The Exchanges

Q Uh, okay. But this is where people will get stuck, right? Because there's, there are people, first off, most people watching this, they're just very bad on camera in general. Some people are even worse at trying to figure out to write scripts to have other people talk like on camera, right? So how are you helping people write scripts that convert like your Facebook ads convert?

A Yeah. Uh, that's a good point. So first I think that's, One skill that, uh, until now was very important, but, uh, that is gonna be completely replaced by AI over time. Um, basically what will happen, I think, is You'll describe your product. AI will come up with a script for you that will use a platform like Arcades automatically, find the perfect actors, put that into your ad account and test that automatically, and then use the data from the ad accounts to come back and generate new versions of the script. I think that's the vision that we have, and that's going to be the future, but right now you can basically help get help from these tools. And we, for example, we built this hook generator Where you can describe your product and it will generate hooks for you. The hook is the most important part of a script. It's the first sentence. It's what will get people to click on, uh, a video. So you can basically go to hook generator, get archives.ai slash hook generator, describe your product, and it will automatically generate the hooks for you and give you ideas of hooks that are, uh, likely to work well. And that's a good starting point.

AI assessment note: “we built this hook generator Where you can describe your product and it will generate hooks”

Q How do you decide what to automate? How are they working? How do you tweak them over time? Before we do that though, I don't want to gloss over those growth tactics you're using. Cause they're obviously working really, really well. You've grown to six million bucks of revenue. And I meant to ask you to, when did you launch the tool? So we can sort of a growth rate.

A Yeah, it was January, 20, 24 when we first, uh, onboarded one customer. It was very private at the time. It was like just a couple of friends in the industry. We were working in the mobile app industry. That's how we got the idea. By the way, we were running a, a seven figures mobile app and advertising it with like a lot of UGC creatives. And we realized, okay, like this is something, uh, that's going to be, Uh, completely changed with AI and we decided, okay, this is a much bigger opportunity that this mobile app studio that we are working on. So we decided to sell it and to start like building, uh, this tool. So in January, we started like onboarding a couple of customers and step by step, it grew like very quickly. In the first months we did a lot of like, uh, revenue, like straight away. So we understood we had product market fit. Now it was about Building a better product and bidding, building the growth strategy to, to reach out to more customers.

AI assessment note: “Yeah, it was January, 20, 24 when we first, uh, onboarded one customer.”

Q Wow. So how much of your day is like just coming in here and like clicking start on all these tasks? I imagine you just have like a bunch of computers up where it's like start, start, start. You get your breakfast, you hit start, start, start, start. You come back an hour later and you like, you just did like 30 days worth of work.

A Actually for a lot of them, I don't even need to click start. Uh, they are, for example, I have one, uh, that, uh, maybe I can show you, but it might be a bit too complex, but that basically will send me, uh, messages in Slack every single day when my competitors run new ads that are worth replicating. It will scrape the ads libraries and yeah, it will download the ads from Facebook ads of my competitors and it will like Take the transcript of this ad, send it to ChatGPT, and then, uh, ChatGPT will rewrite the same ads, but for arcads, and it will send me that in, uh, Slack with a message like, uh, this ad was running on your competitor's account, and now you can replicate it and click this button to replicate it.

AI assessment note: “Actually for a lot of them, I don't even need to click start.”

Q forget to do. You have to be a great, you have to be a historian almost like you have to document history. What does well, because that's how you're going to train anything like a gum loop. But Romain, I think is obviously very good at that. He describes it as, you know, a simple Google doc, but you can see here, walk us through Romain. What have you documented?

