Build and grow a company
with AI, in under an hour.

Cleaned and speaker-labeled. Tool names and proper nouns corrected, no editorial rewriting. Timestamps are from the recording. Recorded Sunday, August 2, 2026.

[0:00] Justin: Welcome in. Today we're gonna walk you through the entire lifecycle of a company built with AI. We're gonna show different companies along the way, but go through what it takes to start, build, and grow a company with AI: the brand, the website, getting traffic, getting customers, the numbers, analytics. And at the end, Ian's 90-10 rule.

[2:00] Justin: For the veterans, you know we mix the format every week. A few weeks ago we built an entire company live in front of you. Today is a bit different: it's more of a show and tell. Everything Ian and Alec are about to show you is a real build. It's polished, it's done, it's working. Nothing is hypothetical or theoretical.

[3:00] Justin: As of 2026, we started a collective. We build agents for companies, so at the end, if you need help building an agent for your business, we'd love to earn your trust and chat with you.

[4:00] Ian: My wife and I started a school last year, Saints Classical. She came up with the idea after getting a master's in education. I jumped in as the tech and brand guy, she knows curriculum, and we launched in fall with 95 students. It was a literal 6-month sprint, and the majority of the work was she and I. We ended up with a full website, a full LMS, an on-demand shop from Brand Sauce, my own company, and full brand everything.

[5:00] Ian: What we learned was that the important thing to start with was the brand kit. Everything flows from that. It's like the brain, the personality, and then you form the skin around it.

[6:00] Ian: I had my OpenClaw take what I had done and build out a brand kit for everybody else to see and work on, so that when other people work on marketing material, they know exactly how to represent your brand. Logo variations, colors, brand fonts, apparel designs, brand images, all downloadable.

[8:00] Ian: Branding, oftentimes people think, is all logo. It's not. It's actually the thinking behind an entity. Even fonts have theories: the serif comes from the Greek, when they used to chisel the letters, so it stayed the higher-learning font style. Each color has its own use case, and you educate the user by the brand representation: our users know the greens and golds are the standard school colors, the navies are clubs and higher learning.

[10:00] Ian: It was just my wife and me. Building a brand is easier with OpenClaw; doing all of this without an AI bot would have been basically impossible. It's a second or third brain: while I'm brainstorming with it, it's building out while I'm building other things. It has exponentially allowed me to reach things I couldn't have reached before, and do things much cheaper.

[12:00] Ian: Rahul, yes, starting with a design.md is a good idea. I've been a designer for 28 years, so I build my own as I go: make this tweak, now save that all as a design skill. If I were starting as a non-designer, I'd find other sites and designs I like, say I really like this part about that, and let your AI build that design MD for you. Then you start building with it and continue to refine it.

[13:00] Ian: Direct the AIs, because otherwise you'll get templated AI builds, which are starting to become pretty noticeable. Behance is a really good place to start. Create Midjourney images, or Nano Banana. Good artists borrow, great artists steal: take someone else's good stuff, tweak it for yourself, and pretty soon it's your own.

[14:00] Justin: You have a brand, you're starting a business, but it's really worthless unless you have a website that converts attention and traffic into customers. Alec, take the baton.

[15:00] Alec: A landing page can now be achieved in half an hour instead of weeks. But if you don't give AI any direction on design, it's gonna end up looking the same; everybody knows what an AI website looks like. We have a specialty in conversion rate optimization, so I'm gonna share an admin panel I've built out.

[16:00] Alec: Everyone's familiar with an A-B test. What you'll see here is an A-B-C-D-E-F-G test: we're rotating different variants of the website. If I give Justin a link and Ian a link, they both get different versions, and we measure what's converting better and direct traffic to the pages that perform. It's very important now because CPMs on Meta and Google are going up and up.

[17:00] Alec: This was Carship Pros: we're generating leads for car shipping companies and we sell those off. We used Lovable, but you can use any of the builders, Claude Code, Codex, Replit. You can see V3 here getting 30% conversion, V1 getting 0%. So now we direct all our traffic to V3, take our top 5, make new versions of those, and continue split testing.

