Shopify ‘26 Winter Editions: Tools for the Commerce Renaissance

Brian: Hello, and welcome back to Future Commerce, the podcast at the intersection of culture and commerce. I'm Brian.

Phillip: I'm Phillip.

Brian: And we have a very special guest today, Mani Fazeli, VP of product at Shopify. Welcome to the show.

Mani Fazeli: Thanks so much. I'm really glad to be here.

Brian: Yeah. We're happy to have you. Mani, tell us a little bit about what you do. I don't think Shopify needs an intro. VP of product, I mean, that's also pretty clear, but why don't you say a few words about your job? {laughter}

Mani Fazeli: Yeah, that's a good point. I've been at Shopify for about five and a half years. I've worked in all sorts of areas of the company. Right now I oversee everything to do with the checkout, which is quite popular. People care about that.  I spent quite a few years building the B2B wholesale tooling at Shopify, but I also oversee lots of other parts, including all the merchandizing capabilities. We have, products, pricing, collections, and the sort...Everything to do with fulfillment and order management and shipping. So, lots of different parts of the experience, customer accounts... So I cross cut enough of it that I have a really good perspective on what happens at our company, how we think about building product, why we build it, who we build it for. And then lastly, I've spent quite a bit of my time and energy on agentic commerce. So that happens to be, I think, pretty important to the conversation we're about to have together.

Brian: {Laughter} Just a little.

Phillip: And I think that's the most important part of it, I think, for a lot of our listeners plans going into 2026 as we're gearing up for everybody's strategic conversations like, where are we putting investments for next year? Brian, we keep talking about like, what are the big themes for next year, right? And I keep saying it's like in one lane, it's like autonomy, right? It's like the age of autonomy. And on another––I think it comes down to people having more agency, right? Like, so we're going to have... It's autonomy in one lane. And then, sort of like personal sovereignty in another. And I think that having these AI, you know, this... It's not just AI revolution, it's agentic commerce is enabling both of those two thematics. Going into 2026, Shopify has made a lot of headlines for AInvestments. Mani, tell me a little bit about you're looking back at how the sectors evolved and your philosophy around AI.

Mani Fazeli: Yeah. So we've got Winter 26 edition coming. It's going to have quite a few things announced. I think, at the time of recording, we're just a couple of days away. And so it's going to have this theme called The Renaissance because we really are in another Renaissance era, very similar and transformative as the original. And there's a little play on words in there. "AI" is in the word renaissance. And so you think about all the ways in which we're looking to help merchants and help their buyers, and taking ideas such as Sidekick, but not Sidekick alone. So, Sidekick sits at the center. It's supporting this entire complex ecosystem of AI, and it's trying to be that intelligent hub. However, it goes much further than that. But really, if you rewind back a bit and think about, you know, what was Shopify's philosophy, how did we get here? Why is it that we think about AI as sitting around all things that we do in supporting merchants for this next leg? That's going to be quite revolutionary. And so the first thing that we had a mantra that we said, "be early and be maxi." So adopt tools internally as early as possible. We ended up building quite of immense toolset. We created an LLM proxy that is able to cut across many different AI gateways and APIs. We use a system called LibreChat. We have many MCP tools. For example, we have something called Vault, which is our internal project management tool, something called Scout, which is our internal merchant feedback tool. We have MCP tools for all that, and we gave this to every employee and said every day, every piece of work you do should be leveraging AI to make you better. And in doing so, we're dogfooding the very systems that then allow us to imagine what we could build for merchants. So that was one of the big things. Be early, be maxi all in on AI.

Mani Fazeli: The second thing was that we have this mantra of "utility above being flashy." So, go right for the heart of what makes a difference in the merchant's life every single day. Helping with automated taxonomy, helping with generation of content. Helping with image manipulation. Helping with writing code. We have an interesting one with that where Sidekicks can help with that. Being a general assistant, be able to answer all the questions or more esoteric things that someone wants to do, and they don't know how to do it. The amount of time they have to spend in a forum or help docs and such. Let's cut that down. So we see this as the means of driving value, and that's the philosophy to and Sidekick the whole idea of having a co-founder for the merchant. And then lastly, it was this idea that the philosophy said platforms before features. So all the hooks and actions that are necessary in the admin for Sidekick to be able to do the magical things that it does. For example, if you've used it, you can see it does this thing that says preview the behavior you've asked for, and then you click it and it autofills everything for you and tells you, okay, is this what you wanted to do? Um, it took a lot of work to build a platform that allows an agent to cut across all these capabilities inside the admin. Also all the product data that lives in Shopify. How do we understand it, manipulate it, categorize it, clean it, and then be able to syndicate it for merchants so that they have reach.

