How This AI Startup Hit $7M ARR in Just 18 Months | Romain Torres, Arcads AI

Today on the Pursuit of Knowledge Show, we have Roma joining us all the way from France. He's the founder of Ards AI. And Archads AI is a platform that makes talking head videos for advertisements and for other stuff using AI. So Roman, what was the inception of the idea for Ards and what can Arcads do today? I come from a background uh of making mobile apps and um you know when you have a mobile app probably there are two very important things you have to nail. The first one is making a good app just making a good product and the second one is marketing. You need to market the app and we were spending a ton of time and money on working with UGC creators. you know those uh creators who have a small audience on Tik Tok or or made up that you pay for the content that they create to promote your product, not really for the audience. And we were spending a ton of time working with them to create the advertisements for our apps. But at some point we realized, okay, AI is going to disrupt this uh and it's going to be a massive market because every single consumer product out there is going to use uh AI to create UGC. So we decided to sell the company, the mobile app that we had at the time. It was a wellness app and to focus 100% on this new idea, which is how are we going to help other consumer companies create winning ads using AI? Okay, before we talk about AI, can you explain what is userenerated content for people who are not familiar with the word UGC? Yeah, 100%. When you have a consumer product, you are always trying to introduce your product in a positive way and very often you put advertising budget on that. Very good way to do that is UGC. So UGC is you find someone who is good at acting, who is good at talking to the camera and at making good content and who will talk about your product in a positive way and then you will use that to run ads basically. And the UGC creator is typically someone who has a very small audience on short form platforms like Instagram res, YouTube shorts, or maybe Tik Tok. They don't necessarily have a big enough audience to make a living out of their audience, but they have the skill of creating good content. And that's a UGC creators. And that's what you use uh when you have a consumer product. So, I have used UGC creators a lot in my businesses and there was this platform called Billow. I don't know if you know about it, but pretty much the workflow was you ship this product whatever jewelry clothing to this person, you send them the script, then they take this whole time to film this video, they send you the video, you give them feedback, and then they change the video, then you approve the video, then you finally have this ad. Then you post the ad or you run it as an ad. Uh you post it as content or you run it as an ad. That was a very long workflow. That was like the most annoying part because when you're a new business like you want to make money like you want to get to revenue like ASAP and that was extremely timeconuming and before that people were having in-house people to shoot content and then these platforms came and also these platforms like below were expensive like one video could cost 100 150 200. So I'm guessing it's pretty natural why AI is going to disrupt that because the workflow becomes really like you can validate uh your product demand really fast with AI and like test these ads. So how what's the workflow with ARADS right now? How fast can you make an ad? Minutes basically you type a script. We have a library of over a thousand AI actors. You can browse the library and find the perfect AI actor and click generate and you will get an engaging video in two minutes. basically. And that's why it's so powerful. As you said, you go from taking multiple weeks, paying multiple hundreds to taking two minutes, and paying a couple of dollars and you get your video. But, uh, what do you think about the uncanny valley or like, you know, the the bad lip sync, uh, people who who say like, hey, these AI ads will not convert or they're not going to be valuable for me. Like, what do you say to those people? Yeah, that's a a great point. So when we started arads and it was almost two years ago so it was in January 2024 that was probably the biggest question we had and already at the time AI was actually quite good at making good content but that's probably the first hypothesis we wanted to validate as a business when we launched archives we wanted to validate multiple hypothesis the first one will was do people want to pay to have access to AI UGC Are they willing to test that? At least give it a try. And that's the first one. The second one was once they give it a try, are they okay to put it in the ad account and to actually expose the brands to the content? And the third one was when it's in the ad account and when they are posting the content, does it perform and does it bring good results? And we were super super lucky because when we first started with the first client that we were working with after two or three iterations the ads that we were producing using AI became one of the top ads when you are doing advertisement typically you do 10 20 50 ads every week and 10% of them become what we call winning ads. Basically, they take most of the budget and they get very positive return on investment and that never happens where uh you test an ad and it becomes the first one