[18:00] Alec: If you wanted to build out 20 landing pages in a drag-and-drop builder, forget about it. Now this is one prompt, literally: make me 20 landing pages and rotate them, and give me an admin panel so I can view the analytics.

[19:00] Alec: I'm feeling generous, let's get that prompt out to anyone who wants it. I'm not of the mindset of gatekeeping anything. This could apply to so many verticals, especially if you're running paid traffic, because you're paying for those clicks.

[20:00] Alec: You don't really know until you know. Let the market tell you what's working. We all have our ideas: I think this will convert. Where's the data? The data's right here. Push more traffic to the good ones, cut the bad ones.

[20:00] Ian: There's actually a tactic to how they look. I did web design for a family of trailer parks and apparently made them look too nice: people started thinking it would cost too much to rent space there. There's a tactic to making it look cheaper, so people don't think you're too expensive. It depends on your audience.

[21:00] Justin: A landing page with no traffic is like a billboard in the desert. Alec, how do you get traffic using AI UGC and AI advertising?

[22:00] Alec: A lot of what we do is run Meta ads, and the algorithm demands an insane volume of creatives. Prior to AI you're paying $250 for a creator to make a UGC video. Now that's achievable with AI in seconds, for free, and the data shows it's working. I put a quick design of a hat together: real advanced prompt engineering, "make a cool dad hat with this logo, give me the product image."

[23:00] Alec: This is all in Higgsfield. It aggregates all the big image and video generation models in a better interface. I took the hat into their Marketing Studio: upload a prompt, the product image, and an avatar. [The ad plays: "If you're using AI every single day but you still couldn't tell me what an API is, this hat is for you. AI4NTP."] Need a better script, but that's not the point. This took 30 seconds to generate, with a prompt I used Claude to help write, and that's literally the hat I generated five minutes before.

[24:00] Alec: It's essentially indistinguishable, or well on the way. By end of year, for certain. These ads let us test hooks and scripts fast, without paying somebody $250 and waiting two weeks for revisions.

[25:00] Alec: And I wouldn't limit it to paid ads. We're all starting to see the rise of AI influencers; this could be repurposed for general social media, or a VSL explaining your product. Any marketing asset where somebody's on camera, it now makes sense to use AI, so you can do it faster, cheaper, and iterate quicker.

[26:00] Justin: Ian, what if we don't have money for paid advertising? Can we position ourselves on ChatGPT, or Google SEO, or GEO, or AEO?

[27:00] Ian: That was exactly my prompt to Enoch, my first OpenClaw. I said: if I give you $100, what would you do? How would you make money? And he said, I wouldn't do any paid advertising. I would build a thing, pay for hosting, and run a bunch of other marketing techniques. He came back with a whole game plan for AEO, SEO, programmatic SEO, GEO, and what I call PEO.

[28:00] Ian: The one he built was Minimatch, a very niche app for minifigure painters. They forget what paints they have, so the app finds paints and matching colors across brands. It's at minimatch.app and on the app store.

[29:00] Ian: AEO is answer engine optimization: LLMs. When you ask your LLM the best way to manage your paints, you want to present answers for exactly that question: FAQs with the right schema, content phrased how people ask, not duplicated, not slop, and accurate.

[30:00] Ian: GEO is generative engine optimization: the AI summary Google pulls up. You may not show up in the search results but you may show up in that GEO spot. PEO is my own phrase, public engine optimization: what the public thinks of you. Reddit, YouTube, blogs, social. Getting mentioned there, managing how you're mentioned, hearing what people say and responding. It's a three-legged stool, and all of them benefit SEO.

[31:00] Ian: This is my Minimatch site. I didn't have anything in February; we launched in March, and most of those visits are me. April, once I started implementing, maybe 55 a day. May, 100 a day. July, about 360, 365 visits a day. And it's only August 2nd and August is already matching April. I was stunned how fast the LLMs picked up on it: ChatGPT is mentioning me a lot.