Mani Fazeli: All these capabilities of platforms before features was really quite important. Maybe the last anecdote I'll share on the philosophy is every company that starts in the world of AI and wants to do something meaningful will have a bit of a cold start problem. Even if they have lots of data, that does not mean they can magically turn that data into some sort of an alpha for their company to generate capabilities in AI. So what we did was we built simulators in the early days of figuring out how should an agent that's trying to interact with a human behave and what kind of answers it should it give and to what length, with what accuracy. All that needed training data. So we created simulators that were able to generate thousands of questions and then have LLMs be able to pretend to be in different verticals and different segments, different maturities of merchants, and then feed all the topics that they would generate into yet other LLMs that were prompted to act like human personas. And you'd get these synthetics conversations that you could generate between the two. Yeah. Then take all that and put it into an eval. There were a lot of smart ideas at Shopify for how to be able to get us up and running. And this was all part of philosophy of what is it we think we need to do so that AI makes a difference.

Brian: There's so many things to unpack here. Man, it's like, it almost feels like a Black Mirror episode practically. The one where they have the the dating simulator take over before the people even meet. It's...we're feeding data into the AI for the AI to talk back to us about yeah.

Phillip: Oh yeah. The holodeck. We can trial things.

Brian: We can trial things. Right. Yeah. And we can trial them ahead of even releasing them, which is incredible. One thing that I really love about this approach is it gives you the hands on experience to know what merchants are up against. And I think Shopify has famously encouraged employees to have their own stores and to have their second shtick, as it will. And I think that's huge in the process of of building an AI first brand. Because what we're looking at right now is and I think you've alluded to this a little bit already. You have net new businesses that are building AI first brands where it is a cofounder to their company. And then you have a whole other set of companies that are in the plus range that are gonna use AI in a very, very different way. And the change management process is going to be very, very different than someone who starts with an AI native. And so, how did you think about that? Because I think the way that people use AI is going to differ immensely depending on the stage of organization they're in.

Mani Fazeli: I think that AIs going to absolutely create some differences in how businesses approach the opportunities in front of them. You can imagine that a company that's just starting out, how fast they can bootstrap all the different activities they have to go through. They have an idea, they have to figure out how to manufacture, they have to figure out how to source, they have to figure out how to market, they have to figure out how to fundraise if that's part of it. They have to figure out how to distribute, how to get affiliates and partners, whether it's wholesale, whether it's direct to consumer, whether they should establish retail. You think about all the parts that go to bootstrapping a business. AI can be assisting in every single one of those places. It's not always just the idea of what you get inside of a Shopify admin. It's every decision, every spreadsheet I need to generate to be able to model something out for the business, or every PO that I need to work through with manufacturing that might be overseas, every deck I need to produce in order to garner some sort of interest in maybe funding part of my business... aI will be able to influence all of that. And the companies that start out with very, very few resources or people, maybe it's a single individual entrepreneur who's brave and has this vision they want to bring to life, they will be forced to do things differently than what the incumbents are doing today. And thus, you should imagine that the very nature of the business, the processes of the business, the DNA of that business, will look different.

Mani Fazeli: This will take some time to play out, but you can already get a sense of it from, let's say, the tech startup ecosystem. They might be a little bit earlier than, let's say, the traditional direct-to-consumer goods-based commerce. You can see that there are the companies that are scaling their revenues and scaling their user base dramatically based off of not just what AI product they bring to life, but the very nature of the number of people who got together and used tools to automate things so that they could build that business. I think that your thesis is correct. However, I also think that we should not imagine that there's going to be some grand stratification that is going to result in businesses being overtly different. In the end, every one of those businesses has to worry about where their buyers will be. The presence of the buyers is what dictates whether they can run a business that's effective. So ones that become like AI Maxi and only sell maybe in one way versus ones that use AI to augment all the different channels they sell in, you can imagine the ones that go more narrow will have a limited TAM, whereas the ones that go broad will be able to reach more buyers and more destinations. So I think that the market forces will have AI become, dominantly, an optimizer, and then of course, you may always see something at the fringe. Just like, prior, there were companies who decided to sell only on social, and that was just way that they were approaching or maybe only video livestream. That's just the way that they were approaching selling. But it limits their TAM.

Brian: I think it's interesting. I agree that it does some of their TAM if they go AI only and they're gonna have to expand their channels. No question. And employ AI throughout the process. Of what I was referring to, and I think you addressed most of it, was just the change management process. That's a part of being in an enterprise organization. Introducing a a net new tool that has so much power and capability and not having plans to, like, use it from the beginning is really hard to manage through.

Mani Fazeli: Yeah, totally. I think Tobi––I can't remember when––he published an internal memo that then leaked and I think everyone saw it, which was this whole idea that we are all in on AI. AI will be the grand accelerator of our ability to serve merchants and our ability to serve the market in totality that each and every one of us every day has to ask ourselves, how can this tool accelerate us? And taking that to heart requires like a hard pendulum swing. But then also, as I mentioned at the top, you've got to make the tools available, right? Like every single person at this company could use cursor. Every single person at this company has access to cloth code. Every single person at this company has access to LibreChat. All the tools, every model, I could go in and dial the models to the max. I can use it from anyone whether it's from Gemini or whether it'll be ChatGPT versions or whether it'll be anything from Anthropic and all of them. I get to play around with every single one. They have different personalities, different skills, different strengths.

