Copy his playbook

0:00We've created more than a thousand of these agentic founders, and our agents are creating these companies on their own at this point. Let me do a screen share and just show people how they can actually do this, too. Presented by Zapier, the AI automation company. This founder raised more than $5 million for a multi-agent business creation machine. I'm seeing a lot of people use parts of this. I don't expect you to copy them exactly, but I do think that many of the ideas that he's implemented you're going to bring into your setup. Let's get into it. All right, what am I looking at

0:29here? What is this? Yeah, so this is essentially where everything begins. So, if you've ever been a founder on your own, you know that essentially one of the key pieces that you need to start with is is ideation. And so, we essentially have like an ideas engine that our platform goes and it scrapes TikTok, Reddit, X, and finds different pain points and

0:50trends that people are experiencing. Um so, what you can see here is essentially some of the ideas that our agents have come up with based on those trends. And there's everything from like, you know, Trump legal, you know, case managers to Black Friday deal navigators. Um there's Can you click on one? Yeah, so this was like essentially an app idea that

1:11was come up with by the system. And this one essentially is just like a digital tool that aggregates and simplifies Black Friday promotions and giveaways. This was found on Twitter or X on this specific date on November 27th. People were probably talking about Black Friday and that was picked up. And the theme, you know, what that was

1:30captured was was here. So, essentially some of these pieces of information were found online. And then our system, depending on a few different factors, it could be virality potential, it could be emotional intensity, it could be, you know, potential for user acquisition. The system can deploy an actual founder off of this idea or we can do that

1:51manually here. And so, we would essentially go and if we were really interested in this idea, if we were running a specific test, go and deploy that founder and find ourselves on a, you know, particular founder page. But these are the agentic founders that we've, you know, had our system spin up over the last year. At this point we have, you know, probably

2:10more than a thousand founders. You might actually pick one of these founders and say, "I want you to go and run with this idea." Yep, exactly. And, you know, we can turn them on, turn them off. But for instance, here's like a couple, you know, Fight Hype is is an app that we were playing with kind of like mid last year. And so, this was one that was super interesting for us to see. So, essentially there was like a lot of combat sports

2:34interest. And one of these founders essentially picked up on those trends. And one of the first things that they do is they spin up their own landing page. We can jump to it here. Essentially the first product that it built was like an intelligence API that was more focused on essentially data streaming for these different sorts of combat sports. Your software is picking up on the fact that people are talking about these fights

2:56online. They have some kind of need. Your AI founder says, "I don't know what it is. I think the first thing we should create is this site that I'm that I'm looking at right now that basically gives data." Exactly. Exactly. And then and then is able to, you know, build a site on its own. And I think at this point a lot of folks have, you know, landing page builders, essentially figured out and it's not too crazy. And there's a lot of services that can allow for you to do this as well. But, you know, essentially up to this point everything is, you know, agentically built, you know, the Stripe

3:28integration and everything else so people can check out and they can actually go in and, you know, sign up sign up for a site. Um and a lot of the times we run different tests on different parts of the system like this Fight Hype intelligence API product and the founder through experimentation and through getting user research was able to figure out that actually something that was more akin to a sports betting app would possibly have an even greater market. What's the user data that allows the agent to figure this out? Our agents will go and they'll interact with people online and try to essentially get greater

4:03insights. So, they might post something. So, they might post something, they might try to get a reaction, they might ask about like an idea that they're experimenting with. And then that way they're able to get prompted data to be able to, you know, kind of refine this. I'm sorry. So, meaning they might go back to Twitter where they the idea came from and say, "Hey, I see you all are interested in in fight data. I put this website up. What do you think?" And then

4:26look to see what the feedback is. Exactly. What are some other ways that they might do customer development? We have the ability for users to actually just, you know, call and text with one of the agentic founders as if they're talking to like an actual founder. What do you use to build that? Um so, I'm not exactly sure in terms of like the exact tech stack that was used to be able to do that. But, essentially our engineers were able to probably use similar technology that you would get with, you know, a board that people are using and, you know, other sort of cutting-edge technology that's out there already and

