chore delegated to an agency.
- Rewrite descriptions for constraints, not for atmosphere. Shopify's merchants are adding breathability and temperature suitability because those are the terms customers state to an agent. Copy written to evoke a feeling does not match against a stated requirement. Inventory your top 200 items and check whether a specification-based query could find them.
- Instrument AI referral traffic as its own channel. If your analytics bundles AI-sourced visits into "direct" or "other," you cannot see the trend Adobe is measuring and will not notice when it bends. Separate the channel, track conversion against it, and report it monthly.
- Look at your long tail before your bestsellers. The Shopify signal is that niche and specification-heavy items benefit disproportionately. If you have a deep catalog you have historically underinvested in, that is where the incremental return now sits.
- Pilot agent checkout; don't re-platform for it. With OpenAI having already shifted its checkout plans and credible analysts arguing the pattern is discovery-in-AI and purchase-on-site, committing a multi-year roadmap to any single agentic transaction standard is a bet on a specification that is still moving. Build the integration points; don't rebuild the funnel.
- Settle liability before you enable autonomous purchase. Who owns a wrongly executed order, a mispriced item, a disputed charge made by software? Get legal and finance to answer in writing before the capability goes live, not after the first chargeback.
The Honest Part of Hoffmeister's Answer
The most useful sentence in the Journal's piece is the hedge. A CFO with a tripling metric in hand and every incentive to declare the inflection has arrived instead said he doesn't know the pace. That is the correct posture, and it maps cleanly onto what the outside data supports: a confirmed, fast-moving shift in how products are discovered, sitting on top of an unresolved question about how they will be bought.
The good news for leaders is that the two investments are not symmetrical in risk. Cleaning up product data, writing for specifications, and tracking AI referral traffic as a distinct channel pay off whether agentic checkout inflects next quarter or in three years — they improve conventional search, conventional conversion and conventional merchandising in the meantime. Betting the roadmap on a checkout protocol that one of its principal backers has already revised does not. Move fast on the layer where the evidence is settled, and let the pace of the rest reveal itself.