Mani Fazeli: So it's all available to me. You can't get an enterprise to shift without really swinging pendulums hard. So we've lived it firsthand. I can tell you that's what it takes. I agree. You know, the other thing that occurred to me...it just occurred to me, an example of the stratification you might get in industry. Remember when delivery became a really big deal in ordering from a restaurant? We eventually had the advent of cloud kitchens and this idea that I'm not going to try to sell to my buyers everywhere they are. I don't have a retail presence that they could come in and sit down at my restaurant, right? But I do serve them only in this one way that you can order online and either pick up or have it delivered to you. So I think that there is some precedence for technological change also shaping the way commerce happens. So we we probably will see some of that with AI, but no one can tell you what exactly is gonna happen in this future. But, uh, that just occurred to me.

Phillip: We shape our tools and thereafter our tools shape us. Right, Brian? There's the, uh, I think that there's that out there.

Brian: Marshall McLuhan quoting somebody else. Right.

Phillip: Yeah. That was right.

Brian: Actually, Marshall––somebody summing up Marshall McLuhan is what it was. {laughter}

Phillip: Yeah. Uh, Michael Scott, Wayne Gretzky, uh, whatever. Mani, you touched on SimGym earlier when we were talking about this democratized testing capability. That's something that in my agency days, I spent a decade helping these businesses that did a billion plus GMV, spent a ton of time doing these sort of like simulated tests, running high traffic tests, doing high volume tests. We had multiple staging QA environments and like private clouds. We did all that sort of thing. Of course, we didn't do it with AI back then. You know, it seems like there's like this level of democratization now in levels of testing and simulation that SimGym is this sort of thing that you're now bringing levels of capability to every kind of a merchant. And I think that's sort of been Shopify's mantra is that you're giving tools that have only been reserved for enterprises to everybody. It's That's what you do. When small merchants can simulate these thousands of interactions across their storefronts, do you think that there is a risk of you know, any sort of like a homogenization? Like they are depending on a Sidekick or ChatGPT. Is there a feedback loop where every merchant's trained on the same patterns and creating sort of like an invisible monoculture of some sort? Like they're basically getting the same outputs. What's preventing them from ostensibly recreating the same inputs?

Mani Fazeli: Yeah. That's a deep question. One that's challenging to answer without having things in the wild and seeing what merchants do with it. But here's my hypothesis. First of all, just earlier I was talking about creating simulations of merchants. SimGym that you're referring to is a new research preview that we're launching at Winter Edition that's different. It's about simulating buyers. Let me explain it a bit and then I'll answer your question about homogenization. So we have this incredibly unique application of AIn our opinion. I think it might be the first in the world or first of its kind and it's designed to help merchants explore the potentials of their ideas using simulations. So the thing that makes us special is we have always been obsessed with helping small businesses succeed the way you said We're constantly looking for ways to give them the superpowers that larger, more established brands have. One of those superpowers is confidence. The confidence that an idea that I have is likely to work and that I do not necessarily need to run all those AB tests and all the validation techniques in order to be able to ship boldly. And we want every entrepreneur to feel brave that way when it comes to trying new ideas. So SimGym uses AI agents that are human like profiles that simulate shopper behavior on storefronts. So these shoppers end up comparing themes. You give it a pair of themes like your live theme and a new theme change that you want to test. And then it recommends changes to those themes before you end up launching so that merchants can be bold and get it out there without necessarily being high traffic. And the way that it does that is that we've had billions upon billions of purchases on our platform and we do so every year.

Mani Fazeli: So we've been able to use all the data at our disposal to train the personas to mimic human behavior and then we're able to tune it based off the traffic of stores and the traffic of segments and demographics and verticals.

Phillip: Right.

Mani Fazeli: And we can create personas such as, let's say, a window shopper or a high intent shopper who will speed run to a checkout and give those different kinds of personas to your store against these themes. So the idea here is it's not meant to say that you wouldn't run AB tests and such if you have that traffic, is that you have yet another piece of signal that allows you to imagine whether your ideas have merit and whether they're worth betting on, and you're not playing as much whack-a-mole, lot of guessing game, and it's a tool in the tool belt. And that's why it's a research preview, by the way. Obviously systems like this are incredibly difficult to build. We spent a lot of energy and time in tuning it. Now we need to put it out there. We need the real world to be using it and then we need to make it more effective and efficient. Also, it's AI based systems. AI based systems have a cost structure around them. We need to make sure that when we're running all these virtualized or simulated agents that we're able to then give that at scale to the whole world so that every one of those brand new entrepreneurs can take advantage of it and that it doesn't just become a tool for the big merchants out there who may be able to throw more money at running simulations and such.

