AI Marketing and Business Context with Alex Greifeld

Confession and AI Dread

0:03[music] >> All right, thank you, Nick. Thank you. So, I'm going to apologize to Nick because I'm going to actually open this with a confession that might get me kicked off the speaker circuit as fast as I've been put on it. >> [snorts] >> I've been feeling a lot of existential dread lately. And I'm not usually this type of person, but like I've been finding myself more and more like sitting in my car in the Walmart parking lot like disassociating, wondering if my role as a marketer is even going to be around in the next 3 to 5 years, if all of my expertise and all of the work that I I've put into this industry has been for nothing because of AI.

0:58And this is the thing that really like broke my brain recently. I'm not sure if

Viral Bot Scam Story

1:04you all have seen this, if you spend a lot of time on X, but essentially this was a long article where this guy explained how he trained his open claw bot to create like six viral Tik Tok slideshows a day to promote his vibe coded app. And he made it sound so easy, like he was laying it out step-by-step, and his premise was, you know, anyone can do this, this is the future. If you're not pushing five vibe coded apps through slideshows every day, you're going to be left behind.

1:40And I was reading through this and thinking like like at a high level this makes sense, but there's something missing here. There's like a layer of actionable information that's missing. And I was kicking this around some networking groups that I'm a part of just like trying to understand this, and it turns out that this is actually a crypto pump and dump scam. He created a coin named after his claw bot, and his whole thing was like using this viral post to get people to invest in this shitcoin, essentially.

2:14So, there's a lot more to this stuff than meets the eye.

Who I Am and No Best Practices

2:19And before we dig any deeper into this topic, I just want to introduce myself. My name's Alex Greifeld. I've been working in e-commerce and digital marketing for close to 15 years at this point. in 2022, I started my own consultancy, and my focus is really helping brands answer one key question, you know, why did we stop growing, or why did the tactics that have always worked for us so well up until this point stop working? And then once that's identified, what are like the quick wins that we can get to reverse course, and then how do we fix like the the higher-level problems that are driving it? And I've worked with brands of every size. I've worked with one-person startups. I've worked with like 10-year-old bootstrapped brands, and I've worked with some of the biggest names that are recognized globally that are shown on this slide.

3:12And in 2020, I actually started a newsletter called No Best Practices, because at that time, I'm not sure how many of you remember this, but there was this kind of like a misconception floating around the e-commerce space where if I just find the right tactic, or like the right app to plug into my brand, it's going to like 10x my ROAS overnight, and somehow not also do that for my direct competitors. So, there was like a lot of hype floating around the DTC ecosystem at this time, but a lot of it fact and myth needed to be separated, and that's really why I started with this positioning.

3:54And that's what I want to do today. I want to take the no best practices approach to AI and investigate some of the biggest, most hyped myths that have been floating around about what AI can do for e-commerce brands.

Myth One AI Takes Jobs

4:08So, myth number one, like let's just eat the frog. AI is going to take your job. Is this true or is this false? Well, here's the truth. Mark Zuckerberg has said he wants to take your job. he said that he's coming for the advertising industry. He wants to create a system where brands don't need marketing teams anymore. They just roll up to Ads Manager, drop a URL in there, press a button, get the perfect campaign structure, and even the ads generated on platform.

4:38And for anyone who's actually been in Ads Manager over the past like 3 to 6 months, you know that this vision is far from reality, but it's good to know that that's the stated intention here.

Testing AI Strategists

4:50So, keeping this in mind, I interviewed four AI creative strategists to see who is the best candidate here to take my job. Because for the past 2 years, one of the main things I've been working on is I've been the fractional head of creative strategy for a protein bar brand called Stars and Honey. So, these are our four candidates. We have Manas, that is the AI agent that Meta itself recently acquired. I really like Manas for data analysis because it runs Python, but you don't need to know how to code Python to do it. So, it doesn't hallucinate on data analysis. Then we have Moby, that's Triple Whale's in-platform AI agent. I love Moby for getting into like the nitty-gritty of media buying analysis.

