Latest / Talk Commerce / Live from Shoptalk: Navigating the Future of Commerce Through Agentic AI with Pano Anthos
Transcript
- Speaker 2 (00:00.13) Thank you. Speaker 1 (00:23.692) Okay, I think these are... These will work. They kind of keep the sound focused on your area. Excellent. Yeah, Pano, thanks for joining us. Welcome everybody to a special episode of Talk Commerce. We're live here at Shop Talk Fall in Chicago. And I'm here with Pano Anthos of XRC Ventures. Pano, the first thing I wanted to ask you is that, you know, reading your press kit, I see you guys are founded at the intersection of consumer behavior and technology. and your founding member. So I wanted to ask you what it is about that combination that you found so intriguing and saw so much potential in. Well, first of all, consumers are responsible for two trillion in spend and a massive portion of our GDP. So whatever they do, we should pay attention to, number one. Number two, they tend to be relatively much faster early adopters of technology than corporations. So if you really want to know what's happening is what is the consumer doing with technology? It could be kids, it be adults, it doesn't really matter. We love enterprises, don't get me wrong, we are big supporters of retail and we have number of retail LPs. But retailers, like most corporations, move to a certain drum beat about adoption. And so if you really want to understand where things are going, get to the consumer first. Speaker 1 (01:52.494) Yeah, that makes a lot of sense. know just personally I tend to make decisions a lot faster with my own spending, with my company's spending, I can definitely see how that plays out. Absolutely. And it actually makes me think of a point that was brought up in the panel that you hosted today, which is that a big part of successful AI implementation has to do not just with the technology itself, but with the culture within a company. for the whole organization to adapt and think about those things and change their mindset. Could you talk a little bit about what a successful organizational mindset is when it comes to adopting tech like agentic AI? think what you heard on the panel was that first of all, stuff has to be written down. And secondly, you have to have a set of goals and then you have to compare what you do today to your goals and decide is this set of processes worth automating, replacing? So there's a lot of process re-engineering that goes on. And then you ultimately decide, all right, so we know here are the rules, we all agree on these rules, that's when you encode them with AI. Mm-hmm Speaker 2 (03:06.298) AI to figure out the rules. You have to really lay out the rules first. And that's the misconception of autonomous AI is that it will make decisions within boundaries. But you have to set those boundaries or you get nothing. You get craziness. George Washington being African American or something stupid in terms of history, right? Yeah. Speaker 1 (03:28.622) Great, well, this might be kind of a step back, but it is a question I wanted to ask and get your answer to, which is just tell us for those who are late to the party, what is agentic AI? How would you define it? Well, so agentic AI has some key components. One is it's adaptive with respect to integration. So you need to have agents that can integrate to existing systems fairly easily. They have to be supple, have to be API oriented, but that's not sufficient. They've got to actually have, they've got to be autonomous. They have to be able to run and make decisions within some boundaries. They you know and when we talk about that the boundaries are corporate governance boundaries as well as consumer boundaries like what what are the consumer boundaries and I'll show you something I won't show it to you but you get the idea that when you think about Sorry, I'm gonna take a second yeah, take your time. Speaker 2 (04:44.428) Yeah, I worked that. So when I went through that definition, right, there were four key things. Autonomous planning, so it does its own decision making. Adaptive reasoning, so it learns and remodels the behavior. It doesn't just simply try to do the same thing over and over again. There are its tool integration, so it's working around the conversing system. And it's goal oriented, it's goal directed. You have to set what the goals are. If you want conversion at one of the, about conversion, if you conversion to be 20%, 30%, 40 % of sale, whatever, you've got to like specify that and it works toward that goal. That is also part of the challenge because goals can conflict, right? You can optimize for the wrong thing if you're not careful. get not bad behavior, get results you really don't want. Right. Speaker 1 (05:43.79) Right, optimize for traffic but let the conversion plummet or something along those lines. Could you tell us maybe about a really exciting concrete example of agentic AI that you've worked with or seen or do you think illustrates its capabilities? actually... We're about to invest in a