Phillip: Yeah. Okay. That makes a lot of sense. I didn't realize that this was part of Winter Edition. I think that definitely having more context is giving a lot more background for our listeners. Think one of the important pieces to understand here too is running these simulations gives us better information before things go out into the wild. I think a lot of folks are necessarily risk averse in these uncertain times. I think that's been one of the challenges of e commerce for the 24 years, 23 years that I've been doing this job, is that we do spend a lot of time and effort perfecting creative. We spend a lot of time and effort perfecting everything, product detail pages, and landing pages. And as highly tuned as they are, customers are unpredictable. I think having any amount of testing and higher amount of certainty before you get to putting things out in front of a customer is a boon to the merchant. So this is an interesting notion to be able to do that, not just in a user testing environment where you might have done it with a trial with real people. This is simulated. And I really I think that's really interesting, especially in a agentic capability.

Brian: It is the Black Mirror episode. It's––

Phillip: Well, that sounds dark when you say it that way.

Brian: No, no. No. I mean, that's like one of those things.

Mani Fazeli: Rebrand it. We need to rebrand it as "white mirror" or something like that.

Phillip: Yeah.

Brian: Totally. No. And and that's like one of the least dark episodes of of Black Mirror of all time, actually. {laughter} And then––well, depending on your viewpoint of, like, being matched up by an AI. {laughter} Actually, to that point, what we're gonna start doing, though, and actually, Future Commerce, we've alluded to this in prior content. But like, the goal is gonna start to be to build for these sort of archetypes that are represented by AI.

Phillip: That's right. Yeah.

Brian: Rather than for individuals. And that's not necessarily a problem, actually, because what that does is allows you to engage with groups of people that an AIs only gonna make them more and more, like, granular. And you're gonna be able to target by different different, like, psychographic markers to build these cohorts eventually. I'm probably not out of the gate. But, I mean, if we're looking down the line, this actually could be really, really powerful and, like, good, because you're gonna be able to address specific groups' needs and test against those needs before you even get to putting stuff in front of them. It saves time and money for everyone.

Phillip: Yeah. Tell us a little more about the the levels of integration, you know, with other, you know, partners. I know seen a lot of, you know, conversations like Harley's written a lot about the ChatGPT integrations and shopping agents. Give us a little more about how those journeys, especially on the consumer side, are being powered through Shopify platforms these days.

Mani Fazeli: Yeah. I'd be happy to. This is a landscape that's changing basically every month. So there are announcements and companies having new capabilities and new partnerships. And so there's a lot changing all the time. So I'll give you a sense of where we fit in, what we're trying to do, but also how I see the market and where the market's at. So the first thing is there is absolutely no doubt that discovery is changing, and chat interfaces are playing more of a role. So ChatGPT is, according to their own numbers, now accounting for 10% of search. That's not corroborative, but let's take them at their word. Google is, very heavily now, using AI mode and Gemini, they recently announced, has 650 million active users. So what does this all mean? We've evolved our nature as humans to go from search that was one-shotting to one-shotting using keywords, one-shotting with an intent to get to a list of blue links and then maybe some ads and deciding which ones xwe want to follow. Now we're in this multi-turn interaction mode with deep faceted searching. We're able to refine, we're able to compare, we're able to do a lot more of our window shopping using facts and not just marketing. We can mix the reviews of experts, we can mix the reviews of forum members, get that with merchant claims, blend it all together.

Mani Fazeli: Right? So our personal desires and requirements are now getting vetted much more quickly and they are at our fingertips, mobile, desktop, wherever we go. So that's the first thing that happened. And once you kind of get there and you realize how much discovery is changing, you realize, well, okay, discovery isn't necessarily just about getting to the point where you are now making the purchase. It's about finding the brands and then craving to see what else those brands have available, What else is in the lineup of products that maybe the chat recommended or distilled for you? Seeing what additional details and materials are available that hit the cutting room floor of any LLM. LLMs are designed to give you succinct answers, right? They're not meant to be super long. Maybe the cynical view would be because of cost, but then the generous view would be because humans don't have the attention span to want to read that much. We're more visual. We want to get to the meat and potatoes of what is this thing I'm really interested in. So the journey does end up oftentimes coming back to an online store and it does come back to discovering in this mixed mode way. I'll use actually an example just over BFCM. I was shopping for a new winter coat because where I live winter is actually upon us.

