WEBVTT 1 00:00:10.140 --> 00:00:13.000 Hello, Stripe Sessions. Everyone enjoying the show today? 2 00:00:13.380 --> 00:00:15.620 It's like the best conference in the biz, in my opinion. 3 00:00:18.400 --> 00:00:22.160 My name is Jonathan Arena. I'm a cofounder of New Generation. 4 00:00:22.760 --> 00:00:27.440 See some friends in the audience. Thanks for coming. Today, 5 00:00:28.440 --> 00:00:30.600 I'm going to tell you a story about websites, 6 00:00:32.100 --> 00:00:33.960 about selling things online, 7 00:00:34.140 --> 00:00:37.240 and where we at New Generation think this is all headed. 8 00:00:39.920 --> 00:00:42.720 But first, a short step back into time. 9 00:00:45.300 --> 00:00:49.420 One of my heroes as a designer is this guy named Bill Atkinson; 10 00:00:49.580 --> 00:00:50.580 early Apple employee, 11 00:00:51.220 --> 00:00:55.920 and credited with inventing HyperCard, if you remember that. 12 00:00:58.120 --> 00:01:02.880 This was really one of the first primitives that was a card format in graphical 13 00:01:02.940 --> 00:01:06.000 user interface that combined images, buttons, links. 14 00:01:06.880 --> 00:01:10.500 Kind of like a product card on an ecommerce website. 15 00:01:13.960 --> 00:01:18.340 No doubt that laid the foundation for one of the earliest screenshots I could 16 00:01:18.400 --> 00:01:22.920 find of an ecommerce website. I think Netmarket made the first one in '94. 17 00:01:24.440 --> 00:01:25.320 Earliest screenshot, 18 00:01:25.380 --> 00:01:29.320 even the Wayback Machine and Internet Archive goes back to 2002, but 19 00:01:31.460 --> 00:01:35.020 it doesn't seem like much has changed. It's pixelated, 20 00:01:36.260 --> 00:01:38.480 but there's still links, a familiar sidebar. 21 00:01:39.580 --> 00:01:41.260 Time is a little bit of a flat circle, 22 00:01:41.500 --> 00:01:43.440 which I think you'll agree by the end of the talk. 23 00:01:44.480 --> 00:01:47.000 But it's fun to see where we've been. 24 00:01:48.100 --> 00:01:52.340 Early ecommerce websites were largely designed actually for Google crawler and 25 00:01:52.380 --> 00:01:55.120 Googlebot, not for humans. 26 00:01:57.120 --> 00:02:01.640 Does anybody know what the first product sold on the internet is? Shout it out. 27 00:02:05.620 --> 00:02:10.160 It was actually Sting's album. It'd be fun to see if Sting knows that. 28 00:02:12.340 --> 00:02:14.960 All that is to say, leading us to present day, 29 00:02:15.120 --> 00:02:17.820 what should a commerce website be in the age of agents? 30 00:02:19.240 --> 00:02:21.580 That's what we're going to explore over the next few minutes. 31 00:02:25.440 --> 00:02:26.720 Let's go shopping to find out. 32 00:02:26.800 --> 00:02:29.580 I'm going to take you on a little bit of a journey that I went through in 33 00:02:29.600 --> 00:02:30.560 preparing for this talk. 34 00:02:32.400 --> 00:02:36.440 And I'm going to go to a store that we at New Generation love. 35 00:02:36.520 --> 00:02:39.360 It's called Quince. Now, 36 00:02:39.580 --> 00:02:42.500 I entered a search on Quince's search bar. 37 00:02:42.640 --> 00:02:47.520 It was a very basic natural language search. I have a big stage talk, 38 00:02:48.660 --> 00:02:50.460 and I need a new top. What would be good? 39 00:02:53.200 --> 00:02:54.720 What do you think Quince shows me? 40 00:02:58.000 --> 00:03:00.880 Skirts and a diaper pail. 41 00:03:02.080 --> 00:03:05.140 I don't know if they're trying to tell me that this is like a garbage talk. 42 00:03:06.200 --> 00:03:10.100 You can be the judge. Unfortunately, 43 00:03:10.160 --> 00:03:13.900 this problem is not unique to Quince. Quince is actually an exceptional website. 44 00:03:14.200 --> 00:03:17.400 And by the way, an exceptional company. You should go check them out. 45 00:03:18.900 --> 00:03:22.960 And the problem is deeper than a cosmetic or surface level search doesn't work. 46 00:03:23.240 --> 00:03:25.400 It actually goes all the way to the infrastructure level. 