Bonus: How an AI CMO Can Jump Your Revenue 25%

0:00Henry AI:So we built LeBeg. It's a tool that tracks your data, your competitors' data, and whole industry. And it figures out what are the latest best practices and trends you can implement on your store to improve your business performance. We look at your creatives and competitors' creatives, and we find what are good elements of your competitor creatives that you are missing on your ads. The big problem with CEOs and business owners is they hyper-focus on one channel, not looking at the whole picture. you know, something that really blew me away, AI visibility as a new marketing channel. Conversion rates of chat GPT traffic are higher than for branded campaigns on Google Ads.

0:36Henry AI:So literally, there is a higher probability that people will purchase your product if chat GPT refers to product than if they organically search for your product.

0:51Josip Begic:Joseph, welcome to the D2C podcast. Excited to dive in with you on some of the interesting developments around AI and e-commerce. Maybe just start with a little overview about why you built Lebeg and Henry AI.

1:05Henry AI:Yes, Eric. Thanks for having me here. Very excited to share what we're doing and what are the current trends in the marketing industry and e-commerce industry altogether. So we built LeBeg. It's a tool that tracks your data, your competitors' data, and whole industry. And it figures out what are the latest best practices and trends you can implement on your store to improve your business performance. A kind of problem that we decided to tackle is if you ask Facebook agency what to do, they will tell you Facebook ads. If you ask Google agency what to do, they will tell you Google ads. And there is no objective player who would look at the holistic picture, what other brands are doing, what are the industry leaders doing, and what are the trends, and tell you objectively without any subjectivity or human bias what you should be doing as a store owner to improve your business numbers.

1:48Josip Begic:Take me back. Who is Lebeg? Because just for everyone wondering, this is L-E-B-E-S-G-U-E on the website if you want to go visit. But who is Lebeg?

1:58Henry AI:Yeah, so Lebeg is a French mathematician. Henri Lebeg is his name. And he reinvented the theory of integration like two centuries ago. and our idea is in the same style. We're kind of going to reinvent the online marketing ecosystem by just being able to measure things better and detect these small spots that everyone is missing on, maybe inventory tracking, inventory issues or emails not performing or competitors running 30 % off promotion. Our idea was to pick up these spots and figure out the best possible device in the marketing industry.

2:27Josip Begic:Very cool. So we're just on the back end of Black Friday Cyber Monday. Give me maybe an example of some of the insights or some of the stories of growth that you guys help unlock with your tools?

2:40Henry AI:Yeah, so I think there's one interesting story. It's not from Black Friday. It's a little bit older, but I think it's a great example of where AI and tools can really help you figure out what is going on. So what was this story? Our customers, it had a classic marketing stack. Facebook ads, Google ads, email marketing, and Shopify store. At some point of time, things stopped to perform. You know, for two or three days, conversion rate was 30 % lower or 20 % lower. MER dropped a bit, revenue dropped a bit. No one knew exactly what's going to go on. Maybe it's just like a weird trend or weird seasonality.

3:12Henry AI:But after seven days, teams started pointing fingers one at the other. So it was a separate entities for Facebook ads, Google ads, creative. And Facebook team said creatives are not good. Creatives team said targeting is not good on Facebook. Google team said that website is not performing and website said email's not performing. So it wasn't clear what's going on. And what we were able to do right around that time, we started building Henry AI, which is like LLM layer on top of all of your business data. And it figured out that exactly when conversion rate dropped, business changed from automatic discount codes to manual discount codes.

3:47Henry AI:And what caused that is, you know, users are not signing up and getting welcome 10 code anymore. They had to sign up, wait two minutes for the email. There was delay in the email flows. and then they would have to copy-paste XYZ XX75 code that was custom for every customer and copy-paste it on the checkout page and convert. And since the audience was a little bit older, like 40 plus, they were not converting on their store. So like there was a 20 to 25 % decrease in conversion rate and AI was able to pick it up like right away and pinpoint the exact problem that hurt the business growth. And they changed that and revenue just increased by 25%.

4:23Henry AI:MER improved and we were happy ever after.

4:26Josip Begic:And is that as simple as literally like a CEO going into the tool and be like, why did revenue drop?

4:33Henry AI:We are trying to make it that simple. I think we are all still learning how to use LLMs in the best possible way. You know, I hear from our developers, there are such smart ways how you can, you know, query LLMs. So it's still, I would say, work in progress. But we literally have like one tool within the AI, which is called the revenue drop investigator. And you just ask him, what did my revenue drop? This is kind of the key concept that we are trying to solve. Is it because of this? So our AI doesn't just understand correlations, but also causalities. For example, if there was stock inventory issue, Facebook ads team had to pull back the ad spend.

5:08Henry AI:There is not enough traffic on the website and emails are not converting anymore. So we can kind of connect these dots and understand the initial problem that caused the revenue drop is stock inventory problem that you have for your store. So kind of also similar with competitors' data, we can pick up that competitor is running 40 % off promotion first week of November. And you should either change the tactic and, you know, start doing the promotion also or pull back the ad spend and save it for the second week or something like that.

