WEBVTT - Solving the Trust Problem in AI Shopping 0:00:15.396 --> 0:00:15.826 Pushkin. 0:00:20.316 --> 0:00:23.476 Hello, hello, Malcolm Gladwell here. Over the past several months 0:00:23.616 --> 0:00:27.376 across revisionist history and business history, we've been exploring the 0:00:27.496 --> 0:00:31.476 idea of trust, where it comes from, the moment it's earned, 0:00:32.016 --> 0:00:35.396 and where we decide to place it. This episode is 0:00:35.436 --> 0:00:40.226 sponsored by PayPal, trusted by consumers, built for all business. 0:00:40.866 --> 0:00:44.066 For over 25 years, they've been at the forefront of commerce, 0:00:44.596 --> 0:00:48.776 helping merchants of all sizes grow. So wherever commerce moves next, 0:00:49.136 --> 0:00:52.676 you can rest assured PayPal has your back. And now 0:00:52.836 --> 0:00:55.276 we want to leave you with one final discussion on 0:00:55.316 --> 0:00:59.806 this special episode of What's Your Problem? Today, Jacob Goldstein 0:01:00.106 --> 0:01:04.166 hosts a conversation about how consumers and companies are trying 0:01:04.206 --> 0:01:07.576 to figure out trust in the era of agentic AI. 0:01:08.296 --> 0:01:08.796 Let's listen. 0:01:20.556 --> 0:01:22.876 I'm Jacob Goldstein, and this is What's Your Problem? My 0:01:22.916 --> 0:01:26.746 guest today is Kartik Hosanagar. Kartik is a professor of 0:01:26.946 --> 0:01:30.306 operations and marketing at the Wharton School at the University 0:01:30.346 --> 0:01:33.646 of Pennsylvania. And I wanted to talk with him because 0:01:34.146 --> 0:01:37.146 we are at this moment when the way people use 0:01:37.226 --> 0:01:40.476 AI is shifting. We're starting to go beyond just having 0:01:40.616 --> 0:01:44.946 AI answer questions and toward having AI actually do stuff 0:01:45.086 --> 0:01:49.146 in the world for us, including having AI agents go 0:01:49.286 --> 0:01:53.306 out onto the internet and buy stuff. Karthik has been 0:01:53.366 --> 0:01:55.926 studying this shift. He's been trying to figure out how 0:01:55.966 --> 0:01:59.706 both customers and merchants can solve the problems that arise 0:01:59.826 --> 0:02:03.246 in the era of shopping with agentic AI. What will 0:02:03.286 --> 0:02:05.906 it take for customers to trust AI agents to make 0:02:05.986 --> 0:02:09.106 purchases on their behalf? And what do merchants need to 0:02:09.166 --> 0:02:11.426 do to prepare for a world where some of their 0:02:11.506 --> 0:02:15.506 customers are AI agents? In the first part of the conversation, 0:02:15.686 --> 0:02:18.346 Karthik and I talked about what this all means from 0:02:18.366 --> 0:02:21.106 the consumer's point of view. In the second part, we 0:02:21.146 --> 0:02:22.946 talked about what it means for merchants. 0:02:23.426 --> 0:02:28.276 You know, I will often begin my shopping research, having 0:02:28.436 --> 0:02:31.016 LLMs do the research for me. I'll send them on 0:02:31.196 --> 0:02:35.676 a deep research task and have them do all of that. 0:02:35.716 --> 0:02:39.016 I'll have them synthesize it for me. In some cases, 0:02:39.096 --> 0:02:41.736 I might be patient enough to read that long report. 0:02:41.776 --> 0:02:45.076 Sometimes I'll say, convert this into a table and just 0:02:45.136 --> 0:02:47.776 give me like, you know, what I need to know. 0:02:48.076 --> 0:02:50.996 This is pretty hardcore shopping. Like what kind of item 0:02:51.136 --> 0:02:54.816 are you having deep research reports written for? Tell me 0:02:54.836 --> 0:02:55.576 it's a t-shirt. 