Mani Fazeli: I definitely use Gemini to help shortlist what are the brands that meet a set of requirements that I have. But in the end I wanted to actually go to each store, take a look at them. It's a sense of fashion, a sense of color, opportunities, options and I didn't end up buying any of the ones that were recommended to me but I did buy from one of the brands that was recommended to me. It was actually Moose Knuckles. Congratulations, you have my purchase. I hope it ends up being an excellent purchase. So when you, you know, we've discussed it before also in many other forums, uh, that maybe we don't need to elaborate here, but humans do actually enjoy the process of shopping. They're not looking for everything to be turned into some sort of utilitarian exchange that LLMs are able to reduce and then there's just the buy button and you click it. So this whole multi touch multimodal thing is I think really important. So what Shopify has done to make all this possible is that we've broken this up into, yes, there is the discovery portion. How can we support that? There's the brand interaction portion. How can we support that? And of course, there's a transaction portion. And how do we support that? And so we've announced this thing called the Shopify catalog where we are solving the problem of how discovery of products happens on the internet.

Mani Fazeli: Surprise, surprise, it's not as easy as just open web search. Even with the super talented companies out there like Google and others who are crawling the web, well, you need accuracy, you need freshness, you need a deep understanding of product models, you have option matrices, you need to normalize cluster filter, all this stuff has to happen. We've been doing that with the Shopify catalog and then syndicating that data to the partners that we work with. For example, the announcement with OpenAI. Then of course, we have systems like the knowledge base app, you to create FAQs, allow you to have citable information on your online store so that more brand interactions can happen if you have questions inside the LLM. And then finally, being the system that brings the merchants customizations about a checkout and making sure that a correct and accurate and fulfillable order is captured and that the transaction happens with the merchant. There's no one actually sitting in between. Someone facilitated getting to the merchant, but the merchant still owns that relationship and that customer. And these are all the places that we're playing. We're making sure that the merchants are never cut out of this process no matter who ends up being the company that facilitates that order happening.

Brian: This is like anecdotally, totally dead on. Like, I just did some shopping for myself for Christmas because my wife's like, hey. I wanna get you some ski boots for Christmas, but you're gonna have to be the one to, like, figure out what you need because ski boots are a particularly difficult thing to buy. And so I was like, alright. Well, I'll buy it over BFCM because that's when there's gonna be the best deals. And so started, like, just kinda kicking around and looking around at different sites, see what I could find. Eventually, I was like, you know what? This is ridiculous. I'm just gonna use ChatGPT. And so I started asking it different questions. What was unbelievable directly to your point is that I ended up buying from one of the retailers that was suggested. Um, and the data that they provided was up to date. But what was incredible was I ended up getting prompted by the LLM to asking asking more questions about what resort I was skiing, how much my height and weight were, and other markers about things other than, like, just my shoe size and my last size. Right? And so it actually pushed me down the road, and that caused me to then eventually call the retailer and talk to I actually talked to the retailer for, like, twenty minutes and got more information as a result.

Brian: And I ended up buying a different boot than the one that was recommended, but the only reason that boot was recommended to me was because I put in some specifications that said that it should be recommended to me. And by by, like, ChatGPT pushing me and then by talking to someone at store, like, I ended up triangulating on a specific pair of boots and that's what I ended up buying. There's no way I would have come to that without having interaction from both the retailer, their store, which I was browsing around actively with the associate, and then ChatGeeBT. And, like, I think that's sort of the end game, actually, for especially for mean, this is obviously a relatively complicated purchase, but, like, the the end game is to have direct interaction with the retailer, you know, be on their site to verify and and and confirm things as you go, and then and be prompted and pushed by ChatGPT to get to the right thing. I think that's endgame.

Mani Fazeli: I agree. And I think that there there's a there's a lot of hoopla right now about agents doing automated shopping. That will come, and it will have its place. But there's also some maybe projection of doom that's out there as a as a topic that I think is is overplayed.

Brian: Well, we certainly couldn't do what I just talked about for toilet paper. Like, there's there's I can't imagine doing what I just did for toilet paper, and that was the kinds of transactions that like well, we saw sort of the Amazon model fail, the Dash Button idea kinda failed. And so maybe even Agencik, like, that that like, until we have tools that just, like, fully understand exactly what our inventory is, personal inventory, and households become full on businesses as a result of the level of data that we have, I think that a gen tech's gonna have trouble because it's really built for, like, most...

Phillip: Moderating personal consumption. I need IOT for my TP is what I need.

Mani Fazeli: It's so funny you bring up the toilet paper example because I was just having a conversation with someone else about friction and where is it that friction is good in commerce and where is it not? And toilet paper came up as an example of, I could not imagine a product that I want to be less frictionful for purchasing because I'm not going to sit there and think about every time over and over again, what brand should I buy? How many ply? What materials should it be? Is not should it be recycled or not? This is something like a decision I make once and maybe I revisit every five to ten years. And otherwise I just want this thing to arrive at my house. And usually, as you put it really well, Brian, is you want it to come with a set of other items in a cart that are typically associated with grocery. There's a lot of repetition in that kind of purchasing. Frictionless, it's what makes it magical. Whereas there are other times when the exact opposite is true, where what you're actually looking for is that friction in the journey is about going through the discovery and the trying ideas on for size, looking for perfection.

