← The Jason & Scot Show - E-Commerce And Retail News
The Jason & Scot Show - E-Commerce And Retail News
EP337 - Agentic Commerce at Citi's Global Consumer & Retail Conference
00:51:55
Loading summary
Transcript185 lines
- [00:02]
A
Welcome to the Jason and Scott show, your source for the latest news and trends in the e commerce industry. Featuring hosts Jason Retail Geek Goldberg, Chief Commerce Strategy Officer and publicist, and Scott Wingo, CEO of Get Spiffy and co founder of Channel Advisor. Here are Jason and Scott.
- [00:23]
B
Welcome to the Jason and Scott show. This is episode 337 being recorded on Tuesday, March 10, 2026. I'm your host, Jason Retailgeek Goldberg and as usual I'm here with your co host Scott Wingo, who you'll hear from in just a minute. On March 10th, Scott and I were invited to participate in a panel at Citi's 2026 Global Consumer and Retail Conference in Miami, Florida. We spoke with Ron Joosse, the Managing director of Citi's Internet Equities Research and and Paul Lajoux, the retailing head of America's Consumer City Research. We had a wide ranging conversation about agentic commerce and the disruption to the traditional retail model and we were able to record the discussion for all of you. So without further ado, here's our conversation with Ron and Paul.
- [01:10]
C
So thank you very much for joining. My name is Ron Josey. I lead the U.S. internet coverage team here at Citi and it's rare that I get invited down to the Consumer Retail Conference. So I'm sort of thrilled. Not sort of I am, but thrilled, excited to be here. And we're talking about a topic that's frankly one of the most debated topics on the Internet space and e commerce overall, which is agentic commerce. With me today we got Jason Goldberg, Sunset Jason Retail Geek Goldberg, Scott Wingo. Scott the serial entrepreneur will hear about their backgrounds here. If you haven't listened to their podcast, the Jason Scott show highly recommend it. As a dumb Internet investor or Internet analyst, not the consumer guys like you all are, I find it super impactful and thoughtful. So we'll get into everything here just to set the stage on what agentic commerce is because there's a lot of definitions, there's a lot of arguments or back and forth overall, the only thing I can really say is coming out of 4Q earnings across my coverage sector, the amount of companies that we spoke with that talked about greater overall traffic from ChatGPT and Gemini that are driving higher conversion rates is a theme that was the case across most every earnings. We heard that from ebay, we heard that from Airbnb, we heard that from Booking, you know, and the list goes on and on. And then on top of that, you know, I think Target's Talked about what 40% growth in traffic month to month from these, from OpenAI, from ChatGPT. Amazon talked about 300 million Rufus users, which is pretty impressive given Rufus just started. I don't, I'm sure everyone here uses Rufus. And then you know, Sparky, which is Walmart talked about 35% higher average order values which tells you that not only are we finding the search or the experience with agentic commerce or these agents more call it to the point, we're finding it higher, converting as well, adding to basket sizes and everything else. So anyway, this is Google builds out their data. OpenAI is announcing all things every, every single day. We'll get into that with ChatGPT. Meta is now into agentic commerce, which social commerce is a big deal for Meta and of course Walmart and Paul the expert here. So before I get into to the questions, I don't know Paul, if you want to say something or no, just
- [03:31]
D
thank you guys for, for being here. It's a, it's a hot topic in retail. So thank you for, for doing this and thanks Ron for coming. We'll probably have you back next year.
- [03:41]
C
I mean, I'm ready to go. This is a test. This is way better than the tech conference in New York.
- [03:45]
E
The three of us normally get invited to the kids table. This is like our first time at the adult table.
- [03:49]
F
Yeah. So we're excited.
- [03:50]
E
We're pretty excited.
- [03:51]
C
All right, so let's get started here. Maybe we'll kick off with introduction just to level set here. So we'll start with you, Jason and Scott. But as you talk about yourself, maybe tease us with in next holiday season what should we be looking for from call it agentic or just commerce perspective. So Jason, we'll start with you. Tell us about yourself, what do you do, what gets you excited and then talk about commerce.
- [04:14]
E
Yeah, yeah, yeah. Thrilled to be here. Thank you so much for inviting me, Ron. I'm Jason Retail geek Goldberg. I'm a fourth generation retailer. I've been doing this for a super long time. I used to joke about how old I am and my wife has told me that you are that old. That's not a joke. But today I lead commerce strategy for a big marketing services and advertising company called Publicis Group. So we're Walmart's agency of record and many other big retailers and brands. So I get to work across a broad spectrum. And back in the dark ages, I helped a bunch of retailers launch E commerce before they were able to hire competent people. So places like Blockbuster Entertainment, Best Buy and Target Perfect. And where are we going to be next year?
- [04:56]
C
Yeah. Or make it one to three years, whatever. What's.
- [04:59]
E
Yeah, well, what I'm predicting, the future is always hard and we record it on the podcast and at the end of the year Scott gets to make fun of how violently wrong I usually am. But. So I think the safe version of the future is disruption is getting dramatically disrupted right now. Right. And so what changes in the transactions and the shopping experiences like there's a lot of room for variation. But I think one thing that's certain is a lot of people are going to start their discovery voyage with an LLM more so than a traditional search engine or potentially even then social.
- [05:37]
C
Okay, I would agree. And we're starting that already. Perfect. Scott.
- [05:43]
F
Yep. Thanks for having me Sidi. And really enjoy the conference so far. First of all, Jason's very humble. He didn't say he's the chief Digital retail Store strategy Tactical Payments Store Officer. I think it's 10 words in your title. I always forget the last three.
- [06:00]
G
Sadly I don't get paid by the word, but yeah.
- [06:02]
F
Yeah. And then it's in his contract every year he gets to add a word so you can tell how long Agentic's coming up. You heard it here. Flash news alert. I'm a serial entrepreneur from North Carolina area. If my name sounds familiar, it's either from our podcast or I started a company called Channel Advisor. Started that 2001. We took that public in 20002013 under the awesome stock symbol E COM E C O M which I think is way better than Shop. Shop is boring. Ecom is rules.