5:33There's Runneth, which is Motion's agent. And I think Motion works really great for creative analysis, especially integrated into like existing Motion dashboards and workflows that you have going. And then finally, there's Parker. Parker launched in January and it's not associated with any existing platform, but it's built with a lot of deep contacts about how ads work and it also has a Slack integration, which I just like personally really appreciate. So, this is what I did. I went to each of these four candidates and like, "Look at my last four days of perform my last 30 days of performance on Meta. Tell me what I should do to get a 1K a day banger like ASAP.

6:17And here's the business context behind the situation. So, the brand that I'm asking this question on their behalf and it was a client brand, it wasn't Stars and Honey. So, total sales were like hockey stick growth. Like they were doing great. But new customer sales, they were following that hockey stick trajectory for a while, but within the last like four to six months, still growing, but kind of like momentum starting to recede. Then ad spend that's like all over the place, like very choppy over the past six months.

6:51Sometimes they're really able to scale it up high, sometimes they're not. And AOV just steadily dropping over the past few years and then within the past six months like just hit this floor past which the brand is not willing to like be unprofitable essentially. So, what's

Why Agents Miss Strategy

7:09actually happening here is this brand has hit a ceiling in their ability to scale. They have a certain setup, it's a certain set of personas they're targeting, a certain set of messages and creative playbooks that they're running and even like a certain set of funnels or products that they're promoting and they can't put any additional money against this and also be profitable. So, what did my AI agents tell me to do? Essentially, they told me to do more of what was already working, which is exactly the opposite thing that we need to do to like break out of this plateau.

7:45So, the number one learning I got from this was, you know, AI agents are only as useful as the questions that you ask them. If you ask really zoomed-in technical questions, you're going to get really zoomed-in technical answers. And if your problem is further upstream than that tactic, it's not going to help you. So, I learned from that and I'm like, "Okay, I'm going to give these guys a second chance to replace me. I'm going to zoom out." So, I asked a different question and I also built out some context. I gave these agents sales data.

8:15I gave them more information about our brand and our goals and what we were trying to accomplish. And I said, "Given all this information, what should I do in the next 90 days to hit our sales goals for this year?" And these were some of the answers that I got. So, these are like all valid, but they're all kind of like floating around in isolation. and I'd say that introduce a higher-priced bundle to get your AOV up is like closest to the strategic solution that we need to pursue.

8:47But, the lesson I learned from this was that, you know, agents are and and AI in general, they can be great for synthesizing a lot of diverse sources of information and helping you like wrap your head around a situation, but they're less good for prioritizing and proposing solutions and like understanding the real sticky human issues that keep us like stuck in conference rooms arguing with each other for 30 to 60 minutes.

Build a Context File

9:15So, the learning from this exercise, you need a context file. if you want to get the best results out of AI, you need to give it this background information about your business, about what your what you're even trying to accomplish. So, what I like to include in a standard context file is, first of all, target customer profiles, ideally validated by data. Like who is it that you're selling to currently, who are you going after?

9:45Key product information, like what are your five to 10 best sellers? What what's in them? What makes them special? How do they work? Your financial KPIs and objectives, this is such a big one. So, not only what are your North Star KPIs for making media budget allocation decisions, but what are you trying to like accomplish as a brand from a high-level financial perspective? Is it pure growth regardless of profit? Is it a certain profit target? Is it an exit? Is it turning this into like a sustainable cash flow business that's all important?

10:21And then finally, you almost want a change log of like every major thing that's happened in the business that you think has impacted top-line revenue. So, this is you know, new product launches, this is price changes, this is new offers, this is new promotions that you've launched, all of that and more. So, if you want to get a sense of what actually is in one of these context files, you can scan that QR code. I built out a sample, it's for a brand called Gloop, no relation to Goop, but it I built it as part of a training exercise that I was developing.

10:57It includes instructions for what to put into each section of the file, and then sample context for this made-up brand.

Myth Two One Shot Ads

11:06Okay, so AI I would close the book on the whole idea that AI is going to take all of our jobs anytime soon. So, let's move on to myth number two. And that is the myth that you can one-shot hundreds of ads a day and flood meta. And this is actually true to a certain extent, but the fact is that AI simply makes it easier for a lot of brands to bit to break the law. You'll see a lot of these examples, 100% AI-generated. Some of them are meant to look like real people, some of them are are not.