company that is using agentic in the supply chain space. So you think, well, supply chain, how's that relate? Well, customers expect products and there's all this work that has to be done between plants and manufacturers to deliver the product. So what happens is when a plant, when a manufacturer says, want to build something, they start contracting with manufacturers to build it usually, not their own. That contraction process, not just the initial contract, but then the placing of the purchase order and then revisions and then the back and forths. There's a very set of manual tasks today. I'm sure there's some automation of the purchase order in terms of like it's not printed, it's PDF, but really the back and forth is very tedious and manual and requires a lot of supervision. We can't make the ship data. 413, okay, then they come back as well. How about 414? You know, so there's back and forth all that this team out of Israel has figured out how to automate using an LLM to basically take all those messages they're going back and forth and Inter just an intermediate Literally intercept them in appropriate ways and then make decisions based on the rules that have been set by the organization So it's delivery dates is like you get a two-day window one day or whatever it is or these other things have to happen for this date to be accepted. So that one's a really exciting one. They're growing really fast because it's a big pain point. We've been investing in AI for a while. I wouldn't have called it agentic AI. That has more to do, I think, with the fact that LLMs now are so accessible. Speaker 2 (07:45.166) You used to to train your data, which was the whole point of machine learning or computer vision. Big exhaustive exercise. It's much easier now. And actually, almost any AI application at its core has the ability to be if its turned loose and let run. And most of them are. So it's really... Uh-huh. Speaker 2 (08:12.078) I wish I could say, it's something different. It really isn't at the end of day. Yeah, I'm getting the idea that it's less of a solid line in the sand and more of a spectrum of highly agentic AI to less and less agentic AI. I mean, even let's take the example with Waymo, Waymo is far more agentic than the Tesla full service, full self-driving feature because Tesla requires you to keep your hands on the wheel. though, well, at least touch it every now and Waymo, there's no one in the car. But his point, Paul's point was like, why is there a steering wheel? And why is there a driver's seat? Right. So even in something that you see as being More loosely coupled or more detached from a person or human, you still have this ability to essentially, and my guess is the steering wheel is there for a reason. Right. something goes wrong, you got to be able to interface, to interdite. Yeah. Okay, so a large part of our audience and really a lot of people I work with are small ecommerce shops, know, maybe three, four, five people running the show. What are some examples or like implementations of agentic AI that those types of entrepreneurs or business people should be looking at? Speaker 2 (09:38.742) So there's a lot being done right now in both customer support, which you heard about, and in search. So the thing that I recommend to these e-commerce providers, these small e-commerce solutions, is you've got to really understand where you sit in the AI stack. Just like with Google, you say, well, where are you when you're ranking, website ranking? You kind of have to understand where you sit by asking the agents, or the chat engines. you know, do you know I exist? And then realizing how they go about, and this is very manual today, but how they go about collecting their data. Because they actually tell you the sources they're pulling from. And then making sure that you go back to those sources and you become the best of the content in those sources. It's fairly manual today. It's not that different than it was in the early days of Google. Right. Backlinking. You would sit there and do backlink off backlinks all day long and just say, hey, can you link to me? Can you link to me? And what kind of increase your status? Yeah, what what have you heard as far as a key methods of getting yourself recognized by? Say chat GPT has an important source Speaker 2 (10:48.91) It's Reddit. I would say if you look at every one of these engines, they go to Reddit. They don't trust, they don't have access to Amazon. So if you look at reviews, you got a problem right away with the Amazon reviews. If you ask any consumer and you look at the agents, they're not looking at the website reviews very much because frankly website reviews, you know, are they really true? But Reddit, you get pretty bald faced on this. yeah, I mean I think a lot of people I've talked to and myself included will type Reddit and then you know whatever you want to do just because you'll get a more unbiased opinion. right. that is, and what we've been saying this about the consumer brand space for years, you've