- [06:32]
G
And we were which stock did better?
- [06:34]
F
Shopify, sadly Shopify. But our name was cooler. We were well known for publishing same store sales data where those largely for our customers so they could kind of see what's going on with ebay Amazon on their marketplaces. What we did at Channel Advisor is we helped brands and retailers sell across all the big marketplaces. Started on ebay, then we added Amazon, then Walmart, Target and then TikTok, Shop, Mercado Libre, all the Chinese marketplaces, so on and so forth. So I'm kind of the marketplace person in our partnership and that company got sold to private equity and then merged with another company and rebranded Rhythm still exists today. R I T H U M. I'm not responsible for that terrible brand. And then just started a new company called Refi Buy and that stands for Research, Find and Buy. And at Challenger we had a set of problems we could never solve in that age of Compute that we can now solve using underlying agentic commerce technology. So we're helping brands and retailers understand agentic commerce, expand out their product catalogs and use that as a way to do better and get integrated with the various engines. So I spend a lot of time marinating. This is 100% of my focus. And yeah, I also publish a lot of thought leadership on this under the brand Retail Gentic. So it's retail and agentic minus the A jammed together. So I may reference that. And what, what, what got me really excited about agentic commerce is I think it's changing consumer. This is a consumer conference. Consumer behavior is changing more in the last six months than it has in the last 10 years from an E commerce perspective. And it's because consumers love this new discovery mechanism. It is the best way to research products and helps you accelerate the purchase funnel. And I think we go from research find to buy and that all gets jammed up and it's going to, it's going to increase E commerce and we're going to see an acceleration of E commerce versus traditional retail over the next. This is my, my prediction. I think you'll start to feel it holiday 26 and then holiday 27 and 28. It will really start to pull away and it'll feel kind of like you know, 2006, 2007, hopefully not 2008, 2009 again we'll forget those years but it'll feel like this like huge new growth opportunity for overall retail that lifts all the boats.
- [08:53]
C
So I want to maybe, maybe we'll take high level now. And Jason, you and I were emailing a few weeks back, put you on the spot about E commerce growth overall. So let's level set up. Everyone knows about retail I guess and consumer. Let's dive in. What are you seeing on the growth space growth aspect of just E commerce in the US we're slowing down quite a bit. Scott says is true that would be tremendous for the broader environment. But talk to us more just level set where we are in E commerce penetration growth rates.
- [09:21]
E
Yeah, well so for the last 10 years average e commerce growth has been about 15% a year. All of retail including E Commerce grows about 3.9% a year. So. So e commerce over the last 10 years has grown 4x brick and mortar. The law of larger numbers is catching up. Depending on how you count. There's a bunch of data sources out there but it's 20 to 25% of all consumer spending is now E commerce.
- [09:46]
G
Last year the rate of growth of
- [09:47]
E
e Commerce slowed considerably and this year the government data will come out, I want to say next week for the final, for the full year, but it's going to be around 5.6%. So it's way, way lower. It's still 2x brick and mortar, but the rate of growth has slowed down a lot. It's not shocking given that it's one out of every four dollars spent. But what's super interesting when you double click on that number is the bifurcation. Right? So the whole E commerce industry in the United States is growing at 5.6%. Amazon, by far the largest E commerce site in the United States of America, is growing and this number is actually harder to suss out than one might think, but is probably growing about 11%. U.S. e commerce.
- [10:30]
G
Walmart, which is the second largest E
- [10:32]
E
commerce site in the U.S. is growing at 27%. So you've got the two dominant players blowing the industry average away. And that leaves really challenging circumstances for folks like Target that are, what were
- [10:47]
G
they, negative 1% or plus 1%? They're very close to flat in E commerce, which like not very long ago you would have said like if you have a pulse, you're going to be
- [10:58]
E
growing faster than that. So there really is a winners and losers situation playing out in E commerce right now.
- [11:04]
D
Can I chime in? Maybe some companies are saying that all of their future growth will come from E Comm. Some companies say retailers will say stores are back. Do you look at Agentic as being something that's really going to drive E Comm to be an accelerating growth vehicle for pretty much everybody across the board? Who's right in that debate? E Comm driving the growth versus stores are back.
- [11:32]
E
So both are probably true in a way.
- [11:35]
G
Right. So the vast majority of organic growth from all retailers, almost all retailers is coming from digital.
- [11:42]
E
And so if you suss out the
- [11:44]
G
Walmart numbers, brick and mortar's probably flat if you look at same store sales and then they have this huge E comm growth. There are other data points that show consumers have a higher preference for stores than they did a few years ago. Not a preference versus E Comm but store traffic has really declined during the pandemic, obviously. And in some categories it's rebounded. It's a little uneven, but I would argue the role of that store has fundamentally shifted. And I'm not an anti store guy by any means. I grew up in stores. I love stores. Super important part of any retailer strategy. And whenever you see a retailer announcing they're closing all their stores to focus on E Commerce, you could put a fork in them because that brick and mortar is a super important part of brand building and awareness. I don't think it's true that the brick and mortar stores are the growth driving engine for everyone. And what's super interesting is one of the doors that E Commerce opened up is a new revenue stream for these retailers, not just the sales they sell because but nobody can find any of the products on the E commerce site unless there is an ad. And I'm in the advertising industry. So we named it Retail Media Networks. And one out of every three dollars of earned income at Walmart now comes from these earned these, these retail media networks At Amazon, the Retail Media network is a larger contributor to earned income than AWS is. Right. So for these big retailers, they have this new revenue stream and you go talk to any CFO of a retailer and they're like, wait a minute, I've been getting 3% gross margins on bananas and now I'm getting 85% gross margins on eyeballs. Let's sell more eyeballs. And so the advertising business has become a super important driver of retail growth. And that's really only unlocked because of all these new digital experiences.
