Shopify powers more than 14% of the U.S. e-commerce market, and three quarters of the orders its merchants attributed to AI last quarter came from product categories outside the 100 largest.
For two decades the way to be found online was to buy a place at the top of the search results. Jess Hertz, Shopify's chief operating officer, said AI agents reward a precise match to what the shopper asked for instead, which sends demand toward the small, specialized sellers that search results buried.
"So 75% of AI-attributed orders last quarter came from categories outside of the top 100, which I find particularly interesting."
Hertz has run revenue, operations, talent, legal and communications at Shopify since October 2025, joined the company as general counsel in 2021, and before that was staff secretary in the Biden White House, a job she calls the central nervous system of the West Wing.
I listened to the full interview so you can skip it. 30 minutes of audio, 15 minutes of reading.
Here are the 11 takeaways that matter.
👤 Guest: Jess Hertz, chief operating officer of Shopify since October 2025, who joined the company as general counsel in 2021 and was staff secretary in the Biden White House
🎙️ Host: Bob Safian, who presents Rapid Response, the interview show on the Masters of Scale network
📰 Published: 1 September 2026 on the Masters of Scale feed
🟢 Spotify | 🟣 Apple Podcasts | 🔗 Show notes | ⏱️ 30 min | ✅ Time saved: 15 min
Key Takeaways
AI shopping agents reward a precise match, and that moves demand to small sellers 75% of the orders merchants attributed to AI last quarter came from categories outside the 100 largest
The growth case rests on merchants who were already there, not on AI Almost 90% of quarterly revenue comes from merchants who have been on the platform more than a year
Handing an agent structured product data doubles the conversion rate against scraped data
Complexity is the defense against the AI platforms building their own commerce layer
Flat headcount for more than eight quarters against 34% revenue growth is the return Hertz points to She grants that adoption is not the same as impact, and says measurement is still being worked out
An AI agent inside Shopify's Slack has taken 15M actions in five months It sits in about 10,000 channels, and roughly 90% of employees have used it
The T-shaped employee is giving way to an X-shaped one Several areas of depth rather than one, and teams built from how those areas fit together
Merchants that grow past $1M in sales almost never leave Retention over the last five years was 92% for merchants reaching that size
1. Tariffs Land On Merchants
Safian opened with the tension in the setup: Shopify is a Canadian company headquartered in Ottawa, most of its customers are in the United States, and the two countries are in a trade war.
Hertz said the discomfort is external rather than internal. "I actually think it's really more outside of the company, to be honest, and we definitely think of ourselves as a company rather than a country."
The company's response to the first round of tariffs was a product, built over a weekend. "Even when the first kind of tariffs came into place, we had a team spending the weekend building a duties and taxes calculator." She said the macro conditions hit smaller merchants hardest, because they cannot absorb the cost without tools to adapt
Asked whether Shopify would behave differently if the tariffs were between two other countries, she said it would not: "Tariffs between any countries, we would be reacting the same way."
2. Politics Taught Her Pace
Safian asked whether watching the current politics makes her want to go back into government. It does not, and her answer was a defense of the crossover.
Her claim is that politics and technology run on the same skills, and that the resemblance is underrated. "It really teaches you how to drive hard, how to develop pace, and how to compress enormous amounts of information in such a short period of time." She started her career working for a behavioral economist in the Obama administration She tied the point to the present: everyone in AI is talking about compressing information and reducing the amount of human context a task needs
She pushed back on the idea that government is the slow one. "My last job in the White House was as White House staff Secretary, which people sometimes don't know, but it's basically the central nervous system of the West Wing and the White House." Most staff secretaries are lawyers, which she said she always found a curious trait of the job Her verdict on the pace inside the West Wing: "That pace is real."
3. The Sales Comp Had Drifted
Her first internal message as chief operating officer told staff the company had drifted from its core. Safian asked what had drifted. The answer was narrower than the phrase suggested: the pay plan for the sales team.
She said the compensation plan had stopped pointing at the same outcome as the business model. "It's not aligned with the merchant outcome and our business model at Shopify, which is that we only succeed when our merchants succeed."
Her stated reason for starting there is that pay design decides behavior. "How you set up systems really does impact outcomes." Her background is partly in behavioral economics; she worked for a law school professor of hers in the field
The fix was meant to be read as evidence rather than as a slogan. "It's actually how we compensate people and put our money where our mouth is and make sure that all of those directions point the same way, which is to make sure that our merchants are set up for success."
4. River Lives In Slack
Safian said chief executive Tobi Lütke has called himself a chaos engine, and asked how much of her job is reining him in.
She rejected the premise outright. "There is no reining in." She added: "There are sometimes chaos people and order people, and I think I'm a chaos person perhaps masked as an order person."