0:02:58.196 --> 0:03:00.596 Yeah, well, you know, sometimes it's like, you know, I'm 0:03:01.316 --> 0:03:04.536 looking to buy a car right now and trying to 0:03:04.556 --> 0:03:10.936 figure out what new electric vehicles are around. But honestly, 0:03:10.976 --> 0:03:14.836 I don't even know the research about whether electric vehicles 0:03:15.556 --> 0:03:19.456 are as environmentally friendly as originally touted to be because 0:03:19.476 --> 0:03:22.836 there's all this criticism that, hey, there's all this rare 0:03:22.856 --> 0:03:25.596 earth mining and the environmental damage of that is very high. 0:03:25.636 --> 0:03:28.676 So I'm not even saying which car should I buy. 0:03:28.716 --> 0:03:31.456 I start with, tell me The pros and cons of 0:03:31.496 --> 0:03:34.316 electric vehicles. In what sense is it good for the 0:03:34.376 --> 0:03:36.956 environment and what sense isn't? And in what sense is 0:03:37.016 --> 0:03:39.976 it convenient to own one and in what sense it 0:03:40.056 --> 0:03:43.416 is not? And then from there, it would be, hey, 0:03:43.516 --> 0:03:45.976 I've seen on the streets people driving Rivian. What do 0:03:45.996 --> 0:03:50.256 you think about Rivian? Is that any good? At times, 0:03:50.336 --> 0:03:53.616 it may not be deep research, but nonetheless, you know, 0:03:53.646 --> 0:03:55.406 I have a startup as well that I started a 0:03:55.426 --> 0:03:59.786 year ago. We're trying to make decisions for the startup like, hey, which. 0:04:01.236 --> 0:04:05.596 Software should I use for cap table management? Which software 0:04:05.616 --> 0:04:09.456 should I use for payroll? Which vendors? Who's right for 0:04:09.496 --> 0:04:12.676 a company our size? We are very small. We're just 0:04:12.736 --> 0:04:16.956 about 10, 12 people. Who's right for California? Who's right 0:04:16.996 --> 0:04:18.886 if we grow to 100 people and so on? 0:04:19.386 --> 0:04:23.846 I do feel like this distinction between using AI, using 0:04:23.906 --> 0:04:26.756 a language model to do research for shopping and and 0:04:26.996 --> 0:04:31.096 actually handing off the purchase decision to the agent is 0:04:31.156 --> 0:04:35.016 really important right now. Like I basically trust the AI 0:04:35.056 --> 0:04:38.256 to give me information about, you know, product reviews and 0:04:38.296 --> 0:04:41.996 that sort of thing. And I'm curious, like when you 0:04:42.156 --> 0:04:45.596 see people who are using AI on the research side, not, 0:04:45.636 --> 0:04:47.276 you know, we'll get back to agentic in a second, 0:04:47.316 --> 0:04:49.516 but on the research side, like how are they using 0:04:49.556 --> 0:04:51.256 it differently than traditional search? 0:04:53.026 --> 0:04:56.506 Well, I think the biggest difference with traditional search is 0:04:56.606 --> 0:05:01.706 that the search engines are merely curating the information and 0:05:02.006 --> 0:05:07.906 then you are taking the trouble to synthesize it. And 0:05:07.986 --> 0:05:12.446 you get exposed to a lot of information about various brands, 0:05:12.566 --> 0:05:15.206 even if you decide not to eventually buy from them, 0:05:15.226 --> 0:05:18.466 or even if it's not in your shortlist, you still 0:05:18.486 --> 0:05:23.096 get exposed to a lot. Yeah. In contrast, when it 0:05:23.156 --> 0:05:28.056 comes to AI for shopping, I'm outsourcing a lot of 0:05:28.106 --> 0:05:32.246 the synthesizing. And then depending on the type of buyer, 0:05:32.846 --> 0:05:37.626 the effort level varies where some might say, synthesize it 0:05:37.666 --> 0:05:40.076 for me. And then there might be a lot of 0:05:40.176 --> 0:05:42.436 back and forth, but what about this? What about that? 0:05:43.236 --> 0:05:46.676 And in other instances, you might outsource the decision itself 0:05:47.076 --> 0:05:51.376 in terms of, bring it down to the short list. 0:05:51.436 --> 0:05:54.036 And so you're starting off with a short list of 0:05:54.096 --> 0:05:56.596 three brands and a comparison table, and that's where the 0:05:56.656 --> 0:05:59.946 human work begins. A lot of it is based for 0:06:00.026 --> 0:06:03.546 me in terms of the risk of making a wrong decision. 