to become a winning ad. But we were super lucky and it actually happened for the second ad we produced. So now we knew okay people are willing to buy this because as you said it's so painful to work with UGC creators also it can bring results. So now it's just our capacity to scale this product as quickly as we can that will make the difference and that's what we did. Man, it's shocking how things have changed in the last two years. We went we just Google released V3, 11 Labs released V3 and it's insane. It's even for me, I'm thinking like, should I still make content because I know a lot of people who have made AI avatars and they got the perfect voice down with 11 Labs. Uh, and they got the perfect video down. They they have the best stock footage. Like the AI gets the hands right now. It gets the text right as well. Like, it's getting to the point of really really good. Like, how how much better do you think this is this is going to get? or are we going to hit a ceiling somewhere? That's actually a very interesting question. Um I think to your point like should I still be making content? I think like two three years from now every single piece of content when it's advertisement or entertainment will be AI generated. I mean there is no doubt uh that that will be the case. But for educational content where you want to actually learn something, I still believe that that's where most of the human content will live because when you consume content on the internet, if it's for entertainment, you don't really care who is actually entertaining you. What you want is to have the entertainment. If it's an advertisement, in some case, you might want to have like the brand associated to uh some personality. But again, even in that case, like who believes like Josh Pony is actually drinking Nespresso every day? I mean, I don't think so. So, even for advertisement, I think most of the content we will consume will be AI generated. For educational content, you want to trust the person who is educating you and trust that they have a very specific knowledge. So, for example, when I'm learning about like how to make a bigger business, I don't want to be learning from a 17 years old who hasn't like ever produced anything, who just good at like creating content. I want to be learning from Jeff Bessos or Elon Musk or whoever is inspiring me. So in that case I'm I might still be consuming human content. But um yeah I think like the the quality won't be the the question of like will the AI be good enough that all the content that we consume will be AI generated. The the question is more on which type of content will be AI generated and which type of content will be generated by humans. I think you're on to something here because Meta just came out and they released a press release saying that by 2026 all the ads will be AI generated. Uh and this kind of like sent a lot of shock waves uh in the community right now because everyone's like oh there's ad agencies are going to die. There's not going to be any like people hiring actors. This will cause a lot of unemployment. Um no doubt for sure because I'm it's like every business I talk to is getting leaner and leaner now. Um, but I want to talk about your team now. How as a creator, you've done incredible numbers. I I want to know how uh how much have you grown in the last two years because it's quite impressive. Yeah. So, basically, we went from zero to almost seven now, about 6 million AR in uh yeah, 18 18 months. That's insane. Do you like uh I feel like you kind of predicted it like I mean you feel like you already you knew it uh why do you think it what was something that hit right and what was like one of your growth marketing uh strategy I see you're active on Twitter like what do you do that drives the continuous growth first of all the thing you described is like I think the mission of any business like taking something that is painful costly and long and making it easier and faster. So the the real aha moment for me was when we first started pitching this idea to clients uh potential clients because at the beginning uh we basically used existing AI tool to produce ads to existing mobile app studios and e-commerce companies and agencies and we're pitching them we will produce your ads using AI and I was like okay so I said I remember I really vividly remember sending an email to one person I wanted to pitch this ID to. And the email was a link to a Google Drive with an AI generated ad. And the email was something like, hey, I generated this ad. It took to it took me 2 minutes to make the ad and here's the result. I think it looks good. What do you think? Something like that. And I was thinking to myself, wow. Like obviously this is going to be huge. Who wouldn't want to use that when you know the time it takes to produce actual UGC content with actual people? So yeah, that was the first aha moment when I was like, okay, this is a big idea. So the fundamentals of entrepreneurship basically, which are make something people want, make something that solves a real problem. Like I know it's boring, but I think that's the main thing, the main reason why why we grew so quickly. The timing was the right one. like probably if someone were to launch the same idea, now it's probably too late because like people already adopted our our tool. Um and then obviously having the right uh channels of acquisition was key. I remember we had some competitors launching at the same time more or less and a lot of them just didn't get any traction. So you can sometimes get the good the right idea but end up having no users or not the same level of traction. Um and so for us yeah it was like I started posting content on Twitter, LinkedIn because we had this experience in the past making uh ads and like growing consumer products. So uh we already grew an app to multiple million in annual recurring revenue. We sold this app. So the market what what was that app again if you don't mind sharing? Yeah. So it was an intermittent fasting app. Um basically uh we were um helping people in the US lose weight with intermittent fasting and um it grew mainly like 99% through paid marketing. So we learned like the hard way how do you make it so people are actually converting into an app through paid marketing and there are a lot of like things I learned during those years building this product and growing this product um that I could talk about on Twitter, LinkedIn and that's a very good way to then push our ads which is a better solution to produce ads. So I kind kind of started like publishing content on like how to grow a consumer product using AI and that was our initial source of traction and then the the product was inherently viral. So uh I remember Sunday uh I I woke up with uh hundreds uh new customers every single hour. Uh so we had this Slack channel where we receive a notification every single time there is someone who is making a a payment and the Slack channel was flooded with payments and it was actually one of our users who shared a video that she made on Twitter and this person didn't have even such a big audience. She probably had like 500 followers on Twitter but the video was completely barrel. a lot of Twitter influencers started reposting it and it like became this like phenomenon. So just the product was inherently super viral. People didn't even believe it was AI generated. So that was the second push and then we obviously uh worked a lot on very a very large number of acquisition channels that you need to optimize when you have a SAS. Who is your ideal customer profile? Like I'm pretty sure like there's a lot of young founders using your tool, but I I'm guessing you're targeting more mainstream companies. Yeah. So the the the ICP is people who understand that you need to be producing a lot of ads to succeed. And actually the biggest companies they don't AV test so many ads. So if um for your audience like if uh they don't know there is this tool called the Facebook ads library where you can go and see all the ads from any existing page and if you go and look for the biggest brands out there Coca-Cola, Uni Lever or any like big big corporates with huge marketing budgets, you'll notice that they don't necessarily run thousands of iterations which you should expect because performance in paid marketing comes from testing a lot of different things. But these big corporates, they have a different approach to performance marketing. Uh they actually don't do performance marketing. They do brand marketing. And the reason for that is they have to protect their brand. When you are Coca-Cola, you cannot afford to have a bad bus. So you have to have a ton of control over the output of every single marketing campaign you do. Thus, they do very few different iterations on the ads that they do. The real ICP is a company that does test a lot of different ads with a very large budget and sometimes the budgets can be very close to the budgets of these like huge uh companies, but they have a performance marketing approach where they test a ton of different u creatives and this is our target audience. So, I'm thinking uh performance marketing agencies, big lead generation companies, mobile app studios, um e-commerce companies with significant budgets, companies like True Plastic. If you look at the ad library of these companies, they do like thousands of ads. Um gaming companies, uh companies like King for Candy Crush, they test thousands of ads as well. So, this is the true recipe. Yeah. the I I'm guessing they don't test hundreds of ads because cost of creating content is so high for them you know uh like Pepsi or whatever like they these campaigns they cost them millions of dollars that's why they cannot test a thousand ads okay before we move on to new ideas other ideas that in AI other people can do two small questions I had what is the business model of Arcad's AI and do you guys publicly reveal which model you use so we have internal models actually uh in some cases we use existing models So you mentioned uh um 11 labs uh they have a crazy uh voice model. So we don't want to compete with them on uh voice. So we just like use their model but um for most of the workflow we actually have internal models. And what was your other question the the pricing the the business model so basically it's a subscription. Um everyone can go ahead and go to arch.ai I and start making ads uh with paid subscription. Um and it's credit based. So basically the larger your volume, the lower the price per individual uh creation will be. And the standard price is €10 per generation which is uh $10. Yeah. I'm guessing $10. Yeah. Yeah. Yeah. Wow. That's a that I mean again it's like entrepreneurship about is it's also about creating something faster, cheaper, and better. And you you're kind of out there cuz I told you like the traditional UGC video would end up costing me like 100 or $150. So you've literally slashed the price by 110th or 115. Yeah, exactly. Yeah. Yeah, I mean as long as they convert, you know, and right now the cool thing about AI ads is because there's still not many ads on Instagram or that are AI. So AI ads tend to stand out a bit more right now. And I feel like that's why they they can be actually like a good strategy to use AI ads because they can convert cuz they they're just unique, you know, compared to other ads. Yeah, 100% they are unique. um we are seeing more and more of them and again um in two to three years it's not going to be a differentiator because everyone is going to use AI for every single piece of ads uh we'll see and I'm pretty sure the adoption will be much faster than one what we anticipate what's the next big thing you're working on on ARADS like uh what else are you guys are innovating on or thinking about yeah so I think LLMs are getting much better and we can assume that they are going to be getting better and better every single month. So when you do an ad, there are really multiple things you need to nail. You need to have a good script. What is actually happening in the ad, you need to have a good person like who looks the way you want to look in the ad. You need the voice to to sound good and then you need to do a good edit. Uh you need the final video to be um amazing, right? And so far we really focused on like the quality of the person and the voice working with 11 Labs. The voice is going to get better like they released their newest model just the voice is going to be automatically better and they are going to keep improving that the faces are going to get better as well and we are like working super super hard to always be the best solution on the market which I think is the case right now the most engaging AI actors if you want to create them that's with our for without any doubt um but the next steps are how do we use LLM to help you come up with these faster so that you don't even have to think about what's the next ads that you should do. How do we use the data from your own ad accounts or from your competitors to come up with scripts that are likely to perform? And as a user, how do we make it easy for you to just come and select manage the AI instead of having to come up with the whole ID yourself and produce the whole ID? And then once the video is generated, when you have someone who is talking, how do we make it easy for you to connect your existing assets? maybe screen records of your websites, of your app, of your uh maybe images of your product. And how do we make it easy for you to produce the full ad without having to um go with an editor and using Adobe or whatever editing software you are using to producing the final ad. That's insane. Wow. So you you're pretty much making it so easy for people that they don't even have to prompt to rewrite the script. You'll even predict that in advance. That's Oh, wow. That's another level, man. You're you're thinking in it like really ahead. I love that. Why is it that all these cool like avatar companies are coming out of France? Do you guys have a certain uh dem like engineer demographic is what's the reason? We have a lot of engineers. Um we also have a lot of talents in the consumer space. I don't know if you're familiar with the company Voodoo but they are like one of the fastest innovative companies in the world in the world of consumer. They recently acquired Beril uh which was one of the fastest growing social networks ever. So we have the combination of understanding consumer and also good engineers basically. Yeah. Even you know the app I don't know if you know Yuka app. Yuka. Yeah they are. Yuka also came out of France and uh there's a lot of I noticed there's a lot of food apps that that are popular food apps all based out of France for some reason. Yeah, it's quite quite interesting like uh what part of France are you in? I'm actually from the south of France. um planning to move to San Francisco uh in the coming uh months because things are moving so fast that I think that's the even though France is like a good country to be in and has very good engineers and good talents I think that the true leaders of the world like the open AI of the world they are in San Francisco so if you want to keep like innovating at the fastest possible pace and building the best company in the world which I think we will we have to be there I think you're right I I used to live in San Francisco and now I'm in Los Angeles, but San Francisco is insane. Like the talent density of founders there, it's uh there's always someone you're competing with and it keeps you on your toes. Um, okay. So, for people who are like, "Hey, Roma, this was amazing. Like, you grew ARADS, now I want to start a company in AI." And what's an area doesn't have to be in consumer, but what are other ideas like you would explore if you had free time? I see a lot of companies having a similar trajectory than ARADS in the space of AI SEO and there are multiple ways to call this but basically that's the idea that now SEO is going to switch uh to Google to LLM so chat VPT perlexity code and people are going to use L&Ms to search for things and there is like basically a blue ocean in this space of like softwares to help people maximize their chances of ranking properly in the LLM's responses and yeah again I talked to a lot of companies in this space and they made me think a lot of us in the early days uh and we are still