- [13:40]
C
It's a good segue into. And Scott, maybe teasing on your comment. In 26 we'll see an acceleration 27, 28 might be back in the heydays when E Commerce was real. I want to unpack that understand a little bit more. But I want to do that in the context of the news last week from OpenAI. Yeah, right. And so we went from call it October, November ish. Walmart joined OpenAI and next thing you know we're going to be buying everything through this chatbot to now we're not going to be a merchant of record. So talk to us more about how you view that change for OpenAI, maybe describe it for the audience. I don't know if everyone's on the same page with it and then how that might impact that acceleration in 27, 28.
- [14:22]
F
Yeah, like all things, the press hates to unpack it and it just becomes a runaway headline that agentic commerce is dead. Which Jason may agree with. I don't know actually, I haven't asked him. No, he doesn't. Okay. And so, so if you look at research, I think the data is pretty clear. Consumers love using the LLMs for research either on site with like a Sparky or Rufus or offsite on a chatgpt. And then the finding is where research is like, okay, I Was talking to one of the folks here and she's going on a bridesmaid trip and she's like, I wish I could. I need to use it for planning that trip and then figuring out, you know, what, what kind of things to get for that trip. And it's great for that. It's great for. I used it for a coffee maker and it answered this question I couldn't find an answer to in the store or online, anywhere else in like two seconds. So it's really good at helping you research products. It's good at helping you find products. So I would argue even if we use Sparky and Rufus, I use them exclusively versus the search engine because it is significantly better. It's better because not only does it have more product data than the search engine and it flows. Filtered nav was the big innovation that I'm sure Jason had a hand in in some way. But it's now and it was very innovative and now it's kind of broken because it just like falls in. We've learned as consumers to say I shouldn't put too many keywords in because it's gonna. Like everyone's had this experience. Like the more I tell Google, Walmart, Amazon about what I'm looking for, the less relevant result. There's a tipping point after four keywords and it's just like, it'll pop up like, you know, a purple hello Kitty croc. And you're like, I don't know where it came up with this random thing, but that's what I've got. And then the same is true for fil. If you add three filters on there, you get like nothing. And so I think that, that helping you research and then actually find the individual products within the research is really there. And what we don't know as an industry and we're trying to figure out is how do you connect the transaction? And ChatGPT is trying to be the most merchant friendly person to come at this. I've helped, you know, literally thousands of merchants sell on Amazon. They are not merchant friendly. They're like, we're Amazon. You're going to use our payment mechanism, you're going to use our fulfill. If you send a single email or an insert in your package to a consumer, we will, this is like, they're saying this to like execs at Nike and Under Armour and stuff. We will kick you off the platform immediately and there is zero tolerance for that. And they will just, you know, they will just kick you off immediately and you know, you can ask Things like, I want to see the purchase journey. No, we don't reveal that. The answer is always no. ChatGPT is trying to be an alternative to that. Now I would put all the, all the agentic commerce, one way to frame this is like it's all the world against Amazon in a way, in my world. So it's almost like oddly, there's a way to say it's ChatGPT, Gemini, Google and all those folks trying to create a viable alternative to Amazon's dominance. They're just trying to figure out how do we get this purchase to happen in a way that's merchant friendly. The first try was something called instant checkout. And they also have another attempt, which is interesting, which is chatgpt apps. There's also a viable way is you can actually just, for many retailers, you can just inject an order. It's called cart to cart. There's a lot of different words for this, but you can kind of start an order somewhere and then inject it into the site. And I think that's going to be like this third thing that they're going to try in this world. They all have a feature that the merchant is a merchant of record, which you don't get in the old marketplace world. On any of the marketplaces, TikTok shops, all that stuff, they're always merchant of record. In this new AI agentic world, the merchant is merchant of record, which is nice. But what I think Tattoo T. I'm kind of reading tea leaves here. I think merchants started saying, well, I really would like my loyalty program plugged in and I'd want all their purchases to come in through a login. And if you start listing out all the things you're going to have to implement in your checkout versus just use the other one, suddenly it starts to tip in favor of that. So I think it's just a reaction to what merchants are saying. And also not a lot of people are using because it's so handicapped right now. It's one item per cart. I don't think a lot of people are using it to buy and they're looking at the consumer behavior and merchant feedback and they're just pivoting a little bit. I don't think they're, I don't think they're shooting it. They explicitly say in that information article, we still think we are going to really make big innovations on discovery. And that's where we want to focus is kind of like right above the purchase part of the funnel.
- [18:42]
C
There's a big debate at least from my section, my sector is this Google 2 point is evolutionary or revolutionary. In other words, if this is just a find and you know better filtering because you can write natural language and then you find it and you go out and buy it with loyalty, that sort of doesn't disrupt. I don't think that disrupts the current environment as much as potentially ChatGPT is a front door to the web.
- [19:05]
F
Yeah. And this is not a refinement on things we've done. It's a totally better way to. And it gets better over time because it learns all about you. It's way it learns it is revolutionary, not evolutionary. Yeah. It is completely disruptive.
- [19:19]
E
I might split the difference on that. It is disruptive. There's a very disruptive progression during our lifetime of retail. For I mentioned 4th generation retailer for thousands of years the greatest source of leads in retail was sis. Saw in store, went to the store to get Oreos, discovered Coke flavored Oreos, which are nasty. Don't try to cook flavored Oreos. And so the retailer that was best at shopper marketing, at helping people discover stuff in store won. That's how it worked. And then e commerce happened and Saw in store got replaced by Google. Like what? SEO, Search engine optimization. What people typed into Google became the most important source of leads for that e commerce site versus saw in store at the retail site. Today the most important source of weeds for digital is actually not Google. It's social. It's TikTok. Right. And that's not talked about as much, but it's the most economically meaningful right now. And I would argue each of those
- [20:18]
G
was a game changer, but it was an evolutionary game changer. If that's a contradiction from saw in store to search to social. And I think social is very clearly going to LLM to prompts. And so I think that's all happening. There's another thing that very easily could happen for some kinds of purchases for the foreseeable future. The most you're going to use the LLM for, for picking your car is probably hey ChatGPT, help me evaluate these two cars and the robot will do a much better job of helping you pick models and compare them. It's very possible that people are going to do something different for toilet paper. Hey ChatGPT, never let me run out of toilet paper again.