0:06:03.826 --> 0:06:05.606 And if the risk is high and I'm spending a 0:06:05.646 --> 0:06:08.166 lot of money, then I'll actually go back and forth 0:06:08.206 --> 0:06:10.886 with the LLMs. But if the risk is low, then 0:06:10.946 --> 0:06:15.516 I'm pretty much, for the most part, outsourcing everything except 0:06:15.576 --> 0:06:17.756 the final approval decision to the LLM. 0:06:18.136 --> 0:06:21.356 And that seems pretty common, right? It seems like, by 0:06:21.416 --> 0:06:23.816 and large, a lot of people are comfortable using AI 0:06:23.856 --> 0:06:26.746 for doing the research, but they're a lot more hesitant 0:06:26.786 --> 0:06:30.326 to let the agent take that last step. They're wary 0:06:30.386 --> 0:06:35.226 of letting AI actually buy the thing. And, you know, 0:06:35.386 --> 0:06:38.396 it reminds me a little bit of the early days 0:06:38.456 --> 0:06:41.256 of the web, basically the 90s, which I am old 0:06:41.296 --> 0:06:43.896 enough to remember. And, you know, you would go and 0:06:44.136 --> 0:06:46.956 companies had websites and you'd look at stuff. But the 0:06:46.996 --> 0:06:50.416 idea of putting your credit card into the web was like, no, 0:06:50.496 --> 0:06:52.516 there is no way I'm going to put my credit 0:06:52.556 --> 0:06:56.326 card into some random website. And of course, you know, 0:06:56.506 --> 0:06:58.706 PayPal is in fact an example of one of the 0:06:58.786 --> 0:07:02.986 early pioneers here. They were navigating these shifts in real time. 0:07:03.446 --> 0:07:07.646 But now, whatever, 25, 30 years later, we have solved that. 0:07:07.926 --> 0:07:11.006 Everybody buys stuff online all the time. And, you know, 0:07:11.676 --> 0:07:13.676 I would say at this moment, I am not ready 0:07:13.736 --> 0:07:16.736 to just have agentic AI go off and buy stuff 0:07:16.976 --> 0:07:21.166 for me without my input. I could imagine getting there. 0:07:21.186 --> 0:07:24.546 I just don't trust it yet. And so I'm curious, 0:07:24.966 --> 0:07:27.646 you know, we've solved the trust problem before. How do 0:07:27.686 --> 0:07:29.946 we solve it this time? Like what has to happen 0:07:30.306 --> 0:07:33.176 to get to the point where consumers trust agents to 0:07:33.276 --> 0:07:36.136 buy stuff on their behalf, on our behalf? 0:07:36.716 --> 0:07:40.256 Yeah, I think there's a few things that, need to happen, 0:07:40.336 --> 0:07:45.456 some of which is probably the tech infrastructure needed to 0:07:45.516 --> 0:07:50.016 support something like this. The second is a behavior change. 0:07:50.276 --> 0:07:55.986 People doubted in 1998, 1999, whether it would ever happen 0:07:56.026 --> 0:07:58.986 that we would put our credit cards and buy. And 0:07:59.006 --> 0:08:01.246 there were others who said, no, no, no, it'll happen 0:08:01.286 --> 0:08:03.566 next year. And there were folks who said, no, it'll 0:08:03.586 --> 0:08:07.826 happen in five years. And so that's the speculative piece, 0:08:07.906 --> 0:08:10.956 when it happens and for what it'll happen and so on. 0:08:11.536 --> 0:08:14.896 I remember back in the day, I had my own 0:08:14.956 --> 0:08:17.596 students who, you know, I teach at Wharton. So my 0:08:17.636 --> 0:08:20.466 students at Wharton who started the company Warby Parker, where 0:08:20.486 --> 0:08:24.926 you buy eyeglasses from, when they first came and actually 0:08:24.966 --> 0:08:28.746 their term paper in my course was a business plan 0:08:28.786 --> 0:08:30.986 for Warby Parker. And at the end of the semester, 0:08:31.106 --> 0:08:34.406 I chatted with them. And to me, it felt like 0:08:34.606 --> 0:08:36.816 at that point when I chatted with them, it was like, 0:08:36.956 --> 0:08:41.376 I can see some people buying their eyeglasses online. But 0:08:41.736 --> 0:08:44.276 I think when you're buying eyeglasses, you're wearing one, I'm 0:08:44.296 --> 0:08:46.436 wearing one. I feel like I want to try it on. 0:08:46.456 --> 0:08:49.536 I want to see how it feels. And that is 0:08:49.606 --> 0:08:52.886 extremely important. So I kind of my reaction was, it 0:08:52.966 --> 0:08:55.666 makes sense, but I still don't see how it's mainstream. 0:08:56.616 --> 0:08:57.366 But they changed it. 0:08:57.416 --> 0:09:00.606 Well, they sent you the glasses. The thing of sending 0:09:00.636 --> 0:09:03.026 you the tester pair, was that because of your feedback? 