47 00:03:28.380 --> 00:03:29.960 What about shopping for a mortgage? 48 00:03:31.480 --> 00:03:33.500 The beautiful Chase mortgage website. 49 00:03:35.520 --> 00:03:38.840 I see this H-roll carousel with features and perks. 50 00:03:39.040 --> 00:03:42.860 Who is going to click on that? Sorry, Chase. Nobody. 51 00:03:45.240 --> 00:03:49.860 And then down below the fold, way below the fold is, "Why choose Chase." 52 00:03:53.140 --> 00:03:55.480 Surely we can do better than this. 53 00:04:01.020 --> 00:04:01.880 What's happening here? 54 00:04:02.000 --> 00:04:06.140 It's because all these websites are built for the average customer, 55 00:04:06.200 --> 00:04:09.660 which everybody knows doesn't exist. And as a consequence, 56 00:04:09.720 --> 00:04:14.480 it's optimized for none of them. The website can't understand your intention. 57 00:04:16.900 --> 00:04:20.880 It can't easily get information to you. You have to go search for it. 58 00:04:23.160 --> 00:04:25.700 And as we've seen, 59 00:04:26.740 --> 00:04:30.540 the display of information doesn't leave the best first impression. 60 00:04:34.180 --> 00:04:38.860 So I have a provocation for you. What if the website itself were intelligent? 61 00:04:40.040 --> 00:04:41.000 What might that be like? 62 00:04:43.800 --> 00:04:48.800 Maybe it's a website that allows you to talk naturally 63 00:04:49.700 --> 00:04:52.200 with your voice or with a natural language query. 64 00:04:53.660 --> 00:04:57.600 But let's not forget that shopping is visual, right? People have eyes, 65 00:04:57.680 --> 00:05:01.640 they have judgment and taste. They want to see what they're shopping for. 66 00:05:03.000 --> 00:05:04.880 What if an interface could be built for you, 67 00:05:06.420 --> 00:05:08.120 based on what you're trying to accomplish? 68 00:05:11.640 --> 00:05:16.320 So let's go back to Quince and imagine a world in which that website had 69 00:05:17.300 --> 00:05:18.740 60 more IQ points. 70 00:05:21.680 --> 00:05:26.300 The first thing you'd probably see is a landing page that 71 00:05:26.600 --> 00:05:30.120 is aware of what's happening. 72 00:05:30.280 --> 00:05:31.780 Maybe it's trends on the internet. 73 00:05:33.740 --> 00:05:36.340 Maybe it knows that Mother's Day is coming up, 74 00:05:36.680 --> 00:05:38.300 and you might be looking for a great gift. 75 00:05:39.640 --> 00:05:42.720 Probably an intelligent website would do this automatically. 76 00:05:46.200 --> 00:05:51.080 So let's try that same query on the version of Quince that went to 77 00:05:51.140 --> 00:05:55.980 grad school. And this was my experience. 78 00:05:57.560 --> 00:06:01.120 It's qualifying me. It's asking which gender I'm shopping for. 79 00:06:05.200 --> 00:06:08.300 And in real time, it's generating a bespoke webpage for me, 80 00:06:09.260 --> 00:06:12.100 tailored to the intent I expressed. 81 00:06:13.340 --> 00:06:15.500 It could ask intelligent follow-up questions, 82 00:06:16.720 --> 00:06:19.300 generate and render dynamic filters. 83 00:06:21.160 --> 00:06:24.960 I like cashmere. That's a soft thing, and Quince is, you know, good price. 84 00:06:26.140 --> 00:06:29.960 And I get made fun of all the time by my wife. My wardrobe is monochromatic, 85 00:06:30.040 --> 00:06:31.240 so I like black. 86 00:06:34.140 --> 00:06:38.240 This entire page is actually AI generated, 87 00:06:40.440 --> 00:06:45.280 and it's contextually relevant to my search. It can help me qualify, 88 00:06:45.280 --> 00:06:49.530 "Is this product right for what I asked for?" It could even ask me smart 89 00:06:49.590 --> 00:06:52.630 follow-up questions. Like is this going to be too warm onstage? 90 00:06:53.210 --> 00:06:57.370 It says probably yes. Thankfully, I've been here before. I know they have AC, 91 00:06:58.110 --> 00:07:02.650 so I bought it anyway. Yes, we dogfood our own products. 92 00:07:08.110 --> 00:07:11.450 Shopping starts with intent, not a keyword. 93 00:07:13.290 --> 00:07:17.630 We like to say that now the internet is denominated in natural language. 