got to build a community. You've got to be on Reddit. You have to be part of the conversation and you can't be selling. I mean, that's the whole point is no one wants to hear you moaning and pitching and blah, blah, but you need to be part of the conversation, which is why do you exist? What's the problem you're solving? Talk about the problem. Yeah, this was going to lead exactly into my next question, which is, you know, Reddit has like no tolerance for BS and self-promotion. So how do companies position themselves on Reddit? authentic and you got to be accurate on the content and the problem that you're going to try to solve and you know you can always say you can always just warn people FYI you know this is my company but your case you're making is independent of your company. Speaker 2 (12:20.942) And that's important. And then people will respect it and then they'll take another stab at it or whatever. But you have to be transparent about it. You just can't hide. I'm an independent consultant. In fact, you're trying to sell your own company. Speaker 1 (12:37.006) Okay, so another area that I saw highlighted from your team was specifically about CAC and agentic AI. Could you talk a little bit about that? You saw one of the fellows frame, co-frame, it's got all about landing pages and conversion rates. mean, CAC is one of those constant whack-a-moles. You you never solve it. You solve for it and I think volume makes a difference. So I like what co-frame is trying to do. I like what Manju is trying to do over at Spanable. Because I do think that anything you can do to improve is converted. is gold. You've paid for that customer, or if that customer's found you one way the other, as long as they're not a bot, what are you gonna do with them? How are you taking them forward? So those solutions are good ones. mean, you know, there's always where the one bounces or the companies that are out there for a long time. They're gonna have to morph if they haven't morphed already. They're all going into this agentic framework. They don't have a choice. If you had a traditional Java-based or Python-based solution or Rails, you have to rethink it with respect to this agentic framework and how to allow personalization to happen on its own. Yeah, makes sense. Manju said something that kind of blew my mind during the talk, which was, you know, people don't always go to your physical store anymore, but you still sell to them. And what we might see next is people don't go to your website anymore. Speaker 2 (14:19.544) I guarantee that's not going to happen. guarantee that's going to happen. I about that, actually have a whole presentation I gave to the French Consulate during NRF in January. I said, your website is toast. Unless you are a fashion-oriented product where discovery is important and inspiration is important and it's truly discretionary, the chat engines are going to take over. So if it's hard good, if it's bottled water, the best bottled water on the market, Yeah. What's the best anything? Anytime you ask what's the best, it's over. They're not going to go to your website. And they're going to take the results. If you look at Perplexity Shopping as an example, they'll take those results and they'll provide a checkout page right there. You don't have to go anywhere else. Dick's does this and Dick's is in danger of losing that website traffic. But at the same time, if they get the sale, it doesn't matter. Yeah. Speaker 1 (15:23.15) Right. Now what they do lose, however, if they don't, there's some compounding effects, and one of those is retail media. So if your website traffic goes down, your retail media is spent, and revenue goes down. You want retail medias, right? Yeah. So that's an important component of ad spend today, is revenue for the brands and retailers. there's going to be some hell to pay, and some business model shifting around. Yeah. Speaker 1 (15:52.522) Right. if checkout really does happen outside of the website, which it can do. Right. Well, and I think a lot of it, you know, might be headed in the direction of completely being automated. Like the bottled water example you gave, you know, just monitor our inventory. We need stuff that delivers within two days and whenever, you know, price. Yeah, I can see this whole set of parameters, you know, with water might be fairly limited, but it's more complicated. You just look at reorders and you automate. But there's also restocking is now, there's a lot of automation and restocking anyway. Because our planning systems are looking at what's to buy, how much we bought, how much we sold and how much do we need going forward. It's really more the consumer experience of the website being unnecessary. already. Speaker 2 (16:37.87) Until you have all these small e-commerce brands, I would figure out how to be on agentic shopping early and not be left behind. Your website is not important, no offense, you know, but unless you're a fashion brand or you're an influencer of some magnitude, your website's going to