Mani Fazeli: Maybe you're buying a gift, maybe it's a high value purchase for yourself. Maybe it's an item like a piece of furniture that's going to last a very long time. These are the kinds of places that you expect more friction to remain in discovery. Like those boots that you're talking about or the coat that I'm talking about, you think about how much meaning and also both in terms of their utility, the importance of getting it right, being warm or not being uncomfortable while you're on the slopes, but also in the style of how it looks, how it makes you feel. All these things are incredibly difficult to just outsource to some automated system because the automated system needs more than just a sense of say the business of the household and an inventory of every item that might be in it. It needs a sense of what you hold dear and what you value and how you feel. And until these systems become emotion aware, it's highly unlikely that they are completely eradicating the entire idea of manual human intervention commerce.

Brian: Frictionless is how I describe my whole toilet paper journey. That's what I want all the way through. All the way through to the end.

Phillip: Oh, my word. Let's get off that track. So, when you...there is some, you know, some friction involved, I think, Mani, you you kind of make a really good point around where some friction is good. I really like that. Instant gratification, I think does devalue what some people buy. I think that might be true. Anecdotally, I believe that to be true. We want a little bit of friction in the things that we acquire and then we might value it more. That's something we've proven out here at Future Commerce through research and through our Insights division over the last few years, beginning with a research report back in 2022 called Services, a New Storefront that we published. But yeah, I have to believe that that's something that you all are proving out every day and the improvements you make. How do you preserve elements of shopping that feel special and authentic? Not about the speed to path to purchase. It's about the balance and mediation of that purchase, right? Because I have to believe that you also care about return rates and customer service inquiries that are generated after the purchase. All those things are also part of the equation, right?

Mani Fazeli: Yeah, absolutely. I think that the premise here is right, but we could hold it a little bit differently. We've already touched on some of the topics. Let's make special what's actually worth being special. And then let's be okay with the fact that the rest of it gets streamlined. So there's a special feeling, let's use an analogy like a wedding. Maybe being on a horse and carriage as entry towards a wedding reception will feel incredibly special despite the fact that it has a lot more friction than driving up with a car. But a horse and carriage for the use of everyday transport is just friction without value. So if we take analogies like that to mind, we should be okay with the fact that the world is going to shift. Social did this. It used to be that you couldn't just see a product in an Instagram feed, fall in love with the idea, click buy. And sure, does the return rate go up a little bit? Yes. But what was the total amount of growth of orders that were never returned? Growth of orders that were beloved or cherished or really valued that you had discovery that you previously would not have discovered because you would have never gotten to that possibility through intent based search. And we should be okay with the fact that there's going to be some frictionless here that has net gain, net positive value. And I think there was this example that I heard our CEO Tobi use. I'm trying to remember which podcast it was. It might've been on Cheeky Pint, where he told a story about he was in search of pillbox, something that many people buy.

Mani Fazeli: For whatever reason, he didn't just want any pillbox. So he went to his favorite LLM and he's like, Hey, I'm looking for a pillbox and I want it to be something that's a little nicer, a little better than the standard plastic you get from pick your favorite mass retailer. And through his inquiry, he ended up being introduced to a brand he would have never found, never, had he just tried to go look for it the traditional ways. And it's this small brand called, I think it's pronounced Ikigai, Ikigai cases, and they make these like artisanal handcrafted metal pill cases. It's a Japanese company and it's exactly what it sounds like. It's just a super beautiful craft forward representation of what is otherwise a utilitarian thing. And you think about the friction removed introduced opportunity that would have never existed had we hung on to the idea that, oh, AIn this case is making something smoother or faster or leaner than would have been in the old world. If just hang on to that, and then we've already talked about it quite a bit as to yes, and then there should be friction where things need to be special And realize that it's just going to be a different mix. Let enough years pass by and this will feel natural. And if we embrace that and every merchant finds their way through that and asks the question of where do I remove friction? Where do I keep it handy? I think that that will change the way we all shop, personalization, all that sort of thing.

Brian: I love this. I think this is actually really consistent with one of my favorite philosophers, technology philosophers from the twentieth century, Norbert Wiener, who wrote a book called The Human Use of Human Beings. He's considered the father of cybernetics. Anyone who's listened to our podcast has probably heard me talk about this a little bit before, but in The Human Use of Human Beings, he has this idea, a theory that there are some things that should be human to human, some things that should be human to machine, and some things that should be machine to machine. And our job is actually to categorize what thing where things fall in those buckets. Because once if we if we if we get it right, it's actually gonna be really helpful for humanity as opposed to, like, saying everything should just be handled by machines or, oh, machines are are are killing humanity. Like, we shouldn't use them at all. Um, I think his thesis is basically there actually is an unnatural order to how things should work. And now that we're moving at the speed of light, we should understand how what that order is. And, you know, you can criticize. There's there's there's critiques of this out there available, but I think you're you're saying basically there are reasons to use tools like AI that result in good outcomes that are, like, actually good for the human, like, the end consumer. However you wanna have what what however you wanna, like, put put people into a box, like, it's actually healthy and helpful and and good. I think that, yeah, that that that is a is a great way to think about certain components of this. And then and then and then the idea that we we do wanna introduce friction in other in other areas because it actually is healthy and good for humans to have friction in those places is it's a it's a really it's a great it's a great way to put it.