- [21:00]
E
And the robot's just going to pick a brand, calculate my consumption rate and keep my pantry stocked with toilet paper. I don't know if people actually put toilet paper in their pantry, but you get the idea. Right.
- [21:09]
C
Don't want to run out.
- [21:10]
E
No, no, exactly. And so if certain purchases ultimately go to zero click to human not present,
- [21:18]
G
that's a complete game changer. And so I would call that the revolutionary change. I would call the migration from TikTok to ChatGPT for, for discovery a huge change, but still a more linear one.
- [21:34]
D
You think the consumer is ready and wanting that service?
- [21:39]
G
No, but I'm pretty sure the consumer wasn't ready for Uber when they, they launched too.
- [21:43]
E
But I noticed we all got here via, via Uber today. Right?
- [21:47]
G
So it's, it's not even the first inning of this game yet. There is absolutely not trust in the robots. But I'm. When I launched E commerce at Blockbuster in 1993, I got called into the
- [21:57]
E
principal's office and the principal's name was Visa. And they're like, hey, you're trying to do something illegal, you're charging a credit card, but the credit card isn't here, card not present. We don't do that. And we had to all invent a new transaction model.
- [22:11]
G
Right. And those same conversations are happening right now. Human not present. And so it's the first inning. This stuff is all still getting figured out. There's absolutely not trust. The consumer's not ready, comma. It would be a bad hypothesis to say, therefore this is foolish and it's never going to happen because I think it is going to happen.
- [22:30]
C
So consumer will go along, depending on trends, et cetera. But why is Amazon not perfectly positioned to win here? Because you've got. Every two weeks we get our toilet paper and our paper towels and now our deodorant Essentials is a core driver of that growth.
- [22:46]
F
They are. But from game theory, it makes sense for them. So remember, Jeff Bezos studied game theory primarily in college, and the game theory is before weightlifting. No, it's different. It's like. Yeah, it's that one movie where he like does all the. Yeah. Anyway, the, the. Yeah, basically your strategy, if you're, if you're already taking share and owning this thing, you don't want to help a competing ideology.
- [23:11]
C
Right.
- [23:12]
F
So. So their strategy is it will always be a walled garden. Out, nothing in, everything out. And that's what they've done. They've put all their data in that walled area in the fortress. And then they have. Rufus is the primary thing. And Rufus first half of 25 was not good. And now it's excellent. They've invested. You can tell they've invested a lot on that. And it's not just the user interface. There's several layers they had to invest at to get that, that humming like it is. And then they have a lot of experiments they're doing. So they have one which is where they have a lot of brands. Birkenstock famously has always said, we'll never sell on Amazon. And there's other brands, there's like a knife one everyone talks about, you probably know it. And so that you can actually now theoretically they started with permissioned, where if someone said, okay, I want to be in this program, you could shop from Amazon and then use Go. Actually kind of throw it into your prime account and shop from Amazon off of Amazon. And then they have an under permissioned one as well. So they're doing a lot of experiment of what if they want to actually. Because the one argument that they can't win is the selection argument. Because there's they, they don't have all the selection. They need to maybe open up the search engine to be all the Internet. So, so that's kind of like what their strategy is right now. At some point, if it did tilt, I think they would change that strategy. Just like we saw Google for a long time was like, oh, these, these, these answer engines, nothing to see over there. We have our own LLM, it's called Gemini. And then suddenly everyone's like, wait a minute, you know, the ChatGPT usage chart started going asymptotic. And then suddenly Google is like, we've actually integrated this in. We have AI overviews now we have AI mode and we have Gemini. And now look, it's, you know, they renamed it. What was it called before? I can't remember.
- [24:55]
C
Bard to begin with.
- [24:56]
F
Yeah, that was Bard.
- [24:58]
C
We're going back to February 23I think was, yeah.
- [25:01]
F
Picturing a medieval guy like singing a song. Yeah, the ode to Google. And then, so now they've had to, they had to, they had like the same game strategy, like, you know, do nothing, protect the fortress. Then they had to like do something.
- [25:16]
E
Yeah. So I think they might win. They have a lot, like, they have a lot of core assets, but they don't exclusively have them. Right. So if you think about it, when you do a search in Google, you type 2.6 words as Scott referenced.
- [25:29]
G
Right.
- [25:29]
E
And if we all did the same
- [25:31]
G
search in Google, we would all largely get the same results. Like they're very nominally personalized. On average, you type 28 words into a LLM per turn. And yes, the. Forget even the, the the that you might do multiple iterations. The LLM's answer is not just based on those 28 words, it's based on every piece of context the LLM has ever learned about you in your entire history. Right? And so the results are completely personalized to the circumstances and are different for every person in every shopping occasion, every single time.
- [26:05]
E
And so that context is enormously valuable. And Amazon has the most context about how you buy stuff digitally in the United States of America today.
- [26:17]
G
They actually don't have the most context about how much peanut butter you use. Because while they sell the most E commerce, Walmart sells a lot more groceries than Amazon groceries 1 out of $4 of consumer spending. And so there's an argument Walmart has more shopping context, Amazon has more digital context. That's going to be a good horse race.
- [26:36]
F
But one other thing is they don't know your fitness regimen. They don't know.