11:43And people are like, "Oh, these ads are so powerful. Like AI is going to take you to the next level." But that's not the causative variable here. It's the fact that you no longer have to pay some disgraced doctor like $10,000 to star in a VSL. It's like, you can make one out of clay for $5. so it's making law breaking more accessible than it's ever been in the past. Moving on to a a legal application of the one-shot idea. So, this is a very cool video and it actually, I'm sorry that this is a a screenshot that says it has 12 views because it has a lot more than 12 views now. So, ignore that. It's this video actually did go viral, but I love Alex. He is like one of the most talented people in creative strategy, working today. And he went through the step-by-step process for essentially building your own static ads one-shot machine. And so, at first glance, this is like, "Oh, this is sick.

12:44I just push a button and I'm like turning out ads." But when you actually watch the video, there's a brand DNA document, there is a library of the ad formats, there's the human discretion required to pick the right ads for your use case, and then there's also a brand copywriting skill. And several of those pieces are like born out of Alex's own talent and experience that he doesn't give away. So, if you implement this system yourself and you're using it and none of the ads are working, like that's the secret hidden sauce that that you don't necessarily have.

Building My One Shot Machine

13:18So, I was curious about this whole idea of like building a one-shot machine. So, I decided to try to build my own. And it's very funny that this came up in the last Q&A section because this is the format I decided to focus on because from a technical perspective, it's relatively straightforward, but it's also applicable to a lot of different use cases. So, I thought, you know, if I'm actually able to pull this off, like this is going to have a a long useful life. And if you haven't seen this type of ad before, it's very popular with drop shippers and affiliates, but it is essentially designed to look like an organic Facebook post from a real person. It's a bizarro image that kind of like gets you wondering what you're even looking at.

14:04There's no text on the image, and then it's like a long story from a whitelisted page. So, even building the copy from that ad is not is not a one-shot. It's like four distinct processes. So, the first is big idea generation, and that is like the most difficult part of any copywriting process where you're trying to find a way into the topic that's going to connect with whoever you're speaking with. You can't just start pitching the product. and then there's the copy points and the story outline. So, like what essentially like what's the narrative flow of the story we're trying to tell. Then there's actually QC'ing the copy to make sure that it's something you'd put out there.

14:48And then finally, it's automating that whole loop. So, you know, you're walking into work on Monday and you have an inbox with like 10 ads that are ready to push live. So, for big ideas, I actually decided to scrape Reddit instead of trying to teach one of these agents to like become a creative thinker from scratch cuz I don't think I don't think I could do that. I think that's beyond my capabilities. for the copy points in the story outline, I had a few examples of these ads that had worked in the past. I used those as training, built out a few story frameworks.

15:22Same thing for writing and queuing the copy. If and I use Claude for this whole process because I thought it would be the easiest to automate once I got each piece. Claude is pretty good at, you know, if you just give it five examples of copy that worked, it can keep writing copy like that. But I also wanted to put together a scoring system to make sure that the outputs were like up to my standard. And then finally, when I had that set up, I'd use co-work to automate it weekly.

15:54So, this is this is like we're like really digging in the weeds here, but I think it's worth mentioning because even for the go find me my big idea piece, you're basically building software without code. You're like build you're breaking each of these pieces into even smaller pieces that you can kind of control the variables around. So, Claude doesn't actually have access to Reddit. So, first I had to build an MCP with Claude code to like even access Reddit in in bulk. That only actually took me like 90 minutes and it was my first Claude code experience. So, like that's a demonstration of how powerful these things are. then in terms of where do I where should I tell it to go look, I decided it would be best for me to say that versus leaving it up to the model's discretion because I wanted to use this for a variety of different situations that might change in the future.