She said the memo everyone quotes was a design decision about the company, not a slogan. "When you really think about that, the memo made being reflexive in AI a baseline expectation for our employee base." She dated it to March 2025
The tool she offered as proof is River, an agent that works inside the company's Slack. "River is basically an agent that lives within our Slack, and you describe a task in a public channel." Every request is made in the open: "It all has to be out in the open in public, so that's a feature, not a bug." "It can read code, it can run tests, and it lives in about 10,000 channels in the company. About 90% of employees have used it." On usage: "It's run 275,000 sessions, and it's taken 15 million actions in five months, and it has this profound effect on how we work as a company."
Her theory of why it matters is about the speed information travels, not about the tasks done. "I really think the velocity that a company moves at can be judged based on how quickly information is moving within the company." River reads every public channel, which she said lowers the cost of a request and leaves staff relying less on human context She added that it improves with every interaction, so use makes it better
5. Agentic Is Upside, Not Core
Safian put the numbers to her: Shopify has posted a fifth consecutive quarter of growth above 30%, and he asked how much of that is share taken from Amazon and how much is AI.
Hertz split the answer in two — a durable existing business, and AI as the addition on top. "On the core business piece, the best statistic I can give you to think about this is that almost 90% of our quarterly revenue is from merchants who've been on the platform for over a year."
She was explicit that the AI-driven volume is still small. Sales made through AI agents are a small share of gross merchandise volume, the total value of goods sold across the platform, though she said it is growing "We do have early signs that Sidekick is helping merchants sell more, and merchants who are a part of Catalog are really starting to make more money."
The two products behind that are worth separating. Sidekick is the assistant merchants use: "For listeners who may not be as Shopify-versed, Sidekick is our AI-powered commerce agent that lives within Shopify." Catalog is the data layer pointed outward at the AI services: "On the Catalog piece, that's our product layer for agentic commerce. That's what makes products understandable, discoverable, and purchasable by AI."
The conversion figure is the argument for the whole product. "When someone comes from an agentic channel, the conversion rate when an agent is using our catalog data is 2x compared to if it's just using scraped data."
6. Complexity Is The Moat
Safian asked the question every infrastructure company gets: Shopify now plugs merchants into ChatGPT, Gemini and Copilot, so what stops those platforms building the commerce layer themselves?
Her first answer was that the fear of AI destroying software businesses has already been tested. "What I would say is I think we're past that SaaSpocalypse moment."
Her second was that the difficulty of the job is the defense. "The way I really think about that is complexity is both the challenge and the moat for Shopify." Shopify has always been the single system running an online store, a physical shop and now agent-driven sales, and merchants cannot easily run those channels themselves "So I think what's even more scarce in this AI world is that infrastructure that you're talking about, right?"
Safian pressed on the version of the story where AI lets small merchants compete alone. Hertz's answer was that AI adds complexity rather than removing it, and that the complexity is what merchants pay someone else to absorb
Asked whether that leaves merchants trapped, she pointed at who legally makes the sale. "I think the nice part about Shopify, even if you look at things like the Universal Commerce Protocol, which is the open commerce standard that we developed with Google, which is really the transaction layer of commerce, is that the merchant continues to be the merchant of record." The merchant of record is the business that legally sells to the customer and takes the payment, so an agent-driven sale is still the merchant's sale
7. Tokens Free, Headcount Flat
Safian raised a complaint he had heard from Uber's chief operating officer, Andrew MacDonald, about the hidden costs of AI, and said he had heard Shopify engineers can spend on model usage mostly without restriction.
Hertz said the return is there, and gave one number for it. "Headcount for Shopify has been flat for more than eight quarters, while revenue is growing 34%." The cost being discussed is tokens, the units of text an AI model charges by
She refused the easy version of the claim. "Now look, the idea of adoption is never going to be the same thing as impact, so that is definitely true, and we are thinking through ways to measure impact."
The spending is open but watched. Engineers are meant to pick the best model for the circumstance, with what she called thoughtful speed bumps built in along the way, and constant monitoring "We are a fairly pragmatic company at bottom."
8. X-Shaped Beats T-Shaped
The productivity claim led into what she thinks the resulting employee looks like. Shopify used to hire for the T shape: broad, with one deep specialty.
She said AI is pushing the company toward people with several areas of depth rather than one. "So you have multiple spikes of expertise, are able to absorb complexity much faster, learn new spaces and be a seven-out-of-10 designer if you want to be, right?" Her own example is that she now does her own data work, and has added commercial depth to a legal background
The point she said matters more is how those people are assembled. "It's actually that team combination that becomes the accelerant to incredible amounts of work that you couldn't even imagine." She described the work as finding where people's edges are and how they fit together
Safian tested it against an older word and she accepted the correction. He suggested the X shape is what used to be called a generalist, then reframed it as generalist specialists who are fitted together in different ways Her own image for it: "I think a little bit about Tetris, maybe."
9. Agents Find The Long Tail
After the break Safian framed the scale of the vantage point — Shopify powers roughly one-seventh of all U.S. commerce, and its merchants are mostly small businesses exposed to tariffs and price shocks.
Hertz said the seat is the point: she can watch competition change in real time. "We power over 14% of the U.S. e-commerce market, and what's incredible about that is you can sit at this kind of wide aperture of commerce, and you can see how competition shakes out in real time."