- [26:40]
G
Well, that's what I'm getting at. So the reason ChatGPT might win and not Walmart or Amazon would be because they're horizontal. Amazon and Walmart you only go to after you already know you have a shopping problem. So we actually call them vertical agents. Marky and Rufus are vertical agents. You already have to. You decided you were going shopping and then you went to Amazon and then you use Rufus, the horizontal agents. The robot that you might use to help make your whole life better is the one you're going to say, how do I stop this leaky toilet? What could this rash be like? All of these random questions that you're likely to ask in your life. And so over time, ChatGPT is getting, or anthropic or whatever your LLM of choice is, is getting this voluminous body of context that a pure vertical shopping engine is not likely to get. So I think it's really dangerous to pick a winner right now because you've
- [27:35]
E
got digital shopping context. Amazon has the lead physical shopping context.
- [27:41]
G
Walmart has the lead horizontal shopping context today. OpenAI has the lead, but they're rapidly losing it to Google and Anthropic.
- [27:48]
E
So it's, it's the wild west of all of this right now. And one argument against Amazon is they potentially are going to be the victim of innovators dilemma here. Like there's a, there's sort of a sweet irony like Amazon's model was launched
- [28:04]
G
based on Walmart's margin as their opportunity, right? Like that these things that are not profitable for Walmart to do could be Amazon's business model. And Amazon was quite successful at that. So now when the horizontal LLMs started wanting to sell stuff when ChatGPT said, hey, buy your stuff from us. Amazon's like, absolutely not. You're the enemy. If they buy it from you, they're not buying it from me. I'm blocking all the bots.
- [28:27]
E
ChatGPT can't know anything about the Amazon ecosystem. They have innovator's dilemma there, right? They have an existing business to protect. Walmart said, huh? If you're going to buy something from the LLM, you probably weren't our customer. You were probably Amazon's customer.
- [28:42]
G
So you know what I'm going to do? I'm going to take my 400 million products and I'm going to list them on ChatGPT and Gemini. And so ironically, there's probably a bunch of giddy execs at Walmart going, Amazon's margin is our opportunity.
- [28:56]
E
And they were the victim of innovators to whim. And now they potentially are the beneficiary.
- [29:02]
F
An example of this horizontal ification that's happening, it just happened in like the last six months. And I know Kroger's coming up. We're a big Kroger household and I watched this in the first three quarters of last year, grocery wasn't on the list of what people were using ChatGPT for. And then it flipped in the fourth quarter. And what happened there is people realized it is amazing at meal planning. So you basically say, I'm a family of three, I live in Chicago, our son likes this, he has a nut allergy, I don't, I have a gluten thing. It remembers all that so you don't have to enter it over and over and over again. And then you also can say, here's like our staple meals, we love having a taco night and pizza night. Those are non negotiable. And, but we, you know, I'm tired of doing meal planning, I need three new ideas. So then it'll like run you through some recipes, you pick those and then it says, then it will actually offer up. It's like, great, here's your meal plan. Do you want me to create an ingredient list? And you're like, well, heck yeah, why not? And it's like, I know you go to the local Kroger here. Would you like me to just kind of put it in a shoppable list for you? And you're like, wow, that's amazing. And it does that. Or it'll say, great, you want me to just shoot it over to Instacart for you? And I've done that and it just goes. And it loads up your instacart and it'll say, I remember the last time we had this conversation. You like your avocados slightly squishy, yet still firm. There's like an avocado thing with instacart that they struggle with. And it'll, it'll tell it, it'll put that in for the shop. Do you want me to add the shop? And people are like, this has just eliminated like three hours and all the pain of my life. And so you've seen this just this just use case just went way up. Amazon will never know that Walmart, all the retailers, the grocery people have tried this over. They've tried a million ways to get into recipes and I don't think you may know more than I do, but I don't think it's really, really stuck. So it's really this interesting horizontal use case that, that is driving a new adoption path.
- [30:55]
D
So you wouldn't think Sparky is a good way to kind of achieve that same goal.
- [31:00]
F
I've never tried to do meal plan. I'm going to guess Sparky isn't great at that.
- [31:04]
E
There is an interesting.
- [31:05]
F
So again, Jason will know the actual
- [31:07]
G
first inning of all these things. They all are not awesome today. They all are as bad as they'll ever be today though, right? The pace of change is happening so fast.
- [31:17]
E
It would be like silly to bet
- [31:20]
G
against them being awesome at this stuff.
- [31:21]
F
Right?
- [31:22]
G
But there is an interesting dichotomy between Rufus and Sparky. Sparky is a very. I'm sorry, Rufus is a very vertical shopping engine. Like it really is optimized for shopping questions. And as soon as you get slightly broader in use case, Sparky doesn't want to play there right now. Now Amazon may or may not stay in that lane, but that's the lane they're in today. It's very focused on recommending products to buy Sparky. Walmart's on site agent is closer to a horizontal LLM. It will not answer as well as OpenAI but it will try to answer as broad a set of questions as OpenAI will. So I would actually argue at the moment, like complex meal planning tasks that Rufus would fail at. Sparky will make a better attempt. And again, is either one going to be a home run today? No. Are they evolving very rapidly?
- [32:17]
E
Yes.
- [32:18]
C
Do we know who's powering Sparky from an LLM perspective? Is that OpenAI?
- [32:22]
E
So we don't officially know the but
- [32:26]
F
Jason knows is going to tell us. Just don't disclose all the trades.
- [32:31]
G
Circle of trust. The thing to know about Walmart is there's this funny but true story I tell all the time that I attended a meeting with Doug McMillan like two and a half years ago at Walmart and he, he asked this scary new question at the beginning of the meeting. He's the recently retired CEO of Walmart who's younger than me, so I'm not
- [32:48]
E
bitter about that at all.
- [32:50]
G
But he, he said like, hey, thank you guys very much for coming to this meeting. Before we get started, quick question. How'd you all prepare use AI to
- [32:57]
E
prepare for this meeting?
- [32:59]
G
And it was like the most awkward thing I'd ever heard, right?