16:50Then I had to tell it how to prioritize posts. Like which basically, you know, look for comments, look for upvotes, look for hot, best. Then I had to define scoring criteria because Claude can't like synthesize the content of these posts while it's looking for them. So, you have to pull down like 500 or 1,000 and then ask it to score it on some various criteria. I think the good example for weight loss transformation is it would get me a lot of stories of people who lost dramatic amounts of weight through like a bariatric surgery and that was like too extreme for the the kind of thing I was trying to promote. So, I just screen that out. and then you resume you review the outputs and you're like, "Okay, am I happy with this?" If not, go back to the beginning and like tweak each part of the loop. So, it it really is like programming and I have like some long-lost background in programming which is why I can say that. But,

17:47things to keep in mind before you embark on one of these one-shot journeys, like one of these one-shot quests, how repeatable is the process? What's the lifespan of what you're trying to build? And do you know enough about why this works to build something functioning? I think the the long text post is a great example because it's very repeatable. it does seem to be like a pretty evergreen format that's worked for years and I personally knew enough about why our winners in that format were winning.

18:21All right. So, yes, it is possible to one-shot ads, but it's not easy.

Myth Three Volume vs Quality

18:26Myth number three, volume is more important than quality. it doesn't matter how good my ads are because I'm going to spam Meta with 500 static a day or whatever it is. So, yeah, this can actually be true to a certain extent, but if you do it, you are destroying the enterprise value of your brand. And AI advertising at scale can be very effective. I mean, we've seen it be effective. It's It's almost impossible to ignore if you're spending a significant amount of time like researching, going on on ads libraries, whatever.

19:05you just can't get caught because AI outputs can be more persuasive than human outputs. But if people realize that you're trying to fool them, obviously they're no longer going to trust your brand. Great tie-in to Sarah's talk right before this one, but especially for Gen Z, if they catch on to the fact that you're trying to misrepresent, you know, the outcomes that someone can expect from using your product or fake testimonials, it will immediately create distrust and they'll turn away from you. And I want to draw a distinction between things that are pure entertainment like Fruit Island and things that are more meant to be deceptive. So, people are happy to watch AI as entertainment. That won't necessarily ruin your brand reputation.

19:54You just can't make promises that you you're not able to deliver on.

The Brand Horseshoe Framework

20:00So, the brand economy in 2026 is not K-shaped. It is a horseshoe. And on one end of the horseshoe, you have your drop shipping demons and they're Yes, my favorite meme to emerge this year. they're using AI to accelerate their ability to break the law. and they're maximizing for like 3 to 6-month MPV. They're not trying to take those brands and sell them to Procter & Gamble in like 5 years. Then on the other end of the horseshoe, you have what I call real brands, brands with a capital B. And yes, they're on e-commerce. Like yes, they're running meta ads, but that is not the only theater of war. they are focused on building great differentiated products.

20:49They're focused on positioning. And they're also focused on creating real-life touchpoints to make the brand something more than just like a Facebook ads brand. they're they're trying to build cultural capital, build brand heat. So, these are the two types of brands that are going to be the most successful in the next 3 to 5 years. At the top of the horseshoe, you have the confused founders. And they're are the ones who are going to struggle the most. So, these are the founders or the brands who can't pick a lane between capital B brand and AI acceleration.

21:25And they tend to get really tripped up in this kind of like technical minutiae like versus CBO, cost caps versus bid caps versus highest volume. Should we be running a promo offer on our email capture? Like, yes, all of those questions can be meaningful, but they're not the foundation of what you're trying to accomplish. So, this actually happened while I was putting these slides together. Like, this wasn't in V1 of this deck. And both of these events happened within 10 days of each other and absolutely proved my point. Dropshipping Demons, you have MedV. It's like a one-man AI-powered startup that reached nine figures. and the the way he did it was by powering an army of affiliates so that he could have plausible deniability of all the ways they were breaking the law to grow his company.

22:16And then on the other hand, you have real brands. You have Grums who exited had a nine-figure exit in like less than 5 years.

AI Amplifies You

22:25So, I want to reframe the conversation around AI. AI is not making me obsolete. It's not making anyone in this room obsolete. Instead, AI is amplifying me and it's amplifying you, but it's not just amplifying what you're good at. It's amplifying your shortfalls and the things that are holding you back almost from a personal development perspective. So, the benefits of AI compound with inner work. The more that you're able to identify those patterns that are like keeping you spinning around in circles or preventing you from achieving your full potential, the more positive impact you're going to get out of AI. And I'm sorry if this sounds like a Tony Robbins like self-development nonsense, but it it's really not.

