Product1 distinct publisher3 min readPublished
Shopify's COO says agents reading its Catalog convert at twice the rate of agents scraping a storefront, and that is the number the company is spending tokens against while its agentic sales are still small.
The Product Desk · Product desk
Compiled by The Product DeskSomething wrong?How this is made
The mechanism under that doubling is dull, which is usually a good sign. An agent sent to buy something either reads a merchant's structured product data or reads whatever it can pull off a rendered page, and Jess Hertz's claim is that the same agent traffic converts twice as well with the structured version [5]. Taken at face value, a merchant needs half as many agent-referred sessions to book the same orders [13]. Worth saying out loud, though: the comparison is Shopify's, run on Shopify's data, against a scraped baseline Shopify defines. Conversion is a ratio, and Hertz says agentic GMV is still small [6], so a doubled rate on a small base is still a small number of orders.
The token policy is the more interesting decision, because it is a classification decision. Fast Company's framing is that Shopify engineers can spend freely on tokens while the rest of the business world agonizes over ROI [2]. What teams usually tell themselves a budget like that buys is optionality. What it actually buys is the removal of a step: nobody building for agent channels has to assemble a payback case for a cost the company has already filed under plumbing. The tradeoff sits in the same sentence. If you never make anyone justify the spend, you find out which part was waste late, from a finance review rather than from the product decision that caused it.
Shopify can hold that posture because of the durability number Hertz gave: almost 90% of quarterly revenue comes from merchants who have been on the platform over a year [7], which leaves a little over 10% coming from merchants in their first year [14]. A base that renews does not need this quarter's experiments to pay for themselves. That is the precondition for the token policy, and it is the part that does not travel to a company with lumpier revenue.
Who this lands on is not Shopify. It is the person maintaining a product feed at a brand selling through Shopify, or through anything else, who now has to decide whether machine-readable product data is a project or a nice-to-have. The useful cut is two questions. First, does your product data exist in structured form outside your storefront's HTML. Second, can you measure any demand arriving from an agent today. If both are no, the work is instrumentation, not a catalog migration. If the data exists and agent demand does not, you have bought cheap insurance and can stop there. If agent demand exists and the structured data does not, you are the merchant on the weak side of the comparison Hertz described, and you can measure your own gap before you take her figure for it. If both are yes, the question turns contractual, which is where the merchant-of-record language in the Universal Commerce Protocol starts to matter [11].
Hertz says the industry is past the "SaaSpocalypse" and that complexity is both Shopify's challenge and its moat [12]. Read plainly, that is a bet that merchants would rather rent agent readability than maintain it, and the token budget is what Shopify has decided it costs to place that bet before the demand shows up in GMV [6].
Ranked by verification strength, evidence, and original report placement.
Hertz said the industry is past the 'SaaSpocalypse' moment and that complexity is both the challenge and the moat for Shopify, which has always acted as a unified operating system and aggregator of ways to sell.
Fast Company published an abridged transcript of an interview with Shopify COO Jess Hertz from the podcast Rapid Response, hosted by former Fast Company editor-in-chief Robert Safian.
Hertz said agentic GMV is small but definitely growing.
Hertz said almost 90% of Shopify's quarterly revenue comes from merchants who have been on the platform for over a year.
Shopify has agentic storefront integrations with ChatGPT, Gemini and Copilot.
Hertz said the Universal Commerce Protocol is an open commerce standard Shopify developed with Google, covering the transaction layer of commerce, under which the merchant continues to be the merchant of record.
Distinct publishers with included, body-backed reporting in this cluster.
1 article · September 1, 2026
Follow any of these and your For You feed starts watching them — no settings page required.
product
Model choice is becoming a line item, and the differentiator moved up the stack1 distinct publisher
invest
Scalable Capital puts ChatGPT, Claude and Grok inside the European order ticket2 distinct publishers
product
OpenAI shipped a teen ChatGPT. The over-65 cohort doubled to 23% and got nothing.1 distinct publisher
build
Under 30% citation overlap between engines makes pooled AI visibility scores unbuyable1 distinct publisher
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
One transcript, one speaker
Every figure that matters here — the doubled conversion, the near-90% revenue base, flat head count against 34% growth — leaves the mouth of the same executive and reaches us through Fast Company's abridged transcript. The tell is small and telling: 'double' sits inside square brackets, supplied in editing rather than spoken, and nothing in the piece points to a filing, a dashboard or a merchant who could confirm it.
Shipped broadly, transacting thinly
The pieces are named and in production: Catalog and Sidekick inside the platform, storefront integrations with the three big assistants, a transaction standard co-written with Google. What is missing is volume, and Shopify supplies that limitation itself — agentic GMV is 'small'. Breadth of deployment is genuine; depth of use is a word, not a number.
Framing outruns the quotes
The promise up front is engineers spending on tokens while everyone else agonises over returns. The answer underneath is more disciplined: healthy ROI, moderated cost, thoughtful speed bumps, constant monitoring, and Hertz volunteering that adoption is not impact. Add a signature 2x with no denominator and an agentic business the company calls small, and the direction of travel looks right while the confidence of the telling sits well ahead of what is shown.
House channel, house argument
A sitting COO talks to Fast Company's former editor-in-chief on a show from the Masters of Scale team, and Fast Company prints it. Each claim that lands doubles as a pitch: Catalog converts better than scraping, complexity protects the platform, merchants stay merchant of record so nobody should feel trapped. The one adversarial thread — will the AI platforms build this layer themselves? — is asked and then answered on Shopify's terms, and no cost figure for the token bill survives the abridgement.
Sure what was said, unsure it holds
Provenance is easy to pin down — we know the speaker, the show, the outlet and where the editing hand shows. Verification is the hard part: not one figure in the interview can be checked against anything else in front of us, so we can describe this story with confidence and certify none of its numbers.