be irrelevant. Yeah, how soon do you think that will be true? It's already happening. But again, it's going to be in those hard goods spaces where comparison really matters. know, like TVs and hard goods, electronics, you can compare them all day long. Anytime you say, what's the best? Right. You're gonna, so that business is happening. would say, I would give it, given the volume of people using the chat engines today, Chat GPT is gonna introduce shopping pretty soon. They've already said they're gonna do that. Alexa already has it, although it's a separate website. It's a weird conglomeration. They haven't really made it part of the core layer. So in a chat service itself, you just type the chat in, request in, you know, I'm gonna have to hit the shopping framework. That's important. Speaker 1 (17:48.942) Yeah. Right? Yeah. Interesting times. So I would say year maybe and you're going to see more more traffic. The traffic is just going to drop. Right. I guess another thing that I see a lot in my line of business is companies like some of larger businesses maybe B2B switching to you know more customized frameworks for their website like Magento or something and I know those do a lot of work on the back end too. Do you think that side of things will still be relevant even as... No. Yeah, I mean it's you're gonna have a checkout page that's gonna be inside perplexity as you do today with shop or you know third party checkout environments and you're just you're gonna get the transaction you'll get the email address as you get the first party data but you will not have you will not have the opportunity to to sell more of other things so my example is the bike radar example right so have you ever used perplexity shop? No, actually Speaker 2 (18:53.454) Your listeners won't be able to kind of figure out what we're doing necessarily, but... It's scary. So, okay. So first of all, come up to our website's screen. And I know why barbecue was showing up, but it is. But let's say, what is the best? This is my favorite. Speaker 2 (19:26.318) still tiller it, right? Speaker 2 (19:32.162) So it looks at searching, Garmin, looks at best cycling gear, looks at all these different sites. There are several. Okay, so right away, you're getting all of the information that you would otherwise go to a website for here. The rating, the reviews, the number of reviews, you can actually click on the reviews right here. The buy now button, right there. So if I choose Wahoo Tracker. This is it. I don't go to their website. Right. I'm buying it through REI. I might even go to REI's website. I'm buying with Pro. That's it. Tons of content. Review sources, all the places. Reddit's here, CycleEat News, CycleEat Subs. It's doing in seconds what it would have taken anyone, including you and me, hours to research. Right, definitely. So why would you ever do it any other way? The minute anyone finds out that this is available, it's game over. Right, and also because you can, I mean I'm sure you can customize it to other things you're looking for. Of course. So you know it's a completely personalized landing page. Speaker 2 (20:38.595) This is your, it's like having a concierge at your fingertips. Yeah, and then to go back to the point you just made, the whole like loss leader or like, you know, get one thing to win their business and sell them, cross sell them other things. You can't do that. Right. It's a big problem. It's gonna be a big big problem on cross-selling Yeah, well I think, you know, we all know AI is disrupting all kinds of things and will continue to do so. I think a big lesson from that is like it's gonna change a lot of things and disrupt a lot of things, but in that area there's obviously a lot of opportunity if you're ahead of the curve in the way that... You are, you know, for that transaction, yes. To build some brand awareness, maybe. Cross-sell, absolutely not. Speaker 2 (21:28.366) Plexi permits it because they're owning that. And then you don't even have control of the cross-sale. They could bring someone else's cross-sale. They could provide advertising. And they will. They're going to have advertising. And they're going to basically allow people to promote their products when bicycle radar tail light gets entered. Yeah. Speaker 1 (21:45.859) Yeah. It's gonna be, it is the Wild West. But as I said, I think the way you win as a small e-commerce brand is you've gotta be on top of every content engine. You gotta be the best. You know, if you're a tail light, if you're making tail lights, if you're Garmin, you better be in cycling news, in all these places. Like it's a war. Right. Okay, so what would you tell your content team who's been creating like web pages and blogs and stuff on your site to start doing? Start using the engines and asking all the questions that any consumer and they give you all the questions that consumers can ask and go figure out whether you're in the top three or top one or top two and if you're not look at the sources and go to those sources and go