5:34Josip Begic:Correlation and causation are the main factors in statistical analysis. It's so easy to get them confused. And, you know, this makes me think of the world of attribution, sort of shared attribution, attribution modeling. How do you guys think about correlation and causality and how are you able to drill in and actually find causality using your tool?

5:54Henry AI:I think the big thing is just having a transparent and objective entity. Facebook team will always say it's about the top of the final Facebook ads, while email team will always say it's because they're top emails that they're sending. It's really hard to pinpoint the exact reason. Here we introduced a concept from game theory called Shapely Value. It won a Nobel Prize in 2012. It computes all of the possible combinations of what would be happening with your store revenue. if certain player, so like marketing channel or ad, was not involved in the advertising process. And based on calculating all of the combinations, it gives you a score for every marketing channel or ad how much it is really impacting the business group.

6:37Henry AI:So that's like one, I think, novel, interesting concept that we are using. It's called shapely value from game theory. But yeah, you know, like even in statistics, they say that every model is wrong, but some is useful. So, you know, like every model just tells a story. You know, it's important who is evaluating the story and what is the key takeaway that you can take as a business owner to change your business results for better, using this data and this analysis.

7:00Josip Begic:And then how big is your data set? How have you guys kind of gathered your data set?

7:05Henry AI:Yeah, so we currently have around like$4 billion of ad spent in our database. And we are also tracking like 25 ,000 brands that are competitors of our customers or industry leaders. So kind of we wanted to enrich this data set, not having first-party data, but also third-party data scraped via ad library or like emails or like Google Transparency Center to enrich the data and the decision-making process for a brand.

7:30Josip Begic:Are there any platforms that you sort of noticed or really pull ahead when it comes to... Because everyone's looking for new customers, right? The tool stack for reactivating customers, whether it's through email, SMS, retargeting, but everyone's looking to put emphasis on new eyeballs to actually grow their brand and not just spend money on the people who might have bought anyway. Anything that you would add from the data that you've looked at about where people are finding the most effective top of funnel new customers in today's environment?

7:59Henry AI:Yeah, great question. So I think definitely there's a TikTok who is growing crazy this year. That's an important thing. But something that really blew me away is AI visibility as a new marketing channel. So we saw recently this year, there was 40x drop in chat GPT referred traffic to Shopify stores. And now Gemini is picking up on chat GPT. So that's a massive change in trends. You get like 10 % of your store visitors via ChatGPT. And more importantly, conversion rates of ChatGPT traffic are higher than for branded campaigns on Google Ads. That's something that blew me away. So literally, there is a higher probability that people will purchase your product if ChatGPT refers the product than if they organically search for your product on Google Ads or organically are paid.

8:44Henry AI:That's something that just I couldn't believe. and revenue per session, more interesting, revenue per session is 30 % higher if ChatGPT refers traffic to websites. So kind of he's also able to upsell more to customers just because I don't know. For me, it's very convincing when ChatGPT tells me to do something.

9:02Josip Begic:It makes sense. It's absolutely wild though, because you're basically, you're already brand aware when you're searching the brand. Obviously, you know about the brand. So you'd think that would be the highest, most intent. But ChatGPT is actually accomplishing awareness and consideration in the same step, which is pretty amazing.

9:19Henry AI:Yeah, correct. Correct. Actually, how we started analyzing that is I was preparing for a conference. And, you know, like ChatGPT gave me some feedback on, you know, the presentation that I have how to speak. And I was like, okay, I'm going to ask him what should I wear for the conference. And he gave me such good advice, what's to wear, that I decided to, you know, wear that in the end. And it was a very successful conference for us. And after that, I was like, okay, if I'm doing that, probably a lot of people are doing that too. So let's start this analysis. You know, let's look at a$4 billion data set and figure out what's going on here.

9:50Josip Begic:What advice do you have for brands who are looking to become more discoverable to the different AI agents and platforms?

9:57Henry AI:Okay, so that's something very researching. And also, I think everyone is researching that. So I read some couple of papers around this topic. No one really, you know, no one really knows how it works. Some things that we saw are extremely important are, so one thing is just that information on your website is, easily fetchable. So you have to allow ChatGPT or Gemini crawlers to crawl their websites to find the key information. They don't have to go too much around your website to figure out the key information. So like returns information, shipping information, reviews should all be on the product page.

10:32Henry AI:Because unlike SEO, AI agents will not crawl your whole website. They will crawl just top two or three pages and figure out what the brand is about. And they will compare the data with other brands. So that's one thing. Just easy access to the website data. And the second thing is off-site information too. So like Trustpilot reviews, Reddit reviews, you know, they're using this as a source of knowledge and information that is on your website should also be available on Trustpilot, Reddit, Captera, G2 Crowd, Product Hunt. These things are very important to cross-reference what information it finds on your website.

11:07Josip Begic:I think that makes perfect sense. It's interesting. It's like AI is more human-like than the web crawlers, where web crawlers are crawling every little nook and cranny of your website that a human might not ever see. AI is doing it kind of like a human brain might do it a little bit, which is an interesting way to think about. So optimizing for the customer, optimizing for AI are probably closer together than optimizing for SEO and optimizing for the customer. Would you agree?