0:09:03.406 --> 0:09:04.426 Did you get equity? 0:09:05.186 --> 0:09:07.786 No, I didn't. I'm sure others brought it up to them. 0:09:07.866 --> 0:09:09.706 I won't take credit for that, but I will give 0:09:09.746 --> 0:09:13.926 them credit for kind of saying, we hear you on 0:09:13.966 --> 0:09:18.456 the problem, but we disagree that there is no solution 0:09:18.496 --> 0:09:22.156 to the problem. And so that's the part that I 0:09:22.176 --> 0:09:26.166 was saying is speculative in terms of, when it will 0:09:26.186 --> 0:09:29.686 happen and what needs to happen for that full behavior 0:09:29.746 --> 0:09:34.106 change to happen. I think someone will crack it. And 0:09:34.186 --> 0:09:36.936 I think it'll happen in phases just like e-commerce did. 0:09:37.126 --> 0:09:39.716 We first felt comfortable buying books and we said, hey, 0:09:40.056 --> 0:09:42.136 a book is a book. I don't need to worry 0:09:42.176 --> 0:09:44.136 about how it feels in my hand and so on. 0:09:44.916 --> 0:09:48.496 And then it became electronics. Hey, I bought books for $ 20. 0:09:48.036 --> 0:09:52.286 Maybe I can buy a $ 200 or $ 500 electronics online 0:09:52.326 --> 0:09:55.236 and it's going to be okay. And then it became clothing. 0:09:55.296 --> 0:09:57.656 And then it just kept expanding. And so I think 0:09:57.696 --> 0:10:00.996 it's going to be similar with agentic where there are 0:10:01.036 --> 0:10:04.886 some categories which will be easier than others. And there'll 0:10:04.926 --> 0:10:07.986 be some customer segments where it'll be easier than the others. 0:10:08.146 --> 0:10:11.006 And that's where things will start. And then how that 0:10:11.326 --> 0:10:16.636 percolates or spreads out and becomes more mainstream across all categories, 0:10:16.756 --> 0:10:19.456 that's the subjective piece. And that's what I was kind 0:10:19.496 --> 0:10:21.296 of saying. It's hard to say when that will happen. 0:10:22.116 --> 0:10:25.396 You mentioned books. I mean, famously, Amazon started out selling books. 0:10:25.996 --> 0:10:26.296 Yeah. 0:10:27.266 --> 0:10:29.006 Do you have a bet for what's going to be 0:10:29.046 --> 0:10:30.906 the books of agentic? What's going to be the first 0:10:31.006 --> 0:10:36.186 category where people are comfortable having an agent go make 0:10:36.206 --> 0:10:37.336 the purchase decision for them? 0:10:38.216 --> 0:10:42.716 Yeah, I actually feel like groceries are a really good 0:10:42.756 --> 0:10:45.516 candidate for me. You know, and this is something that 0:10:45.696 --> 0:10:48.436 I actually ask my students on a regular basis. I've 0:10:48.456 --> 0:10:51.756 been doing this for about two, three years where anytime 0:10:51.826 --> 0:10:56.996 I'm teaching my MBA, undergrad, or executive students, every session 0:10:57.026 --> 0:11:00.636 I have on agentic shopping, AI shopping, I ask them, 0:11:01.616 --> 0:11:06.476 would you use an AI agent to shop for groceries? 0:11:07.376 --> 0:11:11.776 Just your usual grocery list. And perhaps you could argue 0:11:11.856 --> 0:11:14.756 that if I'm using Instacart or if I'm using even 0:11:15.496 --> 0:11:18.316 the Whole Foods or the Amazon app to order my 0:11:18.416 --> 0:11:20.736 groceries where the list is there and I'm saying go 0:11:20.756 --> 0:11:24.276 buy it, at some level, I'm kind of saying, hey, 0:11:25.716 --> 0:11:28.876 here's my usual list. You buy it. You make the 0:11:28.916 --> 0:11:33.936 decisions on substitutions and do it on my behalf. So 0:11:33.956 --> 0:11:37.996 I've already outsourced that. That's not a big deal. Tomorrow, 0:11:38.056 --> 0:11:40.936 if someone were to say, here's an AI agent that 0:11:40.976 --> 0:11:43.876 can do end-to-end, just upload your grocery list once, and 0:11:43.916 --> 0:11:45.916 then you don't have to ever ask again, that's one 0:11:45.966 --> 0:11:49.426 I'd be willing to do. The Instacart, the human