Mani Fazeli: Yeah. I think every tool introduces new opportunities that the world had never imagined prior to that tool being around, but more importantly, people experimenting with that tool. You cannot foresee all the ways in which a tool makes new possibilities happen. And to go back to an analogy I already used, if we just do the horse and carriage versus car thing, the interstate system in The US allowed for all sorts of mobility, transport, commerce, trade because the car was now able to do distances that were not imaginable with horse and carriage in quite the same way. And it began to mimic a train with more flexibility because you're not on rails. Others much smarter than me have written about these concepts. But I think it very instantly resonates for most people to just be like, Yes, tools introduce entire new ways that we could imagine the world around us, but you have to embrace the tool. If you shun it, then you will never reap the rewards that it could give you.

Phillip: There's so much more for us to get into and just not enough time to get to it. I do want to look ahead a bit. So when you talked about sort of like the practical impact over hype with AI. So, by 2026, you guys have been doing these additions. You're it's release after release. You guys have incredible product cadence at Shopify. How are you helping merchants identify which of the AI capabilities actually matter? They have access to things like Sidekick. You're empowering them with things like SimGym. Are there practical onboarding tips, steps, tools, enablement? What are the ways that you're actually putting not just the tool in their hands, but actually getting them skilled up and putting that enablement into use so that they can grow their business using them?

Mani Fazeli: The first thing is we have a philosophy that the tool should be right where you work. So it should be right in your admin, doesn't have to be Shopify admin. It could be any of the administrative interfaces you use to conduct your daily business, but the tool should be right there. And yes, you could have these frontier LLMs that you use as an assistant to ask all your questions and it can scour the web, call tools, do web search, But they will never be quite as good as if you had that kind of intelligence embedded with high knowledge of your data and your interfaces. So the first thing we're doing is 2026 Winter Edition announced a bunch of stuff that has to do with Sidekiq. Sidekiq can now build admin apps for you. So you could generate an app without knowing how to code. Say I want to have a new product launch or new product line and there's a set of activities I need to complete. I want a custom app that gives me the tools to be able to do that. And I don't know how to code, but I can just ask Sidekick to build it for me. That's going be fantastic. We'd love to see what merchants do with it using natural language. Sidekick is going to let you be able to customize themes.

Mani Fazeli: So yes, all the theme editing capabilities between global settings and section settings, block settings are all there already. But now you can go to Sidekiq and say, make this button rounded or change all the fonts in my store to be this other thing. And it's going be able to do that for you through natural language and avoid you having to go through all the menu systems. You're going to be able to do flow automation with Sidekick. So if you want help getting all of your, let's say customers who've ordered more than $200 on your store to be tagged with high volume order customer, Then you can get Sidekick to help you generate that flow to automate this process so that tagging happens on the go without you necessarily knowing how to wire together all the if this then that capabilities of Shopify Flow. We've done a major upgrade to the way that we support product photo editing. So everything from inpainting to outpainting, background removal, upscaling your images and being able to help you again have this state of the art AImage generation right in the tool. Do these tools already exist? Could you go to another company and use it? Of course, but it's even more convenient when it's right in your file editor and you are able to do it inside of Shopify. So these are all things that we are doing on top of the bread and butter stuff that if you weren't using it in 2025, you really need to be using it this year, which is are you writing your alt texts for all your images? Do you have really good product descriptions? Have you done translations? Have you ensured that you have imagery that matches maybe setting or vertical or market? All these places that the tools are able to be used already that of course we're refining behind the scenes and you could take advantage of.

Mani Fazeli: And so if you take that same idea and you expand it to be general advice beyond Shopify, It's really do not look for the companies that are talking about headlines. Look at the companies in the release notes. The goods are in the release notes, not the hype, not the media. Find the tools that have this mantra of I'm shipping value that incrementally makes you better day by day. Forget all the hype and focus on substance. And with that, use the tools, embrace them and just keep experimenting. You won't know what unlocks something for you until you take a process and say, was that a problem of mine? Was it a problem a year ago? It's still a problem now. Let me try the AI tools now. They've gotten so much better.

Brian: It's incredible that you named the AI tool Sidekick because it is practically a homophone for a psychic. {laughter}

Phillip: And I've used it, and it's it's very, very good. And I also I follow a number of people on Twitter who often drop prompts of things you can do with Sidekick. And it gets better all the time. You know, reports you can run with it, you know, various like things that you can use to to pull out interesting like customer behaviors. I just think it is better than, you know, I ever thought that it could be. So yeah.