11:31Henry AI:Absolutely. Absolutely. And that's funny because we built a feature that gives you these suggestions. So it crawls your website and it finds mistakes on your website. And whenever I show that to customers, actually advice is so obvious that they have to implement it. Like compare your products with competitors if you're in a hyper-competitive industry, or add user reviews, or explain to customers what is the shipping process if everything is on pre-order. It's such an obvious thing that you have to have it on the page that AI can really help here even just improve your website and increase conversion rate organically.

12:03Josip Begic:Do you think humans just, we just get lost in the sauce or we just get hyper-focused on specific things? And so what AI does is it really like zooms out, looks at the big picture and finds all these things that humans just kind of like don't think about?

12:17Henry AI:Correct, correct. I mean, AI also needs to have the right data. So if it only sees a small data set like your Facebook ads, it won't know too much. But if it has a whole understanding of your business and competing brands, it will work better. So what is important here, we built our tool to build like CEOs best friends. So it will look at all of the data and tell you what you should be doing. And I think a big problem with like CEOs and like business owners is exactly this. They hyper focus on one channel, not looking at the whole picture. You know, like motivation for me starting this career was my CEO in a previous company where I worked.

12:51Henry AI:He was all about like Facebook ads, Facebook ads, Facebook ads all the time. And I was like, come on, we didn't touch website for 12 months. We introduced like 10 new products and you still just want to iterate creatives on Facebook, you know. And when we changed the website structure where we introduced new email flows, that's when the business started to grow. So sometimes, you know, you can just be too focused on one thing, not seeing the whole picture. And I think AI is really good with that. and we try to further enhance it too.

13:16Josip Begic:One of the things we're talking about on the podcast, just to be specific about Meta for a second, all the time is Andromeda and how big of a deal the Gem Lattice and the Andromeda engine and all of these things and how they have played a big hand in brand success over this year and specifically over this Q4 period. But I've also had people come on and say, oh, it's all overblown, you know, how big of a difference Andromeda has made in the environment. And I'm just curious from someone who looks at the data, looks at things like objectively, like your AI tool does, what's your take on how big of a deal this adoption of AI systems within the meta algorithm has been this year?

13:56Henry AI:a very good question so actually we pulled one interesting data point out on the usage of m dashes on facebook ads uh that's really interesting uh i think i did share it with you previously so uh there was uh currently 12 of facebook ads have m dashes in their copy so that's just like you can see that you know 12 of brands are just copy pasting chat gpt text or cloud text not even thinking to change the m dash to something more human um so that's interesting you know around 10 % just copy-pasted. That's interesting to see. So purely around your question on Andromeda, I don't know, actually, really.

14:34Henry AI:We saw really brands performing really well by following this new Andromeda structure with more creatives, user persona. But usually these teams are already very sophisticated. They knew exactly what they wanted to achieve. I would say more brands, in our case, mess up the fundamentals. You introduce 20 new creatives on Facebook, but you don't have a call to action button on the homepage. Or like you can interact creatives if you really, really have great creatives, but if you don't have the basic creatives set up in the right way, it won't work. I saw sometimes businesses overusing, I don't know, Gemini to create new images for Facebook ads, but these images are bad.

15:16Henry AI:So if image is bad, it is a bad image. No one is going to convert. So I would say solve the fundamentals first and then it's really state of the art. how do you use Andromeda updates to launch the best creatives or iterate on specific user personas for some of your creatives?

15:30Josip Begic:You mentioned the emdash. I see it all the time. The amount of times I'll be scrolling on TikTok or Reels also, and you'll hear someone who's talking to camera and saying like maybe crazy, some crazy philosophical thing, but you can just tell that they're reading. My favorite was actually I follow the Dune franchise really closely. And there's a creator who I love who was talking about the Butlerian Jihad in Dune, which is, to get really nerdy, is about when the people of Dune decided they were not going to use thinking machines anymore. But he was using AI to write his scripts about how humans need to get rid of AI.

16:04Josip Begic:And he was called out on it. And I thought that was hilarious. So, but my question with that statement is when you did the analysis of the MDashes, are people, because generally these copywriting machines, whether it's ChatGPT or Gemini, they're using best practices. And MDashes are actually an incredibly useful grammatical tool that very few people know how to use. So my question is, do you know if those brands were just copy and pasting directly from ChatGP, do you think they were seeing an increase in results? Or are people as wary as all of us technocrats are when we see an emdash or certain formats in the text?

16:40Josip Begic:Or is AI text working well for the average consumer?

16:45Henry AI:That's a very good question. And by the way, I'm also a huge Dune fan, so thanks for that reference. You know what I'm talking about. Yeah, absolutely. I know what you're talking about. But yeah, so regarding M-dashes, we actually analyzed, you know, same businesses over the last four years. So there wasn't business A versus business B. It's all the time, same businesses. And you can see like 0.25 % of them were using it before. And now it's 12%. So obviously, there were some that were using M-dashes in the right way. But kind of this 99 % of them are using it as, you know, I think just chat GPT copy paste.