making 0:11:49.466 --> 0:11:54.526 that decision, for me, is not that different than an 0:11:54.806 --> 0:11:59.466 AI agent saying, hey, I'll substitute this particular brand of 0:11:59.506 --> 0:12:01.666 milk with that particular brand of milk and so on. 0:12:02.196 --> 0:12:05.096 I know what this customer wants. And I'll know in 0:12:05.696 --> 0:12:09.456 two orders that this is working for me or not. 0:12:10.376 --> 0:12:12.356 Am I willing to try two, three times? Yeah. 0:12:12.556 --> 0:12:15.716 And it's relatively low stakes, right? It's not like buying 0:12:15.776 --> 0:12:17.676 a car. If you do it and it kind of sucks, 0:12:17.696 --> 0:12:20.256 you're like, well, I didn't get the right milk this week. 0:12:20.596 --> 0:12:21.996 Not that big of a deal. 0:12:22.176 --> 0:12:25.826 Exactly. So I'll happily try that a few times. So 0:12:25.846 --> 0:12:28.206 I feel that's a good one. I also think the 0:12:28.266 --> 0:12:32.186 other one is where individuals are doing a lot of 0:12:32.506 --> 0:12:35.556 research and the research is a big part of it. 0:12:35.806 --> 0:12:39.236 Insurance is a good example. If I were buying car insurance... Mm-hmm. 0:12:39.826 --> 0:12:42.666 I would be happy to have Muse or somebody else 0:12:42.746 --> 0:12:46.076 go to all the different websites, fill in my information, 0:12:47.026 --> 0:12:49.416 get me a bunch of quotes, and then pick one. 0:12:49.796 --> 0:12:54.996 Maybe with car insurance, I might still want the final decision, 0:12:55.056 --> 0:12:58.536 like an approval gate. To me, that's still agentic if 0:12:59.076 --> 0:12:59.936 it's doing everything. 0:12:59.996 --> 0:13:02.576 I mean, that's a big purchase, and it's a lot 0:13:02.616 --> 0:13:04.796 of money. And if it gets it a little bit wrong, 0:13:05.916 --> 0:13:08.876 you're paying... you know, or buys the wrong policy. Like 0:13:09.516 --> 0:13:11.676 that one feels way scarier to me than milk. 0:13:12.216 --> 0:13:13.936 Yeah, but give me the chance to approve it. And 0:13:13.956 --> 0:13:15.816 that's all I need. You give me all the data. 0:13:15.876 --> 0:13:20.236 You've done all the work. In effect, from the brand's perspective, 0:13:20.356 --> 0:13:22.436 AI is the customer. It's doing all of the work. 0:13:22.576 --> 0:13:25.016 And I'm just kind of saying, okay, everything looks good. 0:13:25.096 --> 0:13:28.966 Let's proceed. That's the one right I want with insurance. 0:13:28.986 --> 0:13:31.096 But with groceries, I don't even need to approve it. 0:13:31.126 --> 0:13:33.446 I'd say just do it. If it breaks two out 0:13:33.466 --> 0:13:35.186 of three times or three out of three times, I'm 0:13:35.226 --> 0:13:37.536 not repeating it. But if it works all three times, 0:13:37.596 --> 0:13:38.296 I'll do it forever. 0:13:38.956 --> 0:13:43.456 So let's talk about what happens when something goes wrong. 0:13:43.896 --> 0:13:45.836 If the agent does something I don't want it to do, 0:13:45.876 --> 0:13:48.646 it makes a mistake. If it goes off the rails, 0:13:49.566 --> 0:13:53.666 we know that businesses are split over who should be 0:13:53.726 --> 0:13:56.716 responsible in this kind of case when there's a dispute 0:13:56.746 --> 0:14:01.206 between the AI platform and the customer. And it's clear 0:14:01.346 --> 0:14:04.866 that most businesses think that the merchant who fulfilled the 0:14:04.986 --> 0:14:08.846 order should not be the one who's financially responsible when 0:14:08.886 --> 0:14:11.766 things go wrong. And so the question is, who should 0:14:11.786 --> 0:14:14.606 be on the hook if the AI, say, makes an 0:14:14.786 --> 0:14:15.726 unauthorized purchase? 0:14:16.346 --> 0:14:20.206 That's a great question. I think anyone who's pushing agents 0:14:20.826 --> 0:14:25.346 to make them mainstream for shopping, any company that's doing 0:14:25.386 --> 0:14:28.386 that will have to assure customers, we are on the hook, 0:14:28.446 --> 0:14:32.426 we'll take care of it. So that's when it goes mainstream. 