[32:00] Ian: A lot of it is programmatic SEO: producing a lot of pages that all have different text, context, and focuses around what you do. You pair things up: a location and a niche, a task and a material. We built around 8,000 pages on Minimatch, each primed for AEO and GEO.

[33:00] Ian: And PSEO on its own: this is Brand Sauce, began in May. Gray means Google hasn't put it in its algorithm yet, green means it has. Right now 3,820 not indexed, still reviewing, and 13,200 pages reviewed and in the index. Thanks to my OpenClaw Saucybot. Google won't ding you for AI writing your content, but it will if it's slop, trickery, or not valuable to humans.

[34:00] Justin: Jan asked: what's the real competitive advantage when everyone starts doing this? There is an arbitrage in the market right now. Building with AI today is analogous to buying Bitcoin when it was under $100. Fortune 500s are scared of AI adoption, there's bureaucracy and red tape, but a nimble small business owner can implement these same day. The advantage is being the first mover, in this window, before it's too late.

[36:00] Alec: Everyone can create this, but nobody will. Most people don't take a ton of action. So the competitive advantage is actually going out and using the tools. We're coming up on a world where anyone can do anything, but most people just won't.

[37:00] Ian: Get into it, learn it, and continue pushing the boundaries. If you get in and then stay there, people will pass you up.

[38:00] Alec: Cold email. How many can a person realistically send a day? This is my Instantly account, January 1st to today: 110,000 emails sent, 3% reply rate, almost 1,000 opportunities. And it's not templated: we're pulling in data, pushing it into an LLM, and pushing that into our email distribution software. 110,000 personalized emails in basically half a year.

[40:00] Alec: The core concept, in pretty much every session: scaling the unscalable. You start with your data source, Apify (a data scraper) or something like Apollo (a contact database). That gets pushed into an LLM prompted to write the personalized email, and that gets pushed into Instantly, which distributes to your list.

[41:00] Ian: Of that 110,000, how many of your hours were involved?

[41:00] Alec: Maybe a couple hours a week, probably an hour. Fully automated. With anything you build with AI there's effort up front, but once it's built, it's built, and it runs. The upward trend is also adding infrastructure: your volume is correlated to how many inboxes you have, because each inbox is constrained on daily sends to keep deliverability high.

[43:00] Alec: Reporting. Basic example: within Claude, I have my Google Drive hooked up. My docs are one click away: a sheet with my data, a call transcript. Not toggling between a million tools is one of the core advantages of AI. And skills are big: I have a skill for proposals, basically instructions on how I like my documents to look.

[44:00] Alec: I drop in the transcript and the skill. One, it has context on the conversation and the client. Two, it has context on how I like my proposals to look. Then I iterate, with a bonus tool, Wispr Flow: go change this, go change that, I don't like the font here. If you're not using Claude Code, n8n, Codex, Make.com, any advanced workflow tools, this can all be accomplished within the tools everybody in this room is already using.

[46:00] Alec: Most of my connectors are native to Claude, but some are third party. Take Plaud, P-L-A-U-D, my note-taker: whenever I record a meeting, the transcription hits a webhook, which pushes it into my Google Drive, which I access via Claude in one click.

[47:00] Justin: You can ask AI to set this up for you. Upload this transcript to Claude and say: I really liked how Alec did ABC, or how Ian did XYZ. Show me how to do it. That's a great jumping-off point. We did, in under an hour: brand, website, traffic, outreach, reporting, PEO, SEO, AEO.

[48:00] Ian: Now some practical wisdom. A friend of mine had an idea around prayer: organizing what he's praying about. I said, what if you could plot where people are praying around the world? He said, let's build this thing. I told him with AI I could build it in a month, a week. Seven months later, we finally just launched to beta.

[49:00] Ian: So the 90-10 rule. It's usually 80-20, but with AI you can go further: 90% of the work will take you about 10% of the time, and the last 10% will take the rest. Here are the logo iterations: the first pass people didn't like, it felt AI. Then the prayer hands with PRYR, it's PRYR.app. Then a tagline, still not right. Then the standard icon, which is what we almost launched with.