94 00:07:17.690 --> 00:07:21.990 This is the primary interaction model that will be how we 95 00:07:22.930 --> 00:07:24.750 engage with websites and the internet at large. 96 00:07:30.280 --> 00:07:34.680 So "commerce agents" may or may not be a new term for you. 97 00:07:36.600 --> 00:07:41.000 This mythical creature of an AI agent-you've heard of coding agents, 98 00:07:41.880 --> 00:07:43.560 maybe customer service agents. 99 00:07:46.160 --> 00:07:47.200 Here's a definition. 100 00:07:47.800 --> 00:07:51.880 The purpose of a commerce agent is to help retailers engage 101 00:07:52.920 --> 00:07:55.120 in selling things online with this new channel. 102 00:07:55.520 --> 00:08:00.440 We think AI is a brand new channel, and it can do things. It can save time. 103 00:08:01.800 --> 00:08:03.870 It can understand and engage with customers. 104 00:08:08.680 --> 00:08:10.640 Interestingly, and I think this is pretty cool, 105 00:08:10.840 --> 00:08:15.200 an agent can also have its own address. There's two reasons for this. One, 106 00:08:15.360 --> 00:08:19.680 obviously, if you're a retailer and you're experimenting with new technology, 107 00:08:19.920 --> 00:08:24.720 this provides a really wonderful way for you to partition what a commerce agent 108 00:08:24.760 --> 00:08:29.160 can do and compare it against your existing website in an A/B test. Number two, 109 00:08:30.080 --> 00:08:34.120 it acts as a dedicated endpoint to receive agentic traffic. 110 00:08:34.720 --> 00:08:37.120 Maybe this is where you expose your MCP tools 111 00:08:39.600 --> 00:08:41.560 or other tools that I'll show you in a moment. 112 00:08:44.440 --> 00:08:46.120 What else can agents do? 113 00:08:46.840 --> 00:08:49.960 Let's quickly revisit Quince and dive a little bit deeper. 114 00:08:52.280 --> 00:08:56.720 So this is a different search I did around-I'm looking for some jewelry for an 115 00:08:56.760 --> 00:09:00.160 anniversary gift for my wife. Quince is great. 116 00:09:00.240 --> 00:09:04.960 They have many different verticals on their website. So again, bespoke webpage, 117 00:09:05.200 --> 00:09:08.920 this collection was dynamically assembled for me, and I'm not sure. 118 00:09:09.120 --> 00:09:12.840 I'm evaluating natural diamonds versus lab grown diamonds. 119 00:09:14.360 --> 00:09:18.240 What if it could render a dynamic comparison table perfectly built for the 120 00:09:18.240 --> 00:09:21.440 vector of comparison I'm interested in? I like that one. 121 00:09:21.480 --> 00:09:24.640 Let's add it to the cart. Oh, interesting. 122 00:09:26.690 --> 00:09:29.730 So a smart website would know that I've added something to my cart and 123 00:09:29.850 --> 00:09:34.010 automatically generate a perfect bundle. This is what I mean. 124 00:09:34.070 --> 00:09:35.130 These are answer types, 125 00:09:35.490 --> 00:09:39.010 and this is UI that the agent is understanding because it knows 126 00:09:40.450 --> 00:09:44.270 about my customer journey. It knows where I am. It knows what I've said. 127 00:09:46.410 --> 00:09:50.750 Are we giving the website a chatbot? Definitely not. 128 00:09:50.810 --> 00:09:53.410 This is at the substrate of the website itself. 129 00:09:56.230 --> 00:09:58.250 Not a little widget in the corner with just text. 130 00:09:58.330 --> 00:10:02.550 This is a fully immersive visual experience that a brand can now control. 131 00:10:11.230 --> 00:10:13.310 Let's take a dive under the hood because I'm sure you're wondering, 132 00:10:13.410 --> 00:10:16.830 how is this being built? And I'm going to show you. 133 00:10:18.750 --> 00:10:19.910 It all starts with knowledge. 134 00:10:19.970 --> 00:10:24.710 This first block is what we call the semantic knowledge 135 00:10:24.770 --> 00:10:25.270 layer. 