- [33:01]
E
Like everybody's totally quiet. It's a big deal to get in
- [33:03]
G
a meeting with Doug. Nobody says anything for what seemed like eight minutes. It was probably like 10 seconds. And then some brave soul chimes in, Doug, we didn't use AI to prepare for this meeting. AI is illegal at Walmart.
- [33:15]
E
And he was absolutely true. The IT department at Walmart had said,
- [33:20]
G
hey, privacy issues, hallucination, you guys aren't
- [33:22]
E
allowed to use any of these tools. And we can make fun of Walmart for that. But that was true of every corporation.
- [33:27]
G
It was certainly true of my company. And in the subsequent two and a half years Walmart, largely led by Doug, went from AI is illegal to we're going to buy ChatGPT for 1.8 million Walmart associates. And then some knucklehead goes in and goes, just so you know, I have a better AI in my pocket than you do in all of Walmart because I have all seven LLMs and you
- [33:48]
E
only have one and a different one
- [33:49]
G
is better every month.
- [33:50]
E
And they're like, you know, you're right,
- [33:52]
G
we're going to buy every associate, all eight LLMs. And then they're like, and we better invent our own because there's things that are going to be unique to us. So you can imagine that the engine driving Sparky is an amalgamation of in house technology. And Walmart should not be underestimated as a technologist. They have 15,000 full time engineers at Walmart. It's an amalgamation of in house technology and licenses that they publicly have with OpenAI, Anthropic and Google. And they were on stage at the National Retail Federation. New CEO of Walmart, John Furner and
- [34:29]
E
Sudha are from Google.
- [34:30]
G
So you can imagine they have a very friendly relationship with all the alums at the moment.
- [34:35]
C
And then the relationship that was just codified with OpenAI and Amazon might suggest that that very verticalized shopping engine might, like Rufus, might actually be more horizontal if they adopt OpenAI.
- [34:48]
F
Yeah.
- [34:48]
C
Although I don't know.
- [34:49]
E
I don't know either. I'm a little cynical. So, first of all, Amazon has a relationship with several alums as well. And you know, obviously Jeff's not super active in Amazon, but I think he has a very large personal investment in Anthropic. If I'm not mist, they can.
- [35:01]
F
So does Amazon. Amazon has like a way more in anti.
- [35:04]
G
Sure.
- [35:05]
E
Okay, so. So they. They have a bigger.
- [35:07]
F
The.
- [35:07]
E
The announcement of OpenAI in my amateur read of the tea leaves is more about AWS that. That AWS wants to sell all of this compute. There's an increasing amount of compute that's being driven by the LLMs. And so Amazon wants to play aws, wants to play in people that are using open AI tool sets. Conspicuously, there was no mention of commerce in this partnership whatsoever. And at the moment, Amazon still very aggressively blocks all the OpenAI bots from crawling Amazon's website.
- [35:40]
G
And they have a very large, what
- [35:42]
E
we call black hat team whose whole job is to not let any of these external agents steal data from Amazon. Sure.
- [35:50]
C
I've got a ton more questions, Paul as well. We have 13, 14 minutes left. If there's any in the audience, please, you know, ask away. This can be as interactive as. As we'd like it to be.
- [36:01]
E
Anticipating the big question. No, I don't think there's a dessert
- [36:05]
C
or we don't get lunch. Yeah, you know, we're. We're the entertainment here.
- [36:08]
E
So, you know, wow, they got ripped off then.
- [36:10]
C
Yeah. Yeah, great. Well, think about questions, you know, raise hands, whatever. We more than happy to do that. I wanted to maybe pivot a little bit more and talk about the advertising side of it all. And so we've been talking about how Amazon sort of like the walled garden. Not sure the horizontal from OpenAI is what it is, but you know, Google has had product listing ads in a carousel forever. Amazon comes in and out of those and Amazon won the product search experience in call it the mid 2015s.ish. So help us understand where. Where do you think? Call it this new horizontal approach, whether it be Gemini or OpenAI or ChatGPT or whatever or Claude. Why can't Amazon play a bigger part? Why can't ebay play a bigger part? Why is this not this Google 2.0 from an agentic engine optimization perspective on the ad side or on the ad side? Yeah.
- [37:08]
F
Okay, so part of our podcast shtick is we typically take opposite sides of an argument and he's Mr. Advertising Guy. So the doomsday I'll be the bear. I'll paint the bear scenario and we'll see if Jason can do the bull scenario. The bear scenario is I'll just pick on Amazon. But this is for all retail media networks. Amazon's ads business is 60, 70 billion, give or take 10 billion between friends. There you go, 68 billion. And so if a fair number of eyeballs move away and agents are communicating with agents over a three or four year period, repeat, you know, the retail media network margin stream and revenue stream is extremely at risk. And that's that, that's a small percent of Amazon's revenue, but it's a very large percent of the EBITDA and same for Walmart and everybody. So that's the doomsday scenario. How likely is that in one year? It's not likely. I think there's a picture in three or four years where it is maybe conceivable.
- [38:09]
C
So, so that, that's on the Amazon. Oh, and O advertising huge business. Amazon is also a huge advertiser on Google and Google is a big director
- [38:18]
F
of traffic to Amazon and they're increasingly adding people to a network. So they also. Amazon is now becoming a network player as well. I'll defer to my advertising colleague on network questions.
- [38:28]
E
Yeah, so caveat.
- [38:30]
G
We like to be contrarians and take opposite sides. He's actually needling me because there's 110,000
- [38:36]
E
employees at my ad agency publicist and I'm the least ad centric one of those 110,000 employees.
- [38:42]
G
So it sucks to be responsible for
- [38:44]
E
advocating for the ads here. Just, you know, like we use this metric in advertising called return Roas return on advertising spend and that. They'll teach it to you in any basic marketing class. The R doesn't stand for return and the S doesn't stand for spend. Like it's a very dubious metric. Right. And the, the magic question is if I turned off all those ads, what
- [39:04]
G
would happen to my sales?