Brian: So I know we're coming up on time here. Last couple minutes looking ahead in in 2026, Mani, what what do you think is gonna be, like, the biggest benefit or, like or maybe even the quiet shift that will fundamentally change commerce?

Mani Fazeli: I like the way you phrase that, the quiet shifts, because there's plenty that everyone talks about, but the quiet shifts, maybe there's I can think of two right off the bat. So let's start with the first. Image generation, the advances of Flux two, Nana Banana Pro, Sea Dream, these other players, is going to fundamentally change and revolutionize production calendars. So don't think of it as just like a replacement for photography. Think of it as a tool that is going to completely change how you go from initial idea of what imagery you need all the way to final production. You can now test 20 hero images instead of three. You can dynamically swap backgrounds for different seasons, different worlds. The text generation has been finally unlocked. So these ideas of I may have content that used to need manual editing because AI was not good enough, that was eradicated less than a month ago. If you haven't played with it, oh my God, Nano Banana Pro can do some ridiculous things. So what it's going to require is for everything that we think about that used to be a bottleneck in production has been removed. The only bottlenecks that will remain are taste, curation and judgment. 20 So '26 is going to see, I think, a dramatic shift when it comes to the world of images and eventually video. We're not there yet, but images for now.

Mani Fazeli: And then the second theme, which should be evident to everyone, but let's say it out loud, is that structured data becomes the new SEO. Every brand is going to have to worry about whether they have clean, well structured, and well understandable schemas of their data that whether their platform like Shopify can reason about or whatever other system or tools that they need to push the data through and pipelines to get into the various forms that they needed in can reason about. So if you do lots of custom things or you stitch it together in unorthodox ways, you're going to find that it's going to be harder to take advantage of this new agentic world because the systems that can represent you, clean up your data and syndicate for you are going to struggle. But if you can find ways to conform and inform, here's how my data is structured, then you're going to be able to unlock a lot of opportunity with all the players that are helping you get your data into this new world because man, everyone knows that the world of LLMs are data hungry, but this is a whole new game because it's not about training, it's about inference. It's about being able to access it in real time. And your data needs to be pristine for that to work.

Phillip: That leads me to believe that there is a new level of, you know, not just user experience that we have to worry about, right? There is an agent experience like it's AX that we are going to have to design for. And that's where the structured data comes in. Right. This level of of, you know, structured data and data exposure on, you know, in our in our you know, in this in this data layer that I think a lot of retailers at a certain size and scale have sort of taken for granted for a really long time. I we talk to a lot of enterprises, we work with a lot of enterprises, And I think the scale of which is very daunting. Tend to not think about it in the same way that, you know, the mid market brands do, where they have a handful of hero products that, you know, carry the org. There is a definite hill to climb there. And so Mani, I think that that's a really, really strong set of predictions that I think should, I would hope start to set the tone for a lot of organizational change because I think Brian, back to your point before, that's a change management problem for the larger organizations because that sounds like, again, a huge shift because it's the new discoverability and channel shift problem that we're going to have for businesses that are going to be like, oh, we're no longer relevant in this, you know, in this world all over again.

Brian: Priorities and budget are going to have to shift.

Phillip: Correct. Yeah. And that's how that's how those orgs think of things. Wow.

Brian: What what a way to end it. What a money killed it at the end there. That was great.

Phillip: Money love. Love this. First time on the show, it shouldn't be your last. We should have you back sometime soon.

Mani Fazeli: Thank you. I hope so.

Phillip: Yeah. And love to have Shopify sharing knowledge with our audience. Thank you so much for the partnership. Where can people find more about Winter Editions?

Mani Fazeli: I recall it being shopify.com/editions will be the but we make so much noise about this. Just find us on social. If you're interested, just type into Google Shopify Editions. You will find it. The neat thing that we do about additions, we keep the history of every single addition we've ever done so you could peruse all of them. They're these incredibly beautiful sites that we create. It's really our love letter to our merchants about all the work and energy we've put behind this work for them to just showcase to them the labor of love of like why we get out of bed every single day. Shopify has come a long way. We're powering 12% of e commerce in The US these days. And that's on the back of every one of those incredible entrepreneurs who wakes up every day to run their business, uh, and trusting Shopify with it. So we're really grateful. Thank you to all the listeners who run on Shopify. For those who don't, we're happy to have you anytime. Uh, and, uh, it was a real pleasure. Thanks for the awesome questions. I really enjoyed the chat.

Phillip: Thank you. Same to you. And thank you all for checking out this episode of Future Commerce. We will we'll link to it here in the show notes below and then make it easy for you. We will be an NRF. Check us out. We have an event coming up. Go sign up for that. It's futurecommerce.com/events. And remember the future is what you make of it. Future commerce helps you shape that future. Thank you so much for checking out this episode of Future Commerce.

‍