136 00:10:25.270 --> 00:10:29.710 And the easy way to think about this is we just give the commerce agent all the 137 00:10:29.750 --> 00:10:33.070 information it needs to know about how to be an exceptional salesperson on your 138 00:10:33.110 --> 00:10:38.070 website. That means it knows about your catalog, your merchant rules, 139 00:10:38.510 --> 00:10:39.343 your policies. 140 00:10:40.710 --> 00:10:43.930 It can enrich your catalog with internet data if that's relevant. 141 00:10:46.010 --> 00:10:48.410 It can access a library 142 00:10:50.290 --> 00:10:51.123 of expertise. 143 00:10:54.880 --> 00:10:59.310 We can then feed that knowledge into what we call the "commerce runtime." And 144 00:10:59.330 --> 00:11:03.270 this is a stateful environment that the agent lives in during an 145 00:11:05.000 --> 00:11:06.490 agentic commerce session. 146 00:11:08.170 --> 00:11:11.690 So in addition to having access to all this information and this knowledge, 147 00:11:13.010 --> 00:11:16.810 the agent can use tools. Some of these tools, as you've already experienced, 148 00:11:17.010 --> 00:11:19.950 are a natural language search API. 149 00:11:22.310 --> 00:11:25.410 State and memory. If you tell the agent something, it should remember. 150 00:11:26.790 --> 00:11:29.190 Think about that for logged in experiences in the future. 151 00:11:29.510 --> 00:11:34.270 You're going to have this preference dataset that you allow to be stored with 152 00:11:34.290 --> 00:11:38.710 the merchant. Composable UI, 153 00:11:39.710 --> 00:11:44.530 the ability for an agent to generate on-demand interfaces and a virtual cart, 154 00:11:45.250 --> 00:11:49.490 which we'll get to in a moment. Lastly, 155 00:11:49.670 --> 00:11:53.210 the outputs. So we've seen some of these outputs already. 156 00:11:54.830 --> 00:11:59.210 The AI storefront hosted on maybe an AI subdomain on a brand's website. Really, 157 00:11:59.270 --> 00:12:00.570 they could host that anywhere. 158 00:12:02.190 --> 00:12:04.870 But we know that many different outputs are racing towards us. 159 00:12:05.130 --> 00:12:09.830 Some of us have probably heard and seen ChatGPT Apps 160 00:12:09.970 --> 00:12:13.270 SDK or Claude MCP Apps. 161 00:12:13.370 --> 00:12:16.490 Imagine if you are a consumer shopping in Claude, 162 00:12:17.530 --> 00:12:22.250 wouldn't it be great if Quince's intelligent generative website could 163 00:12:22.330 --> 00:12:23.890 render right there before the customer? 164 00:12:24.870 --> 00:12:29.390 This is the type of consideration that I'm sharing with the audience today. 165 00:12:33.990 --> 00:12:36.990 And it's really this runtime that it's, I think, the most interesting part. 166 00:12:38.090 --> 00:12:42.690 And diving even deeper yet still, what's most fascinating to me-again, 167 00:12:43.250 --> 00:12:47.890 having spent my career designing software and interfaces-is composable 168 00:12:47.990 --> 00:12:51.310 UI. And we talk to hundreds of merchants around the world, 169 00:12:51.370 --> 00:12:54.850 and one of the things we keep hearing again and again is, 170 00:12:55.730 --> 00:13:00.190 how do you make sure that generative interfaces are on brand 171 00:13:01.010 --> 00:13:01.843 reliably? 172 00:13:02.690 --> 00:13:06.310 It would be kind of disappointing if 173 00:13:07.350 --> 00:13:10.290 Quince's agent rendered UI that looked like Target. 174 00:13:12.070 --> 00:13:13.850 That's a no-go. How can we prevent that? 175 00:13:15.430 --> 00:13:20.170 What we do is we send an agent to understand and extract the design system of 176 00:13:20.210 --> 00:13:24.650 any brand. They could also upload their design system if they have that, 177 00:13:24.710 --> 00:13:25.970 if it's sophisticated company. 178 00:13:27.330 --> 00:13:31.750 But this is everything from the image style, the typography, 179 00:13:31.850 --> 00:13:34.750 the white space, the tone of voice, 180 00:13:36.490 --> 00:13:38.730 the buttons and the components, the motion. 181 00:13:40.630 --> 00:13:44.450 All of this will be programmatic. We then normalize that 182 00:13:46.490 --> 00:13:47.930 and we give it as a tool to the agent. 