in the early days by the way but um they are probably like 8 12 months behind us in terms of growth and revenue but I'm pretty sure they will be at the same stage as us right now in a couple of months um and that's a space I think is very interesting because everything has to be built in this space. Another one I'm very Yeah, sorry. Go ahead. No, I was just going to say a little bit context for the audience. So, what Roma is saying is that SEO is changing. Uh, Chad GPT is eating Google's lunch right now and everyone is finding products on Chad GPT. So, this is this new thing is called GEO. It's like uh GEO which is helping people rank on Chad GPTs. Actually, I'm interviewing one company called promptwatch.com and they are they're solving that problem. Actually, there's a lot of companies trying to solve this problem. Uh the one thing I'm not sure is like how they're going to track the results. Uh cuz cuz OpenAI doesn't publicly reveal uh like chat GPT, you know, like OpenAI doesn't publicly reveal uh like Google like the search results and the Google trends, that kind of thing. So I don't know how these companies are actually tracking if it's working or not. Yeah. Um uh actually I talked to a lot of them. So one of them even told me that they told with Samman and that Samman told them they were not planning to release an API integration where you could see the search volume for certain keywords. So that's that's not happening. But the way they do it right now at least is basically they do API calls to OpenAI API to search for the keyword and to look at the results from different um VPNs from from different IP addresses. Um so that's how they do it right now. I don't know how sustainable this is. Um but um that's a very interesting space where a lot of innovations are going to be made for sure. Cool. What was the second idea? The second topic, it's more a topic. I don't have a very precise idea of how people should do this, but um you probably saw that more and more people are now using note takers and um I use a noteaker for every single customer call I do. And when you are growing quickly, um you end up doing a lot of these calls and then the next step is typically to come back and do a follow-up uh email. So let's say we had a call, I pitch you our um and then two days after I will tell you, hey, by the way, did you try this new feature that we just released or something like that? And right now when you are growing fast, it's very hard to keep up with that because you basically do a lot of call and you don't have a ton of time to manually follow up with every single customer. So basically that's the purpose of a CRM. Uh I would typically have a task in my CRM who would tell me hey you should like reach out to my nav and talk about this new feature that you have. I think this can be done with AI and uh taking the information from one call to another to automatically send sequences and follow-ups to previous conversations you had is something I'm very excited about. So I don't know how this will look like and if someone is building this uh please reach out but um yeah basically I think that using the calls from the content from your calls of the day to potentially reach back and call back previous conversations you had is something that is really very interesting. Yeah, I think that's the next step. I've tested all notetakers. they just do a poor job of following up. It's like using AI agents, right? AI agents take take it one step further and they actually do the work instead of just like giving you they're presenting you the research. Um have you experimented with this tool called Manis? I did at the time when they were kind of viral but um I didn't follow did they update recently? Uh no it's the manis came out of China. It's an AI agent tool that pretty much can take it takes it one step further. It's like charge GBT but on steroids. You uh it actually like goes on the browser and does the like you know like it can book your ticket for you. It can like it takes it one step further. And also this tool came out yesterday called Runner H. Um I'll put a demo on the screen for people watching. This one is super interesting. Like you can connect it to your Gmail and it'll basically write a response to every email already. So you just have to show like put go in each email whatever like unread email and it'll already have a response to each email. So you just have to like correct it and send it. So that's really cool. Yeah, actually that's something I'm very excited about as well. I I tried a bunch of them. Um I tried serif. I tried another one fixer AI. Um, the problem I had with these tools is I hate the Gmail interface for emails. Um, and I'm like a big fan of superhuman, you know, this email client where everything is faster, is more organized. So what I'm really excited about is like when a company will either super human will integrate these cap capacities or new company will uh basically replace superhuman um by building a tool that is like as fast as superhuman but on top of that has the AI capabilities to auto draft the emails and the follow-ups and the follow-up is really the the thing I'm excited about because I think there is a lot of uh money left on the table on follow-ups. So many conversations you had in the past and you didn't close right away but you have the opportunity to close now because there is a new data point but you don't remember about these conversations because