- [39:05]
E
Like we really don't know.
- [39:06]
G
Incrementality.
- [39:07]
E
So huge grain of salt with all of this stuff. There is a dirty secret. Amazon buys $18 billion of eyeballs every year from Google and then they sell
- [39:20]
G
some products, pickleball rackets in the case
- [39:22]
E
of Scott to Scott and make money on that pickleball racket. And then they turn around and sell that eyeball to Procter and gamble for $68 billion. Right. So it's the greatest Eyeball arbitrage business on the planet by 18 billion in ads from Google sell make money on those eyeballs and then sell those eyeballs at a huge profit. Increasingly. My hypothesis is Amazon doesn't need Google for that.
- [39:47]
G
Right. Like, the more people that get in the Amazon ecosystem, they bypass that Google product search and start their product search directly on Amazon.
- [39:57]
E
Right. So we're. If LLMs were never invented, you would say, like, man, a significant portion of Google's advertising revenue is at risk as
- [40:07]
G
Amazon builds this direct relationship with every consumer and disintermediates Google.
- [40:13]
E
And if you don't know, there was a very large unexplained experiment last year. Amazon turned off all $18 million of those ads for a considerable period of time. And they've returned some spending, but not at the same spending level as before. And so that's pretty interesting. The problem is all of those people that might have moved from Google to
- [40:34]
G
Amazon are now leaking to all of
- [40:37]
E
LLMs that we talked about. Right. And so, you know, the originally, like, Sam Altman was like, I hate ads. Like, like, I would never do ads.
- [40:46]
F
Right.
- [40:46]
E
Ads are like, inherently untrustworthy and we want the robots to be trusted.
- [40:51]
G
And you could follow his interviews and
- [40:53]
E
see the progression as he goes.
- [40:55]
G
Actually, there are some kind of ads that are kind of cool. Right.
- [40:59]
E
And then he hired this woman, Fiji Sumo, who was the CEO of Instacart and built the ad network at Facebook.
- [41:05]
G
And today you can all buy ads
- [41:08]
E
that will show up in commerce on chat gbt. Right. And so the it, for better or worse, advertising is a proven economic model. It's a good way to monetize, like, all of this interest. That is another competition at the moment for that eyeball between Google, OpenAI and Amazon. And in this case, of course, Google is the most vulnerable with the innovators dilemma because they were making the most money. Old model.
- [41:36]
C
Right, right. So TBD. But OpenAI is coming on strong.
- [41:40]
E
Yeah.
- [41:40]
C
Amazon could also participate.
- [41:42]
E
I'd be. I'd be following it.
- [41:43]
C
Yeah. Maybe switching topics a little bit, getting more techy.
- [41:48]
E
There's been a. Oh, thank God, because Scott's breaking into a rat.
- [41:51]
C
Let's talk about protocols. So Google has launched their version of their protocol for agentic commerce called universal commerce protocol, I believe.
- [42:01]
E
UCP.
- [42:02]
C
UCP for the cool kids. And then OpenAI has ACP. And so what? Talk to us about how the battle for these protocols, like the importance of these protocols. Sundar was on stage at the NRF unveiling this Gemini is coming on increasingly Strong. I want to understand fast forward a year from now when we go into 27, the acceleration potentially how these protocols create a bigger moat, if you will, for Google and OpenAI or and maybe bring it back to this crowd here. What does that mean for Walmart and Target and Best Buy and everyone else? You need to be on those protocols.
- [42:40]
F
So a lot of times the press likes to frame it as a, you have to pick one and that's it. But they're very open. So what these both companies did is they basically published them, they push them into open source. So unlike old school protocols like edi, I guess EDI ultimately got there, but it was like this closed thing and kind of had to be pushed to that. It's been born open, meaning it's been basically contributed into a foundation. It's on GitHub. This is super nerdy. But you can go to this huge code repository called GitHub, I didn't brand it, and then you can go look at the code and then you can modify it and you can like, you know, and it's in there, it says if you want to build something against this, here's what you do. And then you can go start building and it'll typically give you a prototype to get started.
- [43:24]
E
You can do that, but none of us should do that, right?
- [43:27]
F
Well, it's kind of fun. But yeah, yeah, you can do it on the side. And so they're kind of an implementation at the end of the day, they're an implementation detail and they're very similar in many ways. Google's kind of trick here is they kind of, they let the first one go out and they could come out with a better one. So it's objectively better in many ways. It's more open, it works well with other frameworks. It has a plan for how it's going to do. Loyalty, multi cart and a lot of the things that look like ChatGPT has gotten hung up on. And so that's kind of where we are right now.
- [44:01]
C
Okay, so you need the protocols. Is it important for the cataloging of, for these LLMs Data is so important, you know, knowing what's on sale, what's not on sale, what's in stock, what's not in stock.
- [44:13]
E
The thing I would say, and I'll let Scott pile on, the reason that the protocols are important is because right now the robots are trying to learn about every product that's available for sale by crawling websites. And websites suck and they don't have the right data. And Scott, if you Read his stuff. He finds all these funny examples of
- [44:32]
G
where you went searching for women's shoes and you got a banana or something,
- [44:36]
F
because actually men's shoes.
- [44:37]
E
Okay, yeah, I only skim your article, Scott. And so the data you can get
- [44:44]
G
from this scraping is the retailer's original data sucks. And then you scrape. It's like telephone game on top of telephone game of bad content. The robots need better data. And the protocol is a way to define the structured data that the brands, the product manufacturers should be using to inform the robots about their product. And that is super valuable and super important.
- [45:07]
E
And so whether you embrace any particular protocol, and I tend to agree with
- [45:11]
G
Scott, modern technology, it's kind of easy to concurrently support all these protocols. I mean, I think that's a service refi by would. Would quickly provide you. But the, the real magic here is forcing the discipline of when. When we sold Oreos to Walmart, we. We had a fax order with three attributes about the Oreo. When we launched an e commerce website, we asked Mondelez for 30 attributes about the cookie, and they exploded. How are we going to get those 30 attributes now? The robot wants to know 3,000 attributes about the cookie. And that requires structured data and a protocol.