183 00:13:48.210 --> 00:13:51.130 One metaphor that's helpful to think about how this works is, 184 00:13:52.130 --> 00:13:57.030 we say to brands a lot, "The agent has a bag of Lego blocks. 185 00:13:57.630 --> 00:14:00.010 These are your components. These are your design systems. 186 00:14:00.130 --> 00:14:01.410 This is very partitioned, 187 00:14:02.270 --> 00:14:06.610 and these are the only pieces that the agent can use to assemble interfaces. 188 00:14:08.090 --> 00:14:12.030 That's how we make sure it's accurate and consistent and on 189 00:14:15.710 --> 00:14:17.530 brand." And the purpose of all this, 190 00:14:17.830 --> 00:14:22.530 the big idea is we finally maybe have an opportunity to collapse 191 00:14:23.710 --> 00:14:27.930 and close the gap between human intent and a machine's ability to take the right 192 00:14:28.030 --> 00:14:32.770 action. Everybody has experienced this gap, right? 193 00:14:32.830 --> 00:14:37.350 If you've ever been frustrated on a website, you've experienced this gap. 194 00:14:39.490 --> 00:14:41.190 Commerce agents can help close this gap. 195 00:14:42.950 --> 00:14:46.810 The result is that humans can express themselves freely without constraint, 196 00:14:47.370 --> 00:14:48.203 without concern, 197 00:14:48.650 --> 00:14:53.630 and have delightful experiences where merchants and retailers can actually 198 00:14:53.690 --> 00:14:57.350 have, be the beneficiaries of a lot of new value. 199 00:14:59.830 --> 00:15:01.850 So one agent, 200 00:15:03.270 --> 00:15:05.990 five powerful new jobs the old website could not do. 201 00:15:08.290 --> 00:15:10.590 Generate bespoke answers for every customer at scale. 202 00:15:14.130 --> 00:15:16.210 Format catalogs, PDP data, 203 00:15:16.830 --> 00:15:20.390 policy information for agents and agentic systems. 204 00:15:23.110 --> 00:15:25.330 Provide incredibly rich first-party data. 205 00:15:27.410 --> 00:15:31.990 Use case here is now customers are trying to discover, evaluate, 206 00:15:32.190 --> 00:15:34.690 and buy your products in natural language. 207 00:15:34.930 --> 00:15:39.790 That's information you've never had available to you. Obviously, 208 00:15:39.850 --> 00:15:41.330 let's not forget the business outcomes. 209 00:15:42.630 --> 00:15:47.070 Commerce agents know how to sell your products. Average order value goes up, 210 00:15:47.530 --> 00:15:51.270 conversion goes up. And these are learning systems. 211 00:15:52.030 --> 00:15:55.410 They learn what works, what doesn't. They self-reinforce, they self-learn. 212 00:15:57.970 --> 00:16:02.650 And perhaps where it gets the most interesting is a commerce agent 213 00:16:03.530 --> 00:16:07.410 is actually, in our opinion, the future interface for consumer agents. 214 00:16:12.190 --> 00:16:16.170 So what happens when the visitor to your website is an agent? 215 00:16:18.830 --> 00:16:22.550 We know these are now first-class citizens of the web. 216 00:16:24.650 --> 00:16:26.050 I think recently last year, 217 00:16:26.470 --> 00:16:31.110 we crossed a 50% threshold where more than half of all internet traffic 218 00:16:32.650 --> 00:16:33.483 is nonhuman. 219 00:16:37.590 --> 00:16:41.850 And it seems like every week there's a new type of 220 00:16:42.610 --> 00:16:47.130 agentic system that comes out. It's really interesting. 221 00:16:47.210 --> 00:16:51.510 So we just saw what we like to call "AI shoppers." That's you or me using 222 00:16:51.550 --> 00:16:55.750 natural language. We're now all AI shoppers. Of course, 223 00:16:55.810 --> 00:17:00.770 we have LLMs that are using crawlers and different methods to extract internet 224 00:17:00.810 --> 00:17:02.530 data. They're trained on some of it, of course. 225 00:17:04.050 --> 00:17:06.690 Tool use agents like Claude or OpenClaw 226 00:17:08.710 --> 00:17:11.130 and agentic browsers and browser agents. 227 00:17:12.890 --> 00:17:14.950 Let's take a look at one of these in detail. 