you didn't necessarily do the CRM work to update the company etc. Uh I think that's like a massive massive opportunity and when you save time to people again same idea as archads you take something that people already do so basically filling out the CRM doing follow-ups manually and you make that faster easier then it's a massive opportunity that will generate billions of dollars of value. It's really sad, but I recently made my virtual assistant in Philippines from full-time to part-time. And I replaced all her work with N810, Make.com, and Gum Loop. I figured I pretty much hired someone and figured out every task she had, and I automated it with make.com and Gumloop. And it's it's crazy. It's just a human cannot come compete with Gum Loop. It's it's it's physically not possible. Yeah. and also a huge fan of gum loop. Um, and I think that's the second opportunity with AI. With AI, you can create a business that gets adoption much faster because now you can build a product that wasn't possible before and that solves a problem in a very unique way, but you can also keep the team super super small and grow very quickly um by automating a lot of tasks. So again like that's a business idea for someone who is less technical probably using gum loop and selling automation to businesses. It's like it's a boring idea. Probably tons of people are already working on that. But actually, I think there is a ton of value. If you are like not a developer, you don't have a specific skill. You just have to go on YouTube, learn Gum Loop, and start calling some businesses and offering them to automate some parts of the business and I think you're going to do great. What's your I think uh what's your do you have any predictions? Uh I don't know if you saw this news of uh Chad GPT, no Sam Alman partnering up with Johnny IV and they're creating this new uh we don't know the form factor yet, but they're basically releasing this new tech product which is going to even replace iPhone. Do do you have any predictions if that's going to be sunglasses or is it's going to be glasses or it's going to be like a pendant or it's going to be like a like a band? What is it going to be? Yeah. So actually on on this one specifically um I think it's going to be a super competitive market because it's very easy to build such a product. Uh, all you have to do is to connect to the API of OpenAI and it's basically a wrapper but in the physical world. And I'm pretty sure uh a ton of drop shippers will actually um build like a Raspberry Pi into a bracelet or something and sell that on Tik Tok and make a ton of money before it becomes too competitive. So I'm not even sure that there is going to be a huge winner in the in the tech world for for this product. I I my guess is the winner in this world will be from the drop shipping uh drop shippers and from people who are very good at paid marketing because when the product is easy to copy and there is a lot of competition the differentiation becomes your ability to market it properly. Yeah. Me personally, I don't really enjoy talking to Chad GPT on the voice model that much cuz when it responds to me, it's so slow. And uh instead, my workflow is I use this tool called Whisper Flow. And you just press one button on your Mac and it types whatever you're saying and then it responds back in text. So I can read it really fast. And I noticed for me that saves a lot more time instead of like talking and it's so so slow, you know, compared to me using whisperflow to just immediately type something and then it responds to me and I can read it really fast. Do you use it to to type text? Do you mean like uh to to send your messages? Do you use that to send messages in Slack or Yeah, so I just don't type anymore. I I press F9 key on my Mac and it types it perfectly. And usually, you know, like if you use the traditional Mac voice memo, it gets it wrong. Well, Whisper Flow gets it perfectly right every time. Interesting. So, you never have to type anymore and it just like saves me so much time every day. I'll try this one more. Cool. So, uh, how can people use arcats.ai? And, um, anything else you want people to take away from this podcast? Um, no. So, ars.ai, AI sign up self-s serve and uh I don't know who your audience is but if they are entrepreneurs um learn how to distribute product because now with AI it's going to be easier and easier to create product and the the best way to to make a significant business will be to learn how to differentiate on the communication and marketing use AI to make your marketing you will be faster than competitors and Um um if you master the skill of marketing a product then you are in unstoppable in the age of AI. Amazing. And how can people f people can go on rcast.ai to use a tool. And how can people find you? Are you on LinkedIn? LinkedIn Twitter x uh Roma Torres. Thank you for coming on the show, Roma. We learned a lot today about how Arcad.ai AI is disrupting the ad space and they've really optimized the workflow for people like myself to use UGC content because it it can be really expensive and take forever to validate demand for a product that you don't even know if you're going to be selling in a year from now. So, this is a great product solving a great problem at a great price. That's why it's been so successful. But really happy to have you on the show, Roma. Uh, we should do another like follow-up episode in a year or two to see where things are. All right.