The first thing she said she is seeing is that starting a business has got faster. "Last year, about 10% of new shops made their first sale within a single day of signing up, which is just incredible."
The second is the change in what gets found. Search rewarded whoever was already popular or spending the most; agents, she said, reward a precise match to the request "So 75% of AI-attributed orders last quarter came from categories outside of the top 100, which I find particularly interesting." "And really that's the shift that we're seeing, which is AI can understand the actual need and not just a keyword."
She made the case with her own summer shopping problem rather than with a chart. Her eldest daughter was going to a ranch camp in Wyoming, and the packing list called for a trunk that would fit under a bunk "So I used my AI to find this specialized merchant in Wisconsin that's been doing this for 40 years." She did not know the dimensions of the bunk, and said that kind of seller is exactly who AI can help a buyer find
10. Big Merchants Don't Leave
Safian put the standard risk to a company selling to small businesses: your best customers grow up and move to a platform built for their new size.
Hertz said the range is the design goal, from the smallest merchant to the largest. "We want to be able to serve merchants starting around their kitchen table all the way up to Kim Kardashian at Skims."
Her evidence is retention at the point where a merchant gets big enough to leave. "Over the last five years, merchants that reached a million in GMV had a 92% retention rate, and merchants that reached 10 million had a 97% retention rate."
Safian named a brand he assumed had outgrown Shopify, and it turned out to be a customer. He raised the running brand On — "Now they're up to, whatever, $300 billion in revenue" — and said it would grow beyond what Shopify can do for it "They're a Shopify merchant, actually, and we've had great success with On as well."
Her explanation for why large brands stay is an architectural one. She said today's strategy options are a function of architectural choices made several years ago, and that Shopify built one unified data model and made it extensible "It's how a large brand can run 1,000 different campus stores, 1,200 storefronts, and six million SKUs on one stack."
Asked whether she worries about merchants growing away from Shopify, she inverted it: "I think it's a little bit reversed. We want to be able to help them grow."
11. Unsentimental By Design
Safian quoted her own description of herself as a deeply unsentimental person, and asked whether that is an asset or a hindrance. Her answer was both, and she traced it to a company phrase.
The phrase she named is "Keep your identity lightweight." She said it was one of the things she liked most about Shopify when she arrived "It's this idea that you can wake up smarter, you can wake up better the next day, and you don't have to get so attached to ideas." The practical form is holding ideas loosely enough to drop your own for a better one, which she said let her change careers without carrying the baggage of how things had been done
She rejected the idea that the company is actually chaotic. "I don't actually feel like we're a chaotic company." "There is a deep focus, and I think it's because of this purity of mission, which is really an alignment mechanism that is deeply held within the company."
Safian's reading, which she agreed with, is that the mission is the fixed point and the tactics are the part allowed to move
Bonus Insights
Hertz moves from Washington to Toronto next week. "Let me say, as of next week, I'm actually going to be a Torontonian." She said she gets energy from being with people in person and is not worried about losing her roots in the United States
Asked what is at stake, she gave the same answer as at the start of the company. "It is to be the 100-year company and to grow entrepreneurs everywhere. I think that is the bet, and we are going to win or die trying."
Safian's opening narration framed the booking around Lütke's memo, which he described as telling teams to prove AI cannot do a job before hiring anyone new
Safian's own aside on the X shape was that he is not sure he has made the change himself, and that the point for a listener is to work out how to adapt to the skill set the moment rewards
Hertz's bottom line is that AI does not take commerce away from Shopify, it makes selling more complicated, and the company that absorbs that complexity for merchants is the one that keeps their business as they grow.
Products, Companies & Tools Mentioned
Shopify (The platform Hertz runs, which she says powers over 14% of the U.S. e-commerce market and only makes money when merchants do)
Sidekick and Shopify Catalog (The in-platform AI assistant for merchants, and the product data layer that she says doubles the conversion rate when an agent uses it instead of scraped data)
River and Slack (Shopify's internal agent and the place it lives: 10,000 channels, 90% of employees, and 15 million actions in five months)
ChatGPT, Gemini and Microsoft Copilot (The AI services Shopify plugs merchants into, and the platforms Safian asked whether she fears will build their own commerce layer)
Universal Commerce Protocol and Google (The open commerce standard Shopify developed with Google, which she describes as the transaction layer that keeps the merchant as the merchant of record)
On and Skims (The two large brands she named as evidence that merchants do not outgrow the platform)
Uber (Whose chief operating officer, Andrew MacDonald, started the argument about hidden AI costs that Safian put to her)
Amazon (Named by Safian when he asked whether Shopify's five quarters of 30%-plus growth were sales taken from someone else)
Books & Resources Mentioned
Tobi Lütke's AI memo (The Shopify chief executive's public memo making reflexive AI use a baseline expectation, which Hertz dates to March 2025 and calls a design decision about how the company works)
If this was worth your time, send it to someone closer to the industry than you are.
Get the latest market chatter as it happens: