Subscribe on

Enjoy this episode? Leave us a review.

All Episodes

Episode 119

When AI is Your Buyer with Jonathan Elfreich

AI agents heavily rely on structured data like pricing and star ratings while largely ignoring flashy visuals or banners. Traditional SEO strategies may actually harm your chances of being recommended by AI systems.

This week, Elena and Rob are joined by Jonathan Elfreich, Head AI Architect at Misfits and Machines, to explore how AI is changing marketing. From SEO to GEO optimization to AI-driven TV advertising, learn what marketers need to know about preparing for a world where machines make purchasing decisions.

Topics Covered

• [04:00] How AI differs from automation and learns from data

• [09:00] Why traditional SEO strategies harm AI citation results

• [14:00] Building brand memories in AI systems like ChatGPT

• [18:00] How well-known brands have advantages in AI recommendations

• [21:00] Short-term changes happening in TV advertising with AI

• [24:00] Long-term vision for personalized, generated TV content

• [28:00] The importance of targeting and mass customization

Resources:

2025 AdExchanger Article

Today's Hosts

Elena Jasper

Chief Marketing Officer

Rob DeMars

Chief Product Architect

Jonathan Elfreich

Head AI Architect

Transcript

Jonathan: The way we interact with TV will actually change. You'll go from watching a show that was pre-programmed to literally being the programmer yourself.

Elena: Hello and welcome to the Marketing Architects, a research first podcast dedicated to answering your toughest marketing questions.

I'm Elena Jasper on the marketing team here at Marketing Architects, and I'm joined by my co-host Rob DeMars, the Chief Product Architect of Misfits and Machines.

Rob: Hello, hello.

Elena: And we're joined by a special guest, Jonathan Elfreich, head AI architect at Misfits and Machines.

Jonathan: Yeah. Hey, thanks for having me guys. I'm really excited to be here.

Rob: It's always fun to have another misfit on the podcast. It's super exciting to have someone who was basically responsible for bringing generative AI to Omnicom, a total Star Trek nerd, and also probably a lesser known factoid about Jonathan, a skilled swordsman.

You probably can't see this if you're listening, but if you're on the YouTube, you can see even in the background some of your amazing swordsmanship. Talk to us more about that. How do you go from sword wrangler to AI?

Jonathan: When I was a kid I always promised myself I wanted to be a Renaissance man, and I took that real literally all the way down to the swords. So even when I was a kid, I was basically like, nah, I wanna know a little bit about everything. And I was living out in the rural part of southern Indiana and it turns out there's not a whole lot to do around there when you got a bunch of friends except for beat each other with sticks. So that slowly evolved over time into my very first fencing club. I've always just kind of had a passion for it ever since. I do fence a very weird form of fencing actually.

So instead of more Olympic fencing, I do a version of fencing called HEMA, which is based on historical European martial arts, like two handed long sword fencing. But I'm just a huge collector of swords nowadays. I like to use my brain, not have it hit as much anymore.

Rob: Well, swords, AI, Star Trek. You are a misfit, so thanks for joining us.

Elena: Well, we're really excited you're here today and we're back with our thoughts on some recent marketing news, always trying to root our opinions and data research and what drives business results. And today, Jonathan is joining us to talk about artificial intelligence and marketing, the future job of a marketer, how the way we buy from brands will change, how to prepare for that and more. But first I'm gonna kick us off, as I always do with some research. I chose an article from Ad Exchanger called "Marketing to Machines: A New Performance Strategy in the Age of AI Agents" by Michael Lehman of Nativo. The core argument here is that marketers need to shift their focus not from people to platforms, but from people to machines with AI agents like OpenAI's Operator, Google's Gemini, and Amazon's Rufus stepping into research and recommend products for consumers.

The traditional funnel is breaking down these AI agents, they don't scroll, click, or view banner ads. They read, synthesize, and return what they believe is the best answer. That means ranking number one on Google isn't enough anymore. You need to be embedded in the answer itself. That is driving a move from traditional SEO to GEO generative engine optimization.

It's about creating structured semantically rich content that AI agents can easily parse and trust from. Product pages, Q and A's, reviews and comparisons. According to recent research, these agents heavily rely on structured on-page data like pricing and star ratings, while they largely ignore flashy visuals or banners. Lehman's key takeaway is this: in the age of AI marketing is no longer just about grabbing human attention. It's about shaping machine perception.

If you want your brand to show up in AI driven decisions, you need to become a trusted source, credible, clear, and consistently optimized for how machines think. So that's a lot to take in. Luckily we have Jonathan here to help us break that down, and we're gonna do that in a moment. But before we do, I have what feels to me like a stupid question, but I want to ask it anyways, which is Jonathan, how do you define AI? Artificial intelligence? Because sometimes I think marketers, we might confuse automation with AI and automation is still important, but those are two different things, right?

Jonathan: Yeah, they are two very distinct fields. They're related to one another though. But I think it's especially for marketers to understand the distinction between automation and AI. And I promise I'll loop around and answer your question about how I define AI. So automation is about executing kind of a series of predefined rules. It's like setting up a bunch of dominoes or a Rube Goldberg machine, and then you tip over that first domino and the rest just happens all in sequence, right? Automation is the domain of traditional programming, if you will. You set up a Python script or something from top to bottom and you run it and it just runs all the way through and whatever you program to happen happens. AI though is fundamentally different. It is used to handle complex, uncertain situations, and I think the key thing to understand about AI are that they understand, they learn and they're adaptable as well.

Unlike a traditional program, which somebody would sit down and orchestrate from top to bottom, an AI system actually has to learn from data. So to circle back to your first question, how do I define AI? It's good to look at history actually for the answer to this. And if you look back in the 1950s or so, there was a researcher, his name was Frank Rosenblatt. The guy was a genius way ahead of his time. He invented what we considered the very first AI, it was called the Perceptron, and it was trained using real data.

Rob: The Perceptron woo,

Jonathan: The Perceptron for real, a great name.

Rob: Sounds like a transformer, like the toy, not the AI tool.

Jonathan: Well, I mean, what a surprising connection, right? I mean, Perceptron definitely does feel like he could be a Decepticon or something.

Rob: Yes.

Jonathan: But no, for real, he builds this thing, he calls it the Perceptron, and he actually, he doesn't program it. Instead he shows it real images of men and women, and it just outputs a score. It just outputs men or woman, and when it gets the answer wrong, he presses a little button and it retunes all the knobs and dials inside the AI's mind. So the Perceptron, the very birth of AI, was actually born not by programming a bunch of traditional rules, but by trying to teach a computer to learn those rules from data.

What we've seen happen is in 2017, a paper came out that connects back to what Rob was saying. It was a paper called "Attention is All You Need." And it introduced a new, evolved version of the Perceptron, if you will, called a transformer model. And this model was able to learn incredibly well, it was able to read the entire internet and basically memorize sections from it. And so it has a loose memory or knowledge of everything on the web. These modern AIs do. And when you train it on so much more data, you get these weird emergent capabilities.

Basically what's happening is as you train one of these systems on more and more data, new abilities emerge. So ChatGPT is only really designed to predict the next word in a sequence. So if I go, "it's raining cats and..." Your brain automatically completes that with dogs, right? That's what ChatGPT is designed to do. It's designed to get the sentence that says it's raining cats and predict the word dogs, but just by training on so much more data from the internet, this AI system has learned to reason about math, logic, solve puzzles and other things.

The way I define AI is as any system that learns from data, and we classify these AIs into narrow AIs that can only do one or two little tasks like the perceptron classifying men and women or picking out handwritten digits. Then we're getting what we would call more general intelligences. These transformer models have these emergent capabilities. It can code an entire Python program from scratch.

An AI is a system that learns, but once that system has learned, you can actually use it as a part of automation. So you can have a little automation flow that uses AI as a critical step.

Elena: That's really interesting and I think helpful for marketers to understand some of the context of this stuff. And sometimes it can seem overwhelming and scary, but just having that background I think is helpful when we're talking about learning new things. Coming back to that article we opened with, SEO. It's a big part of most marketers' jobs.

And I know just personally for me, the thought of it being upended is scary because you're trained as a marketer, like, this is how SEO works, this is what you do, here's step by step. And it feels like we don't really know where SEO is going, or at least a lot of marketers aren't aligned on it. So where do you think the future of SEO is going? Is this article, is it more clickbait or is the future going to be fundamentally different because of AI?

Jonathan: One of the things that I like to think about whenever I'm looking at any kind of stuff like this is both history and science fiction. I take a lot of inspiration from shows like Star Trek, sci-fi books like H.G. Wells and others that I read growing up, Isaac Asimov. And so I think that history is a great teacher of what's happened before, and sci-fi is a great look into what's happening in the future. Like Rob said at the beginning, I'm a huge Star Trek nerd.

I love that show. It has predicted the future so many times it's not even funny. I think Star Trek predicted how people will use computers in the future. People don't go type on the computer at all. They just go up to it and go "computer, how far away is that asteroid?" and the computer does all those calculations for you. That is the type of interface we're seeing emerge right now with all these generative agents. People can go to ChatGPT and use their voice and just go, "Hey, I want to go find a bike, what is the best bike for me?"

And then ChatGPT is going to go search the web, recommend results to you. And that is a fundamentally different paradigm from what we all grew up with. So I think it is very important for most marketers to understand that the history that we've all seen of things changing in the past. We've seen changes from radio to TV, from TV to the internet. These types of changes are happening again, and they're happening to SEO this time.

If you just do this experiment, if you go out to a grocery store and you get 30 different flavors of jelly beans and you set them out on a table and you go, "please sample my 30 different jellybeans," very few people will stop. But if you only give them three options, if you go, "Hey, sample my three jellybeans," many more people will stop. And it's because the more complex, the amount of choices somebody has to make is the more cognitive load it is effectively to do that task. And what AI really fundamentally changes about this paradigm is that cognitive load can now just be done by the computer itself.

What's really interesting is, as marketers are thinking about the future of SEO, they need to be definitely thinking about these GEO generative engine optimizations that you can perform.

I've got a buddy of mine, his name is Ian Armstrong. We worked together back at Omnicom and he's currently with a firm called Perplexity and they've been doing a bunch of studies on GEO optimizations as well. And what they've found is that traditional SEO strategies actually harm your results whenever you try to get agents to perform recommendations. They did a large scale survey, over 50,000 prompts to OpenAI, Claude and Gemini testing the relationship between backlinks and AI citations, and what they found was actually astonishing.

It's that AI has an inverse relation to the number of backlinks you actually have on your site to the amount it will cite you. So it means that having what the article that we're discussing says, having a highly fragmented site with tons and tons of backlinks to your site actually will harm how likely you are to be recommended by the AI. So as people are designing their websites, which will be ultimately consumed by these AI people need to be fundamentally rethinking the search approach because more and more traffic over time is going to become these AI systems.

I think it's important for most marketers to consider the fact that there are kids being born today that will never live in a world where ChatGPT doesn't exist. And it's going to just like how TVs and smartphones and everything else got better over time. The same thing is going to happen here. So I really do think that the SEO game is fundamentally changing.

Elena: As a consumer, I find it really interesting. As a marketer, I find it terrifying and very uncomfortable what you're saying. 'Cause everything I've heard before from marketers talking about AI, they've said traditional SEO will help you. So as long as you've got your traditional, your foundation already, AI should be able to find you and clearly that's not, that might not be the case for people. I have a feeling that your brand will still matter a lot, just for consumers when they're making a decision. You know, if AI comes back with a bunch of options, you're probably gonna pick the one that you're familiar with. But if traditional SEO strategy is not gonna help us here, what do you think marketers could do right now to start to prepare for this future? Say, I am already a brand. I know you said you can build a website differently, but say I'm already a marketer, I have a brand. What's something that you think we could do today?

Jonathan: I actually would agree with you. If you think about how brands grow, being the top brand that pops in people's minds when they hit a category entry point is probably the most important thing you can do. If I'm about to have a big party and I know it's gonna be messy, Bounty has spent a huge amount of dollars making sure that their ads pop up in my head and I think of Bounty the quicker picker upper, or when I'm at the store, I pick up some.

They want to be the top thing that shows up in my head when I have the problem. That is the same challenge that we're faced now with AI. Because if you think about what you're saying, sure, if your brand exists and a user asks a super specific question about your brand, it'll probably be able to find you in the search results. But if the user's asking a more general question like, "I'm about to have a party, I need a list of supplies," you really want your brand to be top of mind in the model's mind. It's crazy to think about a little bit how much this is about to change. But it also isn't if you consider history. Like humans, we evolved to process language and the only other source of language we ever had was other humans.

So we evolved the circuits in our head that process language to communicate with one another. It means that when the computer writes in language back to us, our brain interprets it the same way as a friend's recommendation. Our brain doesn't have a built-in filter that goes, "oh, this is AI. It's slightly different." You kind of actually have to train yourself on that a little bit.

And because it doesn't have this filter, it means that a recommendation from an AI is almost, or if not better than a recommendation or word of mouth suggestion from a friend. So like you said, being a trusted brand to the model's mind is very important, but another thing to consider is you can actively influence the model's mind. You can actually go in and set these things called memories or preferences to the model that actually change its behavior slightly. So Rob, you have a ton of memories in your ChatGPT, which directly affects its behavior every day. I'm sure you've seen that.

Rob: I'm just becoming more and more attached to it.

Jonathan: Exactly. If I was a marketer and I were faced with this challenge, I need to optimize my GEO search. I would want my users to be making a campaign where they made memories with ChatGPT about my product or my service. That way, whenever I run into the category entry point, if I have a memory in ChatGPT that says, "ah, he prefers Bounty paper towels," when I tell it, "Hey, I have a party," it is much more likely to recommend for me a product that I have stored in memory as something that I like.

So I think a lot of marketers have to think, SEO is going to evolve. It's going to change from SEO to GEO and the strategies that we use will be different. But if I was a marketer today, that is what I would be focused on for most of my mind, is how do I get my user to make a memory with ChatGPT saying that they enjoy my product.

Rob: That's an amazing, it's 'cause it's actually not even an analogy. You're talking about memory, literally, you know, we talk about consumers' memory and recall, but now you're talking about AI's memory and its ability for recall. Literally not even metaphorically. That's really interesting.

Elena: It seems like, if that's true, what you're saying, so marketers gotta shift your mindset to how do I create memories? With things like ChatGPT, it feels like brands that are already well known, like Bounty is being typed into ChatGPT a lot more. It feels like they're gonna have an advantage. And I also wonder if now is a really important time for things like advertising and helping your brand be better known on a different level. 'Cause if they're learning, you're thinking, if you're a brand new startup right now, it must be more difficult to have someone be proactively searching out your brand in a language model, large language model.

Jonathan: So if I'm running a startup right now, what I'm trying to probably be thinking about doing is getting my brand to be well known to as many consumers as possible. I would immediately start trying to set up websites which are not deep in terms of links, but rich in terms of actual content about your product.

I would also make sure that you have some type of reviews section, because AI seems to really prefer human reviews when making product suggestions. And then I would launch campaign to become a well established brand and I would even focus the campaign around "talk to ChatGPT about us."

Elena: It's a fun call to action.

Jonathan: Yeah, because I mean, honestly, AI is becoming a strong recommendation engine for most folks. They trust it. They feel like it understands them, their preferences, what they like, what they don't like. So if I'm getting a suggestion from an AI who I have this relationship with, all of a sudden it matters more. And I think, Rob, you can even attest to this, just purely by having a memory in the model's mind that you like or dislike something, it's likely to bring it back up again. Even sometimes serendipitously in moments you're not necessarily expecting. So brands can actually use that strategically if they're able to get their consumers to make memories and talk to ChatGPT about their brand. But yet, you should be attempting to get as many people to talk to these models about your brand, to get them to form memories and to index higher in the results.

Elena: And if you're currently dependent on non-branded keyword search and Google, you probably need to shift your strategy to start thinking about more brand building. Speaking of brand building channels and campaigns, the perfect transition into this next topic. You came to Marketing Architects to help in one way disrupt television advertising and bring these new capabilities into TV. So can you talk a little bit about what you think the future of more AI driven TV advertising looks like?

Jonathan: If I'm to break it down, I'd actually try to think about it across three different time horizons. Short term, how is AI affecting TV today? It already is in TV today. I can tell you, I sat down with my wife and we were watching a little TV. We were watching like some old Star Trek and the commercial came—

Rob: You watch Star Trek.

Jonathan: Yeah. Yeah. Just a little bit. Just a little bit.

Rob: Lines, but from memory, I would imagine, do you reenact them?

Elena: Rob, we don't need to get on—

Rob: Okay,

Elena: Stay focused.

Jonathan: I think TV advertising in the short term is actively undergoing a change. Because when I'm watching TV I see commercials and I'll recognize a voice. It won't be a human's voice that I'm recognizing. I'll be like, "oh, I know that model's voice. I've worked with that voice model in ElevenLabs before," and so I can, just by sound sometimes identify and be like, that commercial is a generative AI commercial. And today, we are seeing these types of commercials join into the TV space where there's maybe some voiceover and kind of disconnected scenes. That's what we're seeing today. Short term, it is reality. These ads are going out and most consumers can't even tell that it's AI. Unless you've worked with the voices in ElevenLabs, a whole bunch. It can be hard to tell medium term. This is in a couple of months. If you look at models that have been released by Google recently, like Veo three in a few months, those will be released to the general public via API. Veo three can do things that are genuinely mind blowing to me. It is able to produce up to eight seconds of very coherent video with accompanying audio. So instead of having a bunch of scenes with a voiceover on top of it, you are about to see a another phase shift in a few months where it'll actually be fully generated shots with dialogue between two characters talking about your product, but it'll be short spots.

They'll be mostly limited to things like fifteens and thirties. And the reason for that is that these video generation models, just like any model, has this thing called a context window. That context window can only get so big before the model starts breaking down. It's kinda like trying to hold too many thoughts in your head. If you have too much going on in your head, you just eventually you get so bogged down, you can't keep going. So the models work similarly in that regard where they can only output so much content. In the immediate future, you're about to start seeing fully generated ads where the scenes themselves contain dialogue, complex animation and other things.

And that'll be coming probably by the end of the year. Long-term TV I think is going to fundamentally change. And tools like Veo will not be producing eight second clips. They'll be producing up to a movie's worth of content, two hours of content directed relatively seamlessly for our users. So what I imagine will happen is the way folks actually watch TV will change. Instead of you plan a show out and you do scripting, and then you do casting and all these other things, people will generate shows just on the fly.

People will throw in prompts and be like, "I want an episode of Star Trek: TNG, but set in today's time, right, keep the same actors and I want them to encounter the characters from Seinfeld." People will be able to actually do that. So I think what you're going to see happen is the way we interact with TV will actually change. You'll go from watching a show that was pre-programmed to literally being the programmer yourself. You give the TV a prompt and it actually generates the show for you. And the nature of advertising in that domain will be fundamentally different as well.

The thing that'll become much more common instead of having an ad break, will be product placement. You'll have your Propel or your Mountain Dew or something like that, just show up in the shot. As well, I think you'll see technology advance to the point where people can pause and control the TV purely with their voice. And these models are actually getting pretty good at interpreting what we call intent. So they understand what you mean when you say something. So I think the long-term vision of TV is that people will actually generate whole shows and movies where there will be ads, there'll be product placements and scenes.

Rob: Yeah, no, I love that. I mean, you can look at it through different schools of thought, people that are like yourself or dorks like me, who just love this stuff and kind of lean into it. And then there's other camps that takes early adopters and I think as marketers we're continuing to look at what levers we're gonna be able to pull when it comes to like, building teams and that same sort of polarity between adoption. Some people that are, it's taking a bit longer to jump on the bandwagon. Others who are just on a bleeding edge and annoying everybody else with their Star Trek talk because it's just so cool. You've talked in the past, I've heard you talk about different ways of promoting adoption. Sometimes it's just adopting the tools that are being thrown at you or other schools of thought say, boy, you should really be building things from the ground up. How do you look at that divide and what's the best way to negotiate that?

Jonathan: So that is such an interesting question. I have really seen, even at large enterprises, two large kind of differing schools of thought have emerged. Some folks they already have a product. It's been successful. They see AI, they hear a lot of hype about it, and what they wanna do is they want their system to have AI in it. So they try to figure out some way to just shove AI into a preexisting product. Maybe it can talk about your product or answer a consumer's question, but AI doesn't really control the experience. It's not actually the core thing. You've got your product and you've kind of just duct taped AI onto the top of it. That's one school of thought. Another school of thought is no, AI is actually a pretty big fundamental game changer.

If you go back, all the way back to the very first computers, we've been trying to figure out how to interface with computers and we've been, our very first approach was language. First we tried the language of computers, then we tried our language, but it had to be exact, and that's not how people really communicate. And now the computer can really understand what you're saying. So some folks have looked at that and gone, this is a game changer. The same way that like Windows was a game changer. Let's redesign a whole product around the ability for the computer to understand you. Instead of just being bolted on and answering a couple of questions. It is the guide, it is the way you interface with the site. It's the way people actually talk to your product. That's how I think these two schools of thought are different.

Rob: And Jonathan, we love, as you know, we love contrarian thinking on this podcast, and so do the misfits at Misfits and Machines. What's your most contrarian marketing view? And I bet it's a doozy.

Jonathan: It is a big one. I think people, you and me, we've got different preferences. We've got different beliefs being able to be spoken to. We actually need to hear things slightly differently. I had a great professor of rhetoric when I was in college named Dr. Kerns and he talked to us about how different people are influenced by either emotion or logic or sympathy with one another. And he talked about how different people need to hear different things rhetorically in order to have their minds changed.

And in the past it was pretty difficult to build a spot that spoke to each individual consumer's preferences, how they like to hear things. But that is about to fundamentally change the ability to rapidly generate images with coherent text is already pretty much solved by OpenAI, and the ability to do this for video is going to be solved as well. So if you know the psychological profile of your consumer base, you can actually much easier and much cheaper than ever before mass customize your creative to speak to that individual. You can take those different psychological traits that you might have about your audience, whether they like dogs or cats.

And you can literally fundamentally regenerate a commercial from having a cat in the scene to having a dog in the scene. Just because you know that the user prefers dogs to cats. So the ability to mass customize is going to emerge for a lot of folks, but using mass customization effectively means knowing your target audience incredibly well. So I think knowing how to target people, knowing how to speak to different mindsets is incredibly important.

Rob: Targeting's always been a spicy topic, so on this pod. So thanks.

Elena: I would say we're not anti-targeting at all as a company or a podcast. I think my concerns are when you over target AKA, it costs more to be worth it. Inaccuracies on who you're reaching. Like it really just comes down to cost. Like you're saying, Jonathan, if we live in a world where you can customize and make it cost effective, I'm all for it.

It's just, right now it's typically not worth it. And my ultimately other gripe with this would be, and there is power to cultural imprinting and having a lot of people see the same message about the same product, and knowing that someone else saw it. Reaching people outside of your perfect target. I think that's the other weakness of targeting. I think when you're saying, can we help customize a message to an individual,

I don't think that's necessarily a bad thing, but I think marketers often take that and they end up pinpointing, this is exactly who my customer is. I'm only gonna hit them. And then you're missing influencers and potential new audiences and like you're getting away from that broad reach. So I don't totally disagree with you. I feel like it's a topic with a lot of nuance.

Jonathan: Definitely. And there's a lot of takes on it. What I would say is, what you're saying is cultural touchpoints are huge, important things. They basically ground lots of folks together. What we are about to see, though, or at least what I would predict is that we are actually about to see fewer of those cultural touch points because the ability to make media is actually fundamentally going to change. So instead of a movie studio producing one movie, they're going to produce the basic prompt and lots of folks are gonna customize the endings and interact with it in a different way.

And actually, you're going to end up with kind of this fractal of thousands of different versions of that same initial idea. And I think that it is a big point of discussion is what happens when the cultural touch points, when people don't go to the movie theater anymore to watch movies, they instead customize their movie at home on their screen. How does that fundamentally affect how people are going to perform purchasing behaviors.

Elena: Yeah, it would be weird for sure,

Jonathan: Oh, definitely, definitely.

Elena: Let's wrap up with a fun question. We can lighten it after. I'm just thinking about our SEO strategy at the moment. It's consuming me, but let's wrap up with this. Jonathan, you can go first. If an AI assistant made all your purchases for a week, what is one thing you hope that it would buy and one thing that you fear it would buy?

Jonathan: Ah, well, one thing I hope it would buy would be my entire reading list. I've got a huge reading list from Chuck of all the books, so I would love to just give it that list of books and be like, go on Amazon and order me all of those books. One thing I fear it would buy is, it has a lot of memories of knowing that I wanted a bunch of H100s for compute, so I fear it would go out and buy an entire data center for me.

Elena: Yeah, that would really hit the finances, wouldn't it? Yeah. All right. Rob,

Rob: Well, Cheez-Its, I mean, I just can't have enough Cheez-Its if AI was buying me Cheez-Its all day long, I'd be a happy guy. I've got plenty of shelf space that I would make for Cheez-Its, and on the opposite end, ironically, we've been talking about paper towels a lot. I accidentally buy paper towels all the time. I keep thinking we're out and paper towels take up a lot of space. You know, you buy that big box on Amazon and I'm like, dear God, I accidentally ordered another one. I thought we were out. We weren't. I got a backlog. I need a public storage unit just for the accidental purchases of paper towels that I make. So no more, no paper towels, AI.

Elena: I actually asked ChatGPT, I said, could you like make purchases for me for a week? And first of all, I would say it definitely overspent. It spent like a thousand dollars in a week. So Jonathan, yeah, that concern's definitely real. It thinks I need a lot of stuff that I don't. One thing I love that it suggested was a massage, like, that's nice. ChatGPT. Thank you. And one thing I did fear it would suggest is it suggested I buy a bunch of baby clothes. I do not have a baby. I'm not having a baby, but I have a lot of friends who have had babies recently, and I've been using ChatGPT to come up with gift ideas, ask questions like, when should I visit them? What is it appropriate to bring food? And now I need to tell ChatGPT, I'm not having a child, but it seems to think so.

Jonathan: I wonder if it made a memory. You should go check your memories and see. That'd be such an interesting thing because maybe that's why it's making the suggestions to you.

Elena: It must be, it's like you're a new mom. Like I'm not, I'm a dog mom. But that's about it.

Jonathan: I'm also on that dog dad life. That is the way.

Elena: Nice. Awesome. Well, Jonathan, thank you so much for joining us. You're so smart. Loved the stories and I learned a lot, so thank you. It was entertaining and made me scared at the same time, so perfect, perfect podcast.

Jonathan: Thank you guys so much for having me. I really appreciate it.