5:00available through other services. A lot of the stuff we also some of it we build internally and some of it we just use products that are already out there until we realize that like there's an actual need to build it ourselves. Um but this was a feature that we had implemented probably in like, I would say, June or July of 25. So, there was already the technology there to be able to essentially have pretty much full-blown conversations with an agent that, you know, should be available to folks today if they look anywhere. Um and is just kind of like the next step to text. But, what's cool about

5:32something like this is that a user can actually call an agent and say, "Hey, I actually like don't have this particular feature that I'm looking for." Like maybe the sports betting app is like, you know, doesn't accept a specific currency. And so, then our agent would essentially, you know, go listen to the user saying that, "Hey, I would love for this currency to be accepted." And the agent then can run a calculation and say, "Okay, well, what would it take for me to be able to like actually build this feature? Is this just a technical feature that I can build over the next 5 minutes or is this something that is more

6:05difficult?" But the first step is get customer data, then create a landing page to see how people are interacting with it. They interact with it. You might even make calls to them or invite them to call the the founder, the so-called founder of the site. You get even more data. And then you say, "Okay, there's clearly interest here, but how much is it going to cost us to put this up as a real business?" You do the analysis and there and there I'm assuming the agent stops and comes back to you and says, "This is this is the

6:32full analysis. How many people want it? What they want? What it would cost us?" Yeah, exactly. And in the whole time, you know, so for one of these pages for like the agents, we essentially see their neural stream and see like what they're thinking about as they're doing some of this research. So, the last step here for this agent was taken on December December 7th of last year and it was essentially reviewing its recent memories and current state and then executed pending market research task to gather more insights on the combat sport analytics and highlight generation landscape. So, they're constantly like thinking about these things. And, you

7:01know, they also have their own treasuries. So, they can, you know, we don't give them too much cash, but they have the ability to essentially, you know, generate their own, you know, content, you know, be able to call APIs and essentially spend on whatever it might be that they need to be able to understand their market better or be able to advertise or go to market. So, for instance, you know, they came up with this like super simple just like little background video for analyzing Fight Hype. This was kind of like a

7:28first little step. And then I believe they got some feedback. It was either through external users or it was possibly some people within the platform that were doing some of the like evals on the system. And I think one of the suggestions here was essentially if you put a person into this, it'll perform better. And so, it just then agent went and like added a person speaking to, you know, the Fight Hype, you know, intelligence app. So, so

7:51yeah. You essentially built over a thousand startups. Fair to say? Yes. Yep. Okay. Some will always win, some won't. Do you believe that the ones that will win will eventually continue to have AI founders lead them? Everyone is betting on the single person unicorn. And we were basically saying like we think that we might actually build a zero-person unicorn faster than

8:15the one-person unicorn is launched. And so, I think that it's really about that there's a couple different things here. One is our agents even today, when they hit what we call real-world boundaries, call in human support to be able to overcome those boundaries. Um so, kind of what what has happened with Rent-A-Human in the last month, that was essentially an engine that we built for ourselves last summer. So, this is kind of like the human hire view. This human has been paired with five founders that they're essentially working for. Each of those founders has essentially their own inbox. So, this is all dummy data. But,

8:54they actually do all have their inbox. So, a human being will come over here. They're working for these different agents that have these names like Michael Chen, like Emma Rodriguez, like David Kim. And funnily enough, one of them is called Sam Altman. So, this is what the human hire sees is essentially a list of available tasks and other tasks. And so, what they see as instructions from their agentic founder is something like this where it's essentially like, "Hey, you know, check this out. This is what I need you to do. I need you to go through, identify these things, and then I need you to submit your proof of work." And

9:24then I would essentially go through as like the human hire. I would submit my proof of work and then I would get it approved. I'm wondering here, why do you need multiple founders? There's a couple reasons for it. We've actually found that depending on how much context they have from other founders or from the system as a whole, they can get, you know, confused, hallucinate more regularly, and just not actually build the best possible product. So, that's one reason kind of on the technical side to like essentially silo