A Yeah, a hundred percent. AI, what you have to understand is AI is very good at duplicating stuff. It's, uh, I think it's gonna be good at creating very soon, but right now, like, it's, like, really world class at taking something that is good and doing something similar. So, uh, yeah, I made this list of very good, uh, Twitter posts, and, uh, actually this one was from the founder of Gumloop, and, yeah, I, I saw all of these Twitter posts that I think were very good, so that there are There are 10 of them, and I added them to the Google Doc, and then I, let's go back to Gumloop, I told Gumloop, ok, this is my input, and the input is what I want to talk about, so this time I was, I wanted to talk about a new feature for Arcade, where you can control actors' emotion. In the post, we'll show sad and angry. And so those are the two inputs that I gave to the Gumloop flow. The thing I want to talk about and the list of inspiration posts from the Google Doc. And then Gumloop will combine them together, so it will, like, combine these texts with the content of the Google Doc that it will read, and then it will send that to a prompt and ask ChatGPT, okay, I want you to ingest all these articles from the inspiration list. And then carefully choose one article in the inspiration list. Uh, and you basically need to do a similar post inspired from one of these articles. And I say something …

AI assessment note: “I made this list of very good, uh, Twitter posts”

Q And, and, um, what's driven most of your growth to date in terms of go to market motion? Is there a Affiliates here. Is there an SEO strategy? Is it just you, you post great content on LinkedIn and Twitter? What is it?

A So I think we have three big acquisition channels. The first one and the most important one is like actually paid ads. We use our own tool to promote our tool. So we do some paid ads on Meta to, um, like basically show what the tool can do and We just prove it works by using it, right? That's number one. The second thing is content overall. So content means, as you said, like me creating content, people of the team creating content on LinkedIn, on Twitter, but it's also sometimes partnering with, uh, other people who create, uh, good content. So content creators, uh, who are, um, already posting about, uh, Adds about, like, how to do marketing or about AI in general. That's the second one. And the third one, uh, that, uh, we are pushing more and more is, like, more, you know, direct, uh, led, uh, sales, um, sales-led, like, approaches where we go find people who spend a lot of money in paid ads, and we try to, like, get in touch with them and onboard them on the tour on higher subscriptions.

AI assessment note: “The first one and the most important one is like actually paid ads.”

Q What tools are you using to go and try and find folks that are spending a lot on meta or LinkedIn, et cetera? Is there, you know, is it like an, you know, an Ahrefs or similar web kind of tool or something else?

A There is no like secret weapon. We built a lot of internal tools. So very often when you want to either walk with influencers in the space or with like big companies in the space, like It all comes down to identifying them. So sourcing these people on social media and then reaching out to them. And it turns out AI is very good at this kind of task because like sourcing is a lot of like repetitive work. So it's a lot about scrolling social media, selecting some people based on some criteria, adding them to a Google sheet and then reaching out. And AI is very good at that. So we built some internal AI agents To, to do that. And then once you have like sourced the right profiles, it comes down to how do I reach out to them in a natural way, have a conversation with them to convince them to start using our tool. And again, this AI can do in a very good way now with LLMs. So we ended up like building a lot of internal AI agents to do exactly this.

AI assessment note: “There is no like secret weapon. We built a lot of internal tools.”

Q Yeah. How does a guy, how does a guy build a fasting app for women? How did you get, how did you get into that?

A No, that's a great question. So we were like, basically we knew we wanted to, to do an app. We also knew we wanted to Make something that would help people be a better version of themselves. We were with my, when I say with my co-founder Dylan and I, um, we're very into like health and wellness. And we then kind of opportunistically saw that the, the people who spend the most in this category are women and there is no app specifically for them. And also there is a very specific pain point when you are in the weight loss industry that, which is Like having a community of other people doing the same thing as you is very important and it's very motivating, but as a woman, you don't necessarily want to post pictures of before afters in a group where you will find like men and women. So we thought, okay, what if we built obviously an app, but also a community that is very engaged where, uh, people will feel safe to, uh, talk about their journey and like go into like, like the challenges that they face, et cetera. So yeah, that's why we decided to build this app. And it was very community based where there is a feed, like, like, you know, kind of a Facebook group integrated into the app and people were very, very active about it and like sharing about a lot about their own journey. That's how we came up with this idea. And, um, yeah, again, very quickly, we realized like one of the k…

AI assessment note: “opportunistically saw that the, the people who spend the most in this category are women”

Q What enables your sales team to drive those customers into the six figures? Is it usage-based upselling, number of videos? Is it feature-based upselling, seat-based upselling? Which are those three?