0:14:32.466 --> 0:14:34.606 I think that's when you get past the trust hurdle. 0:14:35.366 --> 0:14:39.886 So if Muse wants to do that for me, or Instinct, 0:14:39.986 --> 0:14:42.826 another app or startup that wants to do this for me, 0:14:43.266 --> 0:14:46.166 if they want me to go full agentic, they've got 0:14:46.186 --> 0:14:50.746 to take on the liability risk and say that if 0:14:50.846 --> 0:14:54.906 our agent does something that deviates from this intent that 0:14:54.936 --> 0:14:56.836 you had, then we will reimburse you. 0:14:57.766 --> 0:15:01.106 You mentioned Muse. I mean, that's the meta... agent that 0:15:01.166 --> 0:15:05.846 is new and buzzy and instinct, which is sort of 0:15:05.866 --> 0:15:09.866 a startup, right, that wants to do something similar. Invite only, 0:15:09.946 --> 0:15:11.596 I believe. I haven't been invited yet. 0:15:12.356 --> 0:15:13.076 I can invite you. 0:15:13.176 --> 0:15:13.536 Okay. 0:15:13.596 --> 0:15:14.016 I'll do that. 0:15:14.036 --> 0:15:16.376 I mean, I'm curious, what do you think it's going 0:15:16.396 --> 0:15:18.956 to take? You know, there is this leap that a 0:15:19.036 --> 0:15:22.796 lot of consumers haven't yet made. That leap is letting 0:15:22.866 --> 0:15:26.706 an AI agent buy something for them. And I'm curious, 0:15:27.466 --> 0:15:29.006 what is it going to take for that to happen? 0:15:29.226 --> 0:15:32.546 And And do you think agents like Muse are going 0:15:32.566 --> 0:15:34.866 to be part of the key to this transition? 0:15:36.086 --> 0:15:39.306 Yeah, I think agents like Muse are key to the transition. 0:15:39.386 --> 0:15:42.026 So what Muse and other agents can do is you 0:15:42.066 --> 0:15:45.306 can articulate your preference, just like you do in a 0:15:45.346 --> 0:15:48.846 prompt on an LLM. They can go to different websites, 0:15:48.926 --> 0:15:53.226 gather the information, reason through that, fill forms if need be, 0:15:54.206 --> 0:15:57.866 make a decision, present it to us. So all of that, 0:15:57.966 --> 0:16:00.846 those systems do. And like I said, I use it 0:16:01.026 --> 0:16:04.006 up to that point. They can also go and transact 0:16:04.046 --> 0:16:06.356 and that's where they need to get past the trust hurdle. 0:16:06.436 --> 0:16:11.416 And I think for that to happen, first, they need 0:16:11.456 --> 0:16:14.536 to insure me about things going wrong, as you were 0:16:14.576 --> 0:16:17.516 just alluding to earlier. So they have to tell me 0:16:17.576 --> 0:16:21.136 that there is some protection. Just like, by the way, 0:16:21.196 --> 0:16:23.676 a credit card company, when it's giving us a credit card, 0:16:23.696 --> 0:16:28.206 they're saying if there's some fraudulent transaction, You're not liable. 0:16:28.266 --> 0:16:30.586 You can appeal and you can say, this was not me. 0:16:31.346 --> 0:16:35.846 Something's gone wrong. And they will make sure that no 0:16:35.866 --> 0:16:39.286 harm is done. And they're providing that guarantee. That's what 0:16:39.346 --> 0:16:41.986 needs to happen in this world. So someone needs to 0:16:42.026 --> 0:16:46.556 be the guarantor or the trust builder here. So that's 0:16:46.656 --> 0:16:49.356 the piece that is needed beyond just the technology needs 0:16:49.396 --> 0:16:49.716 to work. 0:16:50.796 --> 0:16:53.776 And it seems like that will happen, right? It seems 0:16:53.816 --> 0:16:59.096 like the nature of the history of finance is... lowering 0:16:59.216 --> 0:17:03.356 friction is the way things go. There's less and less 0:17:03.436 --> 0:17:06.086 friction to buy things. A credit card is lower friction 0:17:06.106 --> 0:17:08.606 than cash. Shopping online is lower friction than going to 0:17:08.646 --> 0:17:12.086 a store. And it seems like ultimately, agentic AI will 0:17:12.126 --> 0:17:15.566 be lower friction still, and someone will figure out how 0:17:15.606 --> 0:17:19.126 to solve the trust problem, and people will use agents 0:17:19.186 --> 0:17:21.346 to