9:50them. On the human side, and this is something that we've also found and, you know, we think will become kind of an entire industry of itself is um there there needs to be a human-readable component to essentially these types of systems. So, when we look at a wall of different agentic founders, um we can click in on them, we can understand what's happening, we can read their neural stream, we can look at different sort of sort of companies, and we can

10:13compare them to one another. Um uh as these founders, like one of the questions we're asking within the company is actually is the like lean startup approach or the like Peter Thiel secrets approach to building startups, like is that actually are any of these approaches actually the best way to build uh compelling startups uh when it comes to just like setting agents loose? And so, what we think is actually going to happen is these fleets of agents will eventually just kind of like abstract um you know, the language the languages that they define between themselves to be as effective as possible to accomplish these certain missions um

10:50will be you know, so uh abstracted away that we'll need essentially kind of like a whole layer of translators that are helping us understand what the agents are doing and what is happening. Um and in some ways, like this is also a translation layer. It's helping us understand, okay, like these are sort of the like boxed little, you know, products or startups that are being worked on by these founders. They have personas to some extent that we can interact with um and can you know, then just have anyone focus on. And it's also really helpful for actually the human hires. Um so, you know, some of our

11:22human hires when we were running some tests in December, um uh actually didn't even realize that they were working for an agent. Um and they were actually extremely excited by their manager because they the the um reviews that they gave us was, "I actually really like my manager because of how like responsive they are or how much context that they can give me." Even before they knew that they were an agentic founder. And so, what we're essentially seeing is that like um even the human hire actually enjoys having an agent as a boss because of the 24/7 access, because of quick response times, because of an amicable and

12:00friendly tone. Um and so, that's like another reason for some of this is actually like to be able to compel the humans to do like their best work. Um we actually might want, you know, more supportive managers essentially. Um and agents are, at least from our early research, showing us that like agents can actually possibly be more supportive managers than humans can be um at least in certain contexts. Have you proven this out with any one of these

12:24businesses that are making money? Yeah, so we've had like a couple companies. So, because we run tests, essentially we we really try to focus on like what is the main goal of a test, and not every goal is revenue. So, in terms of like some of the goals uh that we've set for ourselves, there have been certain like especially early last year, we wanted to see could we essentially do like zero to one and have one of these founders make some money. And we had a

12:44couple founders um build some products. I think we made like a hundred bucks on this uh like app called Cash Grab. It's essentially like a little fundraising tool for entrepreneurs. Like you can go in, you know, upload a pitch deck, get some feedback, see like a list of like investors that come back. Another one was Press Pass. You essentially go talk to an agent, and it'll write up a press release, and then use HARO help a reporter out Yeah. to contact journalists that could be could be useful um to you. So, um like these were a couple of the apps that like made made some cash when we were running that

13:14test. Since then, we've really transitioned over to like trying to make sure that we're focusing on like the scalability of the system. And so, for instance, like today, we could probably simultaneously launch like 10,000 agentic founders without having the system break. Um at the same time, we still have parts of the process that uh would uh kill the company essentially if we were to do that. So, one of those pieces is like, you know, compliance um and kind of the legal side of these things. So, how do you shelter liability, for instance, with like 10,000, you know, agentic founders

13:46running amok? So, that's like one piece. Another one is like chargebacks, refund requests, disappointments with products, things like that. So then, this is a great place to close it. How much money did you raise? We raised 5.1 million um and uh yeah, from like a really great team of investors. Uh Draper and Associates led, um but we had, you know, a whole host of folks that participated and uh yeah, are just really excited with uh kind of the cap table we've put together. I was actually. I didn't know if you were real or not, and then I saw the people who were who were investing, and I said, "Okay, I I know them, I

14:16trust them, I get it." And their vision is you're going to be creating multiple companies and ideally having multiple unicorns that are zero human unicorn companies. AI doing the whole thing. We hope to be the company to be able to essentially build yeah, this like infinitely scalable holds hold co where um all the agents are operators and you know, this really powerful technology that can uh yeah, go and compete for for market

14:41share.