figure out how to get yourself exposed to the sources. Now if your product sucks, then you got to hope. Right. Because you can't fake the engines aren't faked easily. They'll be gamed maybe by advertising, but they're not gonna be faked. So Reddit's not faked. So you can't really get away with that most people don't know what Reddit is. Over the age of 40, no one knows what Reddit is. So that's problem is that, or 45, whatever. So there's a whole class of customer that brands can sell today because they have an inferior product, but no one really knows that. Speaker 1 (22:54.932) Right. Speaker 1 (23:16.426) Right. But now the AI is about to bring Reddit to the masses. Exactly. it's going to bring transparency, which is great for everyone, it's going to make manufacturers make better products. So I think that's a good thing. I always think transparency is a better thing than obscuring or hiding or believing that inaccessibility is your best. Yeah. Well, thanks for all those insights on small businesses. guess to wrap things up though, let me go out to a bigger perspective. What do you think happens to like Google and the like legacy search engines? I mean Google, it's not toast, let's be honest. But the judges in the trial that just came out last week or two weeks ago, it's pretty obvious that the judge knows that what we all know is Google search in the traditional SEO, SEM world, it's over. And it's moving hyper fast. I think about how many users are on ChatGPT today. and on Perplexity and on Claude. And then you imagine it's only a minute. So they're already asking these questions. They're not immediately getting a shopping experience, but they're gonna get one. And if they don't get a shopping experience, they'll at least get answers. You don't need Google anymore. So what Google has been betting on is that you're gonna use their search bar. But, and this is the big but, I think the only way they win on the search front with it, Speaker 1 (24:31.832) that. Speaker 2 (24:51.382) and I or whatever to kind of is by buying their way onto your phone, by buying their way onto anywhere where you would, you you the browser and the way you get first. Right now, at Safari, you get Google. Because they pay for it. They pay billions for that. They're going to pay a lot more because, maybe not, but they're going to have to pay dearly to keep that title. And I think that's Right. Speaker 2 (25:21.232) Then it's question of the browsers don't get the action because you know, perplexity has got their own browser now. It's not the best browser in the world, it's kind of painful to use. I've used it. You can't just type in domain names easily, it's like perplexity dot dot dot. Focus. Speaker 1 (25:38.795) Yeah, it's double early in the game. It's too early, but... But you can see where their head is. It was just like, wait a minute, I could have a browser do 50, 30, 80 tasks engineered differently and change what a browser does to be more of a butler than a browser. So browsers made sense in the Google world. It doesn't make sense going forward. So it would be really interesting to see what happens. So I think Google, I think, again. Right. Speaker 2 (26:10.414) They'll last a lot longer than anyone can possibly imagine. I I think there's still a Kmart store around somewhere. It's hard to kill a company and I don't think Google wants to go down and they've got plenty of money. But they're gonna have to spend a lot to keep people's attention. A lot. You know, if I didn't use Gemini in my little presentation, I use Claude, perplexity, and chat-chip. Gemini is like, I don't even think about Gemini. Yeah. Speaker 1 (26:36.36) Yeah. pray a same. Exactly. Think about that for a moment. When was the... Ten years ago? Google was everything. Five, I would say. Five, I would say. Five years ago. Google that. I'm Google this. Google that. It's a verb. Google it. It's not gonna be the case anymore. Speaker 1 (26:59.746) Yeah. Well, and then think about Yahoo, which, or AOL, you know, they were massive in their day and now they're these dinosaurs that are kind of irrelevant. It's the innovator's dilemma. And the reality is Google had a shot at a very easily built ChatGPT That's the classic innovator's dilemma, and they screwed it up. Because they were sitting too fat and happy with too much revenue and still believing that they could be disrupted. Yeah. Speaker 1 (27:26.796) Yeah. Well, thanks so much for giving us a lot to think about. Absolutely. We usually end with a chance for you to like pitch one thing or send people one place. You know, our listeners, what should they where should they go to learn more or to where would you like? I mean, think our website's good. Our LinkedIn page, we do a lot of LinkedIn content. We have two great sub stacks, a consumer tech sub stack with about 15, 20,000 subs already, and a brand fund sub stack. Our sub stacks are really a lot of activity today. And, you know, use technology, use our startups. Yeah. Awesome. Well, thanks so much for your time. Thanks.