- [45:45]
F
Yeah, yeah, the protocols have. A product data element is a big pillar of that. And in our world, it's kind of funny, we use product data feeds to do this. And this is actually part of the spec is they have the feed spec, and at the end of the day, it's basically a CSV file. It's kind of embarrassing. It's 20, 26 and we sling CSV files around, but that's part of what's inside of there. And then it's. And then it spends the bulk of its time on the orchestration of the checkout, which is really complicated because you have a lot of things going on there. You have PC this, credit card compliance, the thing Jason got almost put in jail for at Blockbuster, you have this like, PCI compliance rule. You can't just put cards over the open Internet or they get stolen. So the credit card people will get all upset about that. And rightfully so. So, so most of it, this conference
- [46:33]
G
is being put on by a credit card company, just so you know.
- [46:35]
F
And city credit cards are great. Everybody phenomenal. Yes. Yeah, I got. That's part of our. We have to say that. But we love our Citi cards and the. So three quarters of it, a quarter is kind of product stuff and three quarters is. Okay, here's all the things that we all know, can go wrong in a checkout. You get to the end and the thing, the Nintendo Switch got swiped out from underneath you. What happens? How does it fail gracefully, that kind of thing. So a lot of it is in there. And then Google went a little bit further and they said, well, what happens in the return process? Let's put some loyalty information in here. How do we package up more of these little intent nuggets and pass it back to the merchant? They added a little bit of flavor in there, but the protocols share more than they're different.
- [47:17]
C
Got it. So we have about two minutes left. I don't know if there's any questions the audience or Paul, if you've.
- [47:23]
F
Sure.
- [47:23]
D
I mean, we've talked about. You talked about Walmart, talked about Amazon, I guess. Can you maybe just talk about in an agentic world, if there are other retailers that you're thinking of that are particularly well positioned or poorly positioned that this audience cover mostly retail, might be interested in hearing.
- [47:37]
G
Well, an inside baseball one is going to be Kroger, right? Like they're the largest specialty grocery store in the country. They sell less groceries than Walmart.
- [47:46]
F
But the.
- [47:47]
G
And I would argue they had sort of lost their way and the proposed merger was highly distracting and all these things. So they just hired a new CEO, Greg Peran. Greg is the greatest grocer in the world and largely responsible for the grocery growth at Walmart. And he left Walmart to run New Zealand air a week before the pandemic,
- [48:06]
E
which is not a good career move, it turns out.
- [48:08]
G
And so now he's back in grocery at Kroger. Don't know what he's going to do. It's going to be interesting, but I suspect he's going to be a lot more pragmatic than the previous management about figuring out this. So I'm going to be watching Kroger closely. They also have a lot of consumer spending context and they're closer to Scott's house than Walmart is. Right. So I think that's interesting. I do think what is interesting is all of this stuff really favors the biggest people in the ecosystem that can make the biggest investments. A lot of other retailers are just struggling with the basic resources right now to like afford a TikTok presence. And so they're not developing a robust agentic strategy. And so I think more so than previous disruptions, this disruption seems like it's favoring the incumbents.
- [48:54]
E
And so I'm really curious, does a new competitor emerge?
- [48:58]
G
When digital happened, Amazon and ebay emerged. Who's going to be the new competitor that competes in digital and anthropic and OpenAI are certainly viable candidates, but I don't know.
- [49:12]
F
One thing we haven't talked about is there's a primordial soup of startups and they tend to be highly verticalized. So there's like two in grocery storage, there's like 50 in fashion. I ran across my first auto parts one and they're trying to say I'm going to use an LLM to have the best fashion buying experience. And because I'm focused on a vertical, I can, I can have a better experience than ChatGPT. You know, today they're not there, but that's there. And these are some of the most famous, you know, founders in our, our world. Who's the specifics lady I always want to see?
- [49:41]
G
Yeah, Katrina.
- [49:42]
F
Yeah, yeah. So she has one called Daydream, so everyone's watching that one. And of course she went out and raised like $75 million in a pre seed round kind. There's some Chinese founders that have gensmo, so there's a lot of really interesting things going on there. So far they haven't really popped from a consumer standpoint where they've shown up on the data. So that's interesting to watch. The categories to watch are beauty and fashion. That's where a lot of the action is because a lot of the generational adoption of not only ChatGPT, but using it for shopping skews younger. So it's millennial and Gen Zs and they're, they're really starting to light it up. And the other thing we haven't talked about is going back to the Amazon era. There was always this pull of Amazon, which is like this cool new marketplace you could sell on. But for the first time we have this interesting push. I talked to specifically these fashion retailers and their traffic from Google is down 20, 30, 40% right now because Google is drying up clicks. So we're a lot of people in our world worry and what keeps them awake is a future with no clicks. What does that look like if no one's clicking around anymore? And what that means is if I go to Google and I just get the answer, I lost the traffic. It never came to me. I never got a shot at inserting my widget I wanted to sell. Or they go to ChatGPT or Anthropic and I never see them. What does that world look like? We're starting to see it in the data where the, the clicks that you used to get, especially the free ones. The SEO clicks are just going away very, very rapidly right now. And that that's a big, big trend
- [51:22]
C
that, that changes a lot of things on the overall web. Well, we are in overtime here. I want to say thank you very much, Jason and Scott and Paul for, for having us. And yeah, thank you. This was wonderful.
- [51:35]
F
Yeah, thanks. So and just just as a reminder,
- [51:38]
C
we still have open.
- [51:39]
A
You've been listening to the Jason and Scott show for all the latest news and trends on e commerce and shopper marketing. Subscribe to us in itunes or visit www.jasonandscott.com.