228 00:17:16.170 --> 00:17:21.110 What happens when browser agents interact with websites? So about 229 00:17:23.150 --> 00:17:23.983 eight months ago, 230 00:17:25.870 --> 00:17:29.550 we sent an agent to Samsung's website, 231 00:17:31.270 --> 00:17:34.890 and we also sent an agent to Samsung's website that had the very early 232 00:17:34.970 --> 00:17:37.910 beginnings of a commerce agent. And I want to show you what happened. 233 00:17:40.230 --> 00:17:41.790 Now it's a little fast. You got to pay attention. 234 00:17:41.850 --> 00:17:45.050 On the left is regular Samsung, and on the right is the smart Samsung. 235 00:17:46.310 --> 00:17:48.190 Same query, same product catalog. 236 00:17:56.730 --> 00:17:58.190 And the smart website's already finished. 237 00:17:58.450 --> 00:18:01.690 And if you have to take my word for it, it's actually a really good answer. 238 00:18:02.290 --> 00:18:04.850 High quality. And even 8 months ago, 239 00:18:04.920 --> 00:18:09.630 this website was 10 times faster than Samsung.com at answering the agent's 240 00:18:09.670 --> 00:18:13.490 question. Why? What's happening? Well, 241 00:18:14.950 --> 00:18:19.370 agents don't like human UI. They get stuck on pop-up windows. 242 00:18:19.770 --> 00:18:24.430 They have to go from PDP to PDP, extract all the data, reason over it, 243 00:18:24.770 --> 00:18:29.510 look for reviews. They get stuck on JavaScript. There's all sorts of problems. 244 00:18:30.170 --> 00:18:32.010 Surely we thought there must be a better way. 245 00:18:33.950 --> 00:18:37.610 And so over the last eight months, we've been building that better way. 246 00:18:40.150 --> 00:18:40.990 And as I mentioned, 247 00:18:41.270 --> 00:18:45.370 the pace of change here is so incredibly fast, 248 00:18:46.410 --> 00:18:51.210 that we actually think a wonderful use case for agents is simply abstracting all 249 00:18:51.230 --> 00:18:55.250 that complexity and helping you keep up and experiment with the efficient 250 00:18:55.270 --> 00:18:56.103 frontier. 251 00:18:59.090 --> 00:19:01.950 So I'll show you where we are today. And before I play this, 252 00:19:02.070 --> 00:19:06.170 I want to set up what you're about to see. This is Comet, 253 00:19:06.550 --> 00:19:08.750 Perplexity's agentic browser, 254 00:19:09.770 --> 00:19:14.730 and this is a demonstration of that interacting directly with 255 00:19:14.770 --> 00:19:16.550 Quince's commerce agent that we built. 256 00:19:18.770 --> 00:19:21.370 The last thing that's very interesting about browsers here, 257 00:19:21.650 --> 00:19:25.730 is that they know who you are because they have your history and you can opt in 258 00:19:25.770 --> 00:19:28.770 to share preferences. So automatically, 259 00:19:28.810 --> 00:19:33.230 the browser can tell the conversation of Quince: "This is Jonathan. 260 00:19:33.590 --> 00:19:36.530 This is where he lives. These are his preferences, his shoe size, 261 00:19:37.050 --> 00:19:41.470 you name it." And what happens is pretty remarkable. 262 00:19:43.810 --> 00:19:46.670 And we asked a pretty tough question to test it out. 263 00:19:47.230 --> 00:19:50.910 My wife and I are going to Japan. We want capsule collections for both of us. 264 00:19:51.010 --> 00:19:52.990 We have a $1,000 budget. Go figure it out. 265 00:19:55.530 --> 00:19:58.730 What's happening here is, the agents are working together 266 00:20:00.650 --> 00:20:02.350 to try to get me the best answer possible. 267 00:20:04.850 --> 00:20:07.630 And this is going to be groundbreaking for merchants, 268 00:20:09.270 --> 00:20:14.190 because this means that they can still maintain merchandising control and have 269 00:20:15.210 --> 00:20:18.260 an ability to share their expertise with the agent. 270 00:20:20.970 --> 00:20:24.900 It did its job. It printed me out a whole collection. It was under budget. 