buy things on their behalf as often as we 0:17:21.396 --> 0:17:22.676 buy stuff online now. 0:17:23.256 --> 0:17:25.536 Yeah. I mean, look, I'm not a betting man, but 0:17:25.576 --> 0:17:28.666 if you were to say, hey, What's your prediction here? 0:17:29.186 --> 0:17:32.306 Pick a side. For me, it would be, you know, 0:17:32.346 --> 0:17:36.326 there's no doubt the history of shopping shows that we're 0:17:37.026 --> 0:17:43.436 always willing to adopt technologies that add convenience while guaranteeing trust. Now, 0:17:43.876 --> 0:17:45.736 is it going to happen for every kind of shopping? 0:17:46.056 --> 0:17:49.776 Maybe not. Is it going to happen for every person 0:17:49.856 --> 0:17:53.136 on earth? Maybe not. But is it going to be mainstreamed? 0:17:53.616 --> 0:17:56.676 For large pockets of shopping, my bet is that it 0:17:56.716 --> 0:17:57.696 will happen. 0:18:01.196 --> 0:18:14.906 We'll be back in just a minute. So we've been 0:18:14.946 --> 0:18:18.906 talking essentially about the consumer point of view, about what 0:18:18.946 --> 0:18:22.186 it's going to take for people to let agents buy 0:18:22.246 --> 0:18:25.446 stuff on their behalf. I want to talk now about 0:18:25.546 --> 0:18:28.106 the merchant point of view, about the retailer point of view. 0:18:28.726 --> 0:18:30.986 This is a thing you have written about. You wrote 0:18:31.006 --> 0:18:35.086 an article about it recently in the Harvard Business Review. And, 0:18:35.946 --> 0:18:40.666 you know, it does seem like most business owners largely 0:18:40.806 --> 0:18:45.066 trust AI platforms to represent their brand to customers. But 0:18:45.136 --> 0:18:47.456 it is clear that there is, you know, to some extent, 0:18:47.566 --> 0:18:52.576 a trust issue here, that the AI won't accurately represent 0:18:52.616 --> 0:18:55.036 the brand. What are you seeing here? What are you 0:18:55.076 --> 0:18:56.876 hearing from the merchants you talk to? 0:18:57.356 --> 0:19:01.696 Yeah, I think merchants have to worry about this too. 0:19:01.796 --> 0:19:05.526 And there's many layers of the merchant concerns. One is, 0:19:05.726 --> 0:19:09.986 are the agents representing the brands correctly? And so I'll 0:19:10.006 --> 0:19:13.526 give you an example. I was talking to this seller 0:19:14.246 --> 0:19:19.246 of vintage clothes in London. And so this person owns 0:19:19.326 --> 0:19:22.626 this brand. They sell in stores. They also sell online. 0:19:23.676 --> 0:19:27.676 And they sell vintage women's clothing. He actually bought the 0:19:27.716 --> 0:19:31.166 brand off of the previous owner. And he was kind 0:19:31.206 --> 0:19:35.926 of saying, it's shocking that we are the number one 0:19:36.246 --> 0:19:41.846 seller of vintage clothes in the UK. And you go 0:19:41.886 --> 0:19:45.346 to ChatGPT, you go to Gemini, and you ask this question. 0:19:45.406 --> 0:19:49.096 I'm in London. Where can I buy vintage clothes with 0:19:49.216 --> 0:19:52.566 70s look, 60s look, and things like that? or Vintage 0:19:52.666 --> 0:19:55.836 Inspired rather is their brand, and his brand would never 0:19:55.876 --> 0:19:59.156 show up. And he was like, what's going on? And 0:19:59.216 --> 0:20:02.396 so I actually went through this a little bit. We 0:20:02.476 --> 0:20:05.616 found what was happening was that there was a post 0:20:05.696 --> 0:20:09.166 on Reddit where someone had mentioned that this brand is 0:20:09.186 --> 0:20:13.546 shutting down. And the history there is that the previous 0:20:14.266 --> 0:20:19.056 owner had put it up for sale And probably the 0:20:19.096 --> 0:20:21.756 word got around that the previous owner was struggling to 0:20:21.796 --> 0:20:24.526 make it work. And then, of course, new buyers took over. 0:20:24.566 --> 0:20:27.906 But there was this post online. So ChatGPT had probably 0:20:27.966 --> 0:20:30.746 decided that it shut down or something like that and 0:20:30.786 --> 0:20:33.846 was not showing the brand. And so that's an example 0:20:34.286 --> 0:20:39.906 of the brand being misrepresented by an LLM. 0:20:40.966 --> 0:20:43.946 What do you do about that? Like, this is a problem, 0:20:44.026 --> 0:20:46.706 and presumably it's not in the LLM's interest, right? The 0:20:46.746 --> 0:20:49.006 maker of the model doesn't want it to be wrong. 