[50:00] Ian: What we ended up with, a month ago, was no icon at all, just PRYR. We went for a Fight Club feel; we wanted a more masculine app. What I learned: I can overestimate what I can get done because AI goes so fast. Be patient. You'll get really far really fast, but that last part takes a lot.

[51:00] Ian: It's the diamonds example: lab-made diamonds are perfect and a tenth of the price, because we value the imperfections, the little human touches. Once you get your perfect app, start to add your own touch and feel. We call it the Instagram of Prayer. If you're interested in the beta, ping me and I'll send you an invite.

[52:00] Ian: A lot of it was human suffering, honestly: mulling it over, showing it to people, being embarrassed. Be willing to take those lumps. AI can only get you so far: it's always going to think every idea you have is great, and not every idea you have is great.

[52:00] Ian: I used to be a beekeeper. Bees get nectar, take it into their stomach, throw it up into another bee's mouth, seven times, and then it's capped off as honey. Seven iterations. Nature tells us there's a lot to taking in an idea and mulling it over a few times to get a pure thing.

[53:00] Alec: Two core themes: the timeline has been rapidly condensed (that would've been a six-month build a few years ago), and the costs. Hiring developers you might be at $10K on the low end, $50,000 on the higher end for an app like that. Now it's token costs and hosting; it's really just sweat equity.

[54:00] Ian: This app cost us maybe a few hundred dollars, under a thousand, in my spare time, because I have a full-time job. And I'd be happier with this than what I'd probably get from a developer.

[55:00] Justin: The three of us are an AI collective. We build AI agents, automations, marketing automations for a wide range of companies. If you want to chat about us helping you out: ai4ntp.com/calendar. Now, Q&A: "Inspired, but I can't spend $1,500 on each A-B test. Any advice?"

[55:00] Alec: Cold email is very cheap to run: inboxes are about $3 per inbox per month, you can send about 30 emails a day per inbox, domains are about 5 bucks with 3 inboxes each, and the sequencer maybe $50 a month. SEO and AEO: relatively inexpensive, more work than paid ads. And if those are too expensive, it's time to start hitting the phones.

[56:00] Justin: Ian, for the programmatic SEO, AEO, GEO build-out, any capital requirements?

[57:00] Ian: Purely the server costs and whatever your model costs. OpenClaw is free, open source, at openclaw.ai. Once you install it, it's a matter of knowing what you want to do. Very, very cheap.

[58:00] Ian: On security: with OpenClaw, the first thing I do is put it on a server and say put up the firewalls, install Fail2Ban, which bans IPs after about 5 failed login attempts. Then make sure your guardrails are set up: no API keys, no passwords in the codebase or on GitHub, nothing you wouldn't want the whole world seeing.

[59:00] Alec: A lot of people use the cybersecurity argument to avoid building anything. With every build, look at the blast radius: what does the system have access to? If you're housing social security numbers, you've got to be buttoned up without question, maybe bring in a third party. If it's internal and the data isn't super important, you don't need the same level. If the codebase ended up in a bad actor's hands, what would happen? That determines the level of security you need.

[60:00] Justin: We had a lawyer reach out who wanted a server on-prem because their client data can't be exposed to any outside server. It depends on the use case and the blast radius.

[61:00] Alec: On targeting: you really don't need to anymore. Everybody I know, people spending tens of thousands a day, everybody's running broad. Your creative's doing the targeting. Retargeting is different, but pretty much everybody's running broad.

[62:00] Justin: That's a wrap. We're trying to continue doing these once a week. Each one is unique, but the commonality is always sharing how to build with AI. We'll send everyone a follow-up with the recording, show notes, and next steps to collaborate with us.

[63:00] Alec: If something inspired you today: anyone can do it, but most won't, so be one of the ones that do. Get out there and try the tools.

[63:00] Ian: And when you do, send them to us. We want to see what you guys are building with this. We're really excited to see what you build.

This is a lightly condensed version for skimming. Watch the replay for everything, verbatim.