271 00:20:25.420 --> 00:20:27.280 It split the collection between my wife and I. 272 00:20:27.470 --> 00:20:30.520 I was actually pretty impressed with this. To be clear, 273 00:20:30.900 --> 00:20:34.970 this is completely impossible today with regular websites, 274 00:20:36.990 --> 00:20:40.940 or it would take way too long and you'd leave. 275 00:20:43.780 --> 00:20:45.400 And of course, everyone's wondering about checkout. 276 00:20:45.540 --> 00:20:47.160 This is Stripe conference after all, right? 277 00:20:47.710 --> 00:20:52.640 So what if I wanted to buy all that stuff after some back and forth, say, "Yeah, 278 00:20:52.700 --> 00:20:55.120 go pull the trigger." Well, 279 00:20:55.400 --> 00:20:57.540 our point of view here is a little bit different and unique. 280 00:20:58.900 --> 00:21:03.300 In contrast to agentic checkout in a closed ecosystem, 281 00:21:03.760 --> 00:21:04.760 like within ChatGPT, 282 00:21:06.420 --> 00:21:10.900 what we're describing here is how any agent with a 283 00:21:10.960 --> 00:21:15.660 tokenized credit card could go to the open web to any merchant 284 00:21:15.680 --> 00:21:18.020 that has a commerce agent and check out. 285 00:21:20.520 --> 00:21:24.100 What you just saw with Comet are these pieces. 286 00:21:26.710 --> 00:21:28.620 And the simple way that this works is, 287 00:21:29.490 --> 00:21:32.990 the commerce agent can now take the buy intent from the customer, 288 00:21:33.580 --> 00:21:36.460 given to it from the agent, and the credit card, 289 00:21:38.460 --> 00:21:41.220 and it can use those tools we talked about before-a virtual cart. 290 00:21:42.000 --> 00:21:43.320 It then goes and looks up, 291 00:21:43.860 --> 00:21:47.020 is this product in stock? What's the final shipping cost? 292 00:21:47.480 --> 00:21:48.980 All those ticky-tacky details. 293 00:21:50.240 --> 00:21:55.220 And it's going to take that information and pass it through a checkout session, 294 00:21:55.520 --> 00:21:59.780 maybe similar to what Stripe is doing with Agentic Commerce Protocol, 295 00:22:01.140 --> 00:22:04.660 but it's really this merchant-side infrastructure that allows any merchant to be 296 00:22:04.740 --> 00:22:05.700 ready for this at scale. 297 00:22:09.560 --> 00:22:13.360 So how prepared is the average website to receive these new first-class 298 00:22:13.420 --> 00:22:16.100 visitors? These are three questions that we ask. 299 00:22:19.800 --> 00:22:21.680 And we were asking these questions so much 300 00:22:23.440 --> 00:22:25.940 that we built something called the "Agent Commerce Score." 301 00:22:27.740 --> 00:22:31.700 And this is probably the most advanced evaluation of a website's preparedness to 302 00:22:31.760 --> 00:22:33.940 receive agents and answer their questions. 303 00:22:35.640 --> 00:22:36.860 I'll show you how it works. 304 00:22:38.920 --> 00:22:43.300 We send 100 agents to Quince, and we ask 100 different questions, 305 00:22:43.380 --> 00:22:47.260 and these questions are automatically generated to be very relevant for what 306 00:22:47.320 --> 00:22:49.120 Quince sells and who they are. 307 00:22:51.340 --> 00:22:53.720 We ask easy questions, we call them table stakes questions, 308 00:22:53.780 --> 00:22:55.560 and we ask data-intensive questions like, 309 00:22:55.600 --> 00:22:58.100 compare reviews across these variables. 310 00:22:59.100 --> 00:23:02.220 You can go all the way down to a single question and take a look at how the 311 00:23:02.280 --> 00:23:06.560 agent did. Which products was it served? Are these the right products? 312 00:23:07.820 --> 00:23:10.980 You can even click on a replay of the agent experience, 313 00:23:12.420 --> 00:23:15.460 and it can tell you how it thinks it did. 314 00:23:17.220 --> 00:23:21.420 This provides incredible data again and very clear next steps for any brand to 315 00:23:21.480 --> 00:23:22.920 evaluate, "Here's where we're doing well, 316 00:23:23.360 --> 00:23:24.840 and here's where we need to take next steps." 317 00:23:27.880 --> 00:23:30.880 And we found a really interesting one in looking at this 318 00:23:30.940 --> 00:23:34.560 report-a softball question. 