0:20:49.436 --> 0:20:52.196 The merchant doesn't want it to be wrong. Is this 0:20:52.256 --> 0:20:53.356 a solvable problem? 0:20:54.476 --> 0:20:56.756 It's a tough one to solve because the maker of 0:20:56.796 --> 0:21:00.116 the LLM is not going to say, every time you're 0:21:00.156 --> 0:21:02.426 not happy with an answer, appeal to us, and then 0:21:02.466 --> 0:21:04.846 we'll have a human look through it and correct it. 0:21:04.886 --> 0:21:09.046 That doesn't work for them. And so while it is 0:21:09.106 --> 0:21:13.666 in their interest, there's just too much information on the web, 0:21:14.606 --> 0:21:18.706 too many vested interests. And then how do you also 0:21:18.766 --> 0:21:22.746 address people who might complain about subjective things and say that, 0:21:22.806 --> 0:21:26.846 how come I'm not listed as the top three vintage brands? 0:21:26.906 --> 0:21:29.426 And in fact, our survey shows that we're among the 0:21:29.466 --> 0:21:33.266 top three. And so they're not going to provide resources 0:21:33.326 --> 0:21:36.126 for that. So unfortunately, what it ends up becoming is 0:21:37.136 --> 0:21:39.616 This is what I tell a lot of brands, that 0:21:39.676 --> 0:21:44.186 this world is a lot like your customers are sometimes misinformed. 0:21:45.086 --> 0:21:48.146 Your customers also need to be persuaded. And just like 0:21:48.226 --> 0:21:51.926 you've spent decades thinking about, how do I ensure I 0:21:52.086 --> 0:21:55.566 inform my customers about my brand? How do I make 0:21:55.606 --> 0:21:59.386 sure I persuade my customers about your brand? You have 0:21:59.426 --> 0:22:02.106 to do the same for AI agents now. And you 0:22:02.146 --> 0:22:04.886 have to start thinking about the AI agents as another customer, 0:22:05.416 --> 0:22:08.186 And the same questions you had to have for human customers, 0:22:08.276 --> 0:22:12.626 now you need to answer those for AI agent customers. 0:22:12.706 --> 0:22:16.866 And it's a new problem for marketers and brands and 0:22:16.886 --> 0:22:18.066 so on. So, right. 0:22:18.086 --> 0:22:20.846 So, I mean, the kind of central theme of the 0:22:20.886 --> 0:22:26.086 merchant point of view section of your article is companies, merchants, 0:22:26.196 --> 0:22:29.316 marketers need to think about the agent as a customer. 0:22:29.796 --> 0:22:32.436 And sort of you dig into that and you lay 0:22:32.536 --> 0:22:37.416 out these three common misconceptions that you have heard, you know, 0:22:37.496 --> 0:22:40.026 in your work with retailers and marketers. And so I 0:22:40.056 --> 0:22:42.546 actually want to walk through those and have you kind 0:22:42.586 --> 0:22:44.726 of talk us through them. So I'm going to read 0:22:44.766 --> 0:22:45.926 them to you and then I'm going to ask you 0:22:45.946 --> 0:22:49.626 what they mean. So the first one, misconception number one, 0:22:49.986 --> 0:22:53.306 is there will always be a human at the other end. 0:22:53.986 --> 0:22:56.246 Tell me about that. Like, what are you hearing? Why 0:22:56.626 --> 0:22:57.636 is that a misconception? 0:22:58.756 --> 0:23:01.986 Yeah, so I think a common one that a lot 0:23:02.006 --> 0:23:06.546 of marketers and brands think is, okay, AI is probably 0:23:06.586 --> 0:23:10.986 another channel like search, where in search engines, we have 0:23:11.026 --> 0:23:14.416 to think through how we show up in Google's results pages, 0:23:14.506 --> 0:23:17.156 but there's always a human reading it and making the decision. 0:23:18.026 --> 0:23:19.756 And so if you think about it that way, then 0:23:19.776 --> 0:23:21.956 you would say, hey, maybe the human is doing the 0:23:21.996 --> 0:23:25.636