A For us, it's been usage-based. So, if you, if you look at, like, the market, like, two years ago, no one was doing Gen.ai for advertising. Then, during the past two years, people started implementing that, and the usage has been growing up and up and up, and went from, like, maybe zero percent at the time of video on the Facebook ads library of a client being AI generated, to five, 10%, And now some companies are a hundred percent AI generated ads. So the usage is growing more and more and more again, and it's becoming more and more competitive to create ads and to, to get, to create profitable ads. So for this reason, uh, yeah, the usage has been significantly increasing over the past few years.

AI assessment note: “For us, it's been usage-based.”

Q Are these brand accounts or are these individual influencers, like, like, like digital influencers, not real people, but characters?

A Oh yeah. Basically they are like characters, characters you create and, um, You create them with using archives and you create a face for the account using the archives character. So I was thinking, okay, what if I can start posting, uh, 10 times a day? Actually, I very quickly realized that if you post 10 times a day on the same accounts, TikTok will not push your content too much because they don't really want people to be posting that much. So you have to find creative ways to be posting more content on multiple accounts. So I ended up like, um, uh, thinking what if I create multiple accounts and start posting on them? And then you'd have to do these like creative tactics where you have to have TikTok think you are a different person and you have to use VPNs, multiple devices. And I was just like playing with this and experimenting with this. And I decided, okay, let's record a video to showcase what I'm like, um, tinkering with. And let's see, uh, if, uh, I can like get, uh, some people interested in that and like to give me some, Tips on that. And I actually have a friend who is like very deep into this stuff. So I didn't like end up, uh, spending too much time on it yet, but, uh, that's something I think very, uh, uh, interesting to pressure. And I see a lot of Arcade's customers, usually teens, like, 16 years old, and they go really deep into this strategy using Arcade. …

AI assessment note: “Basically they are like characters, characters you create”

Q but no, it's really impressive. Bootstrapped to 500 grand a month in revenue, six million ARR, uh, six FTEs. And again, one of the things you do really, really well, uh, you just touched on paid ads. We touched on content. I want to touch on this piece of content before we jump into how you're building these, these agents. What are we seeing here? Is this your workstation here?

A Yeah, actually, uh, that's kind of a weekend project. So, um, uh, basically my thinking was simple when I did this and, uh, I mean, I didn't end up spending too much time on it. Like, uh, it was just something I do for fun. Uh, my core focus is just like focusing on the growing archives, but I was thinking, okay, with archives now you can create a lot of these UGC ads, right? With AI talking avatars, et cetera. What if, um, uh, instead of posting them on paid ads, which is what I really know how to do well, what if I started posting these on organic accounts? So basically TikTok accounts with no, uh, uh, no money behind it, no paid budget.

AI assessment note: “that's kind of a weekend project... you can create a lot of these UGC ads”

Q live. Cause remain, I think what a lot of people think about is, you know, they're not, you strike me as someone you're pretty darn technical, right? Um, you know how to write code and you're also a great growth engineer, which is rare combination, but for people that are just great marketing and sales folks, how could they use gum loop to run some of these AI agent playbooks?

A Yeah, I will try to find a simple one to give you a good intro about what you can do, and then I will try to show a little bit of a more complex one. But by the way, Gameloup, the team is very, um, good at teaching you how to use it. They have this AI agent inside Gameloup that you can chat to, and that you can like use to kind of understand better how to build them. And even the team, like, can reach out to them and, um, just, like, Ask them to, to guide you on how to do more complex workflows. So, uh, let me share my screen. Where is the, uh, yeah. It's right in Riverside.

AI assessment note: “They have this AI agent inside Gameloup that you can chat to”