319 00:23:34.620 --> 00:23:39.420 I think it's a pretty easy question like, "Can I self-launder a suit?" No, 320 00:23:39.580 --> 00:23:40.540 you should probably dry-clean it, 321 00:23:40.860 --> 00:23:43.960 but we wanted to test if the website could answer a basic question like that. 322 00:23:44.740 --> 00:23:48.020 It didn't do well. It says, "Poor," so we wanted to know why. 323 00:23:51.880 --> 00:23:52.980 Kind of a disappointing answer. 324 00:23:53.040 --> 00:23:56.480 It got stuck on a pop-up window that it could not dismiss. 325 00:23:58.260 --> 00:24:02.020 Now, I'm sharing this really to describe how brittle 326 00:24:03.800 --> 00:24:08.520 the standard website is. And of note, 327 00:24:09.820 --> 00:24:14.700 perhaps this was the modal that the agent saw and it says, 328 00:24:14.700 --> 00:24:18.880 "Hold on." Maybe it just got stuck here and obeyed instructions. 329 00:24:19.100 --> 00:24:19.933 We don't know. 330 00:24:26.460 --> 00:24:28.820 So when you're asking a question to a website, 331 00:24:29.200 --> 00:24:31.340 and the website can answer back effectively, 332 00:24:31.900 --> 00:24:34.800 you have a much better chance of converting that customer, 333 00:24:34.860 --> 00:24:38.060 be it an agent customer or a human customer. If you can't, 334 00:24:39.520 --> 00:24:40.353 they're going to bounce. 335 00:24:42.760 --> 00:24:44.640 This is what we call the bounce rate of the future, 336 00:24:45.580 --> 00:24:49.160 and we think every owner of every website needs to be paying attention to this, 337 00:24:50.740 --> 00:24:54.520 because it's not like demand goes away. 338 00:24:55.500 --> 00:24:57.540 It just goes somewhere else. It goes to your competitor. 339 00:24:58.980 --> 00:25:01.080 You have a high-intent customer saying what they want. 340 00:25:02.800 --> 00:25:05.480 If you're not ready to give a good answer, 341 00:25:05.680 --> 00:25:06.980 they're going to find it somewhere else. 342 00:25:09.280 --> 00:25:14.220 So I will leave you with a few takeaways that you can share back with 343 00:25:14.240 --> 00:25:19.060 your teams. The website becomes programmable. 344 00:25:20.120 --> 00:25:24.120 We are moving out of the era of static one-size-fits-all websites and into 345 00:25:24.160 --> 00:25:25.160 intelligence systems. 346 00:25:28.500 --> 00:25:33.400 The storefront becomes portable. Yes, everyone will still need a website. 347 00:25:33.780 --> 00:25:37.840 We believe that you will still have traffic to your website. But in addition, 348 00:25:38.840 --> 00:25:42.260 you probably need your website to go out and meet customers where they are. 349 00:25:43.800 --> 00:25:48.160 Maybe thousands, if not millions of AI surfaces where customers are shopping. 350 00:25:51.400 --> 00:25:55.200 Every brand gets their own commerce agent, and this allows teams 351 00:25:57.060 --> 00:25:59.660 to turn into orchestrators and directors, 352 00:26:01.240 --> 00:26:03.820 saving CapEx and overhead and time. 353 00:26:08.400 --> 00:26:10.180 So here's your checklist and your homework. 354 00:26:12.940 --> 00:26:16.380 I'm going to ground this again. Can you answer natural language questions? 355 00:26:18.240 --> 00:26:19.680 Can you receive these new customers? 356 00:26:24.600 --> 00:26:27.340 Are you aware of the intention of your customers and where they are in their 357 00:26:27.380 --> 00:26:31.880 shopping journey? And of course, 358 00:26:34.740 --> 00:26:36.740 can it dynamically bend and scale? 359 00:26:40.320 --> 00:26:42.600 If you're curious, you can go get a score. 360 00:26:43.440 --> 00:26:47.080 It takes about 15 minutes for the agents to do their job. 361 00:26:48.320 --> 00:26:49.900 We'd love to hear from you if you do that. 362 00:26:51.880 --> 00:26:53.980 And if you think that this is interesting, 363 00:26:54.200 --> 00:26:57.020 and you're struggling to answer these questions yourself, 364 00:27:00.380 --> 00:27:04.560 we think about this every day. Thank you so much.