23:19AI forces you to confront what you're afraid to admit because if you can't tell it what you really want, if you can't give it a solid outcome to shoot for, then it's not going to be able to give you meaningful feedback. And that's where the context file piece of it is so important. >> [snorts] >> Yes, AI reduces cognitive switching costs and it reduces busy work and it makes it easier for one person to accomplish a lot of different things in a day, but if you're working on the wrong things, it's just helping you run around in a circle faster.

23:52And then finally, AI draws a binary between a brand and just like a collection of things that one guy or girl likes. And this is one that I've encountered a lot in in my career because I I've spent a lot of time working in the fashion and apparel space. A brand can be delegated. you can turn out as many statics and videos as you want, but if they're bottlenecked by the approval of one person who can't give feedback that another person could actually learn from and get better from, like if it's all just vibes-based feedback, your production capacity doesn't matter.

Toolkit Plug Better Ads

24:30So, the epilogue to my personal AI story is I feel that I have replaced myself successfully to a certain extent. I really dove into these tools and when in a lot of my client projects, honestly, like the what they hear out in the marketplace is you got to make better ads, bro. And that is like the most condescending and unactionable advice it drives me nuts. but I actually built out a toolkit of five different workflows and prompts that help brands figure out what make better ads means for them and how to execute on it because honestly, as much as it pains me to admit this, the way out of a lot of your growth issues is actually to make better ads.

25:18but that means a different thing for every person in this room. So, I call I called this package of prompts and workflows, creative strategy for operators. And this is my little plug that I'm just tapping into this presentation. but if you want to learn more about that, you can I have it pinned to the top of my X profile. my handle is heyitsalex p as in Peter. Or you could just come up to me after this and ask me about it.

25:46plug is done. These should be your top

Six Month AI Priorities

25:49priorities for the next 6 months if you want your brand to get the most out of AI, especially in the lead up to Q4. So, the first is become what you are. Think about where you sit on the horseshoe. Are you closer to the drop shipping demon or are you closer to the capital B brand? And you also have to think about what you're willing and not willing to do both from a marketing strategy and from a financial perspective to go into either of those lanes because building a brand with all of these touch points outside of Meta, like it can be more capital intensive than, you know, running a direct response brand with Meta as your only growth lever.

26:30the second is you need to define where your context as an individual and as an organization is a unique advantage versus where it makes more sense to go out and rent context. So, thinking back to like the one-shot ad generator, I think that's the perfect example where the average agency owner is going to have more context on what ads work and don't work versus the average brand owner. So, like, does it necessarily make sense for each brand to be building their own one-shot ad generator?

27:01Probably not. The third is to build your core context file for the your brand and even if you have to bring in what I call a context therapist, which isn't a real role that exists yet, but it's kind of like a combination between an executive coach and a consultant. Like someone who can bring your executive team into alignment about exactly what it is you're trying to achieve from a financial perspective, from a growth perspective, and how you put KPIs against that.

27:34And then the fourth, the most technical, is to just pick one task at a minimum to automate each month. and it can be something as simple as like inbox triage or train clo- Claude to write like me. But it's not it's not just like the act of accomplishing these tasks. It's like then you have to step back and try to almost explain to yourself why you chose the given approach and think about like is there a way I can make this more more efficient because the best AI workflows it it's like building software. There's the minimum viable solution phase and then there's the refactoring phase. So, doing this is going to help you get better and better at this process.

Thanks

28:21And the last slide of my talk was a thank you slide, but it's it's gone. Oh, there it is. Okay. so thank you. Thank you all for your time. And if you want to follow along as I continue to explore these tools and just share my insights on like what I'm seeing out in the e-commerce space, what's working, what's not, you can scan this new QR code and you can sign up for my newsletter. Alex, thank you so much.

28:47>> [applause] [cheering] >> I I've been waiting for this talk for quite some time. I'm very thankful for the way you think and the way you speak. Thank you so much. Any questions for her on this? Understood. Thank you guys. You guys are going to reflect on that one for the rest of the night. I'll tell you that much. Alex, thank you so much. >> [applause]