A kid calls a barbershop.
He tells the owner he builds websites, and that he already built theirs. It's done. Want to see it?
Third call of the day, the barbershop says yes and pays him $850.
The website took 2 minutes to make. AI built it before the phone even rang.
That clip is from Brett Malinowski's latest video, and it's the best example I've seen of what he calls the most profitable AI business right now. Not selling AI software. Selling AI-generated websites and other done-for-you services, where AI does the fulfillment and you keep service prices.
I watched the whole thing, pulled out the playbook, and I'm breaking it down here with the real numbers. Because this maps almost 1:1 to what I keep telling founders in the Club: the money is rarely in the tool. It's in the outcome.
Watch the original video first
Everything below comes from this video on Brett's channel. Worth the full watch:
Why services beat software right now
Brett opens with an argument that some of the biggest AI investors are now making out loud.
The most defensible businesses might not be the ones selling the software. They might be the ones using the software to do the work.
Think about it from the software side. If you sell an AI website builder, you're competing with every other builder AND with the next model update that makes yours look old. Brutal.
But if you sell the outcome ("here is your finished website, pay me"), every model update makes your business better. Your delivery cost drops while your price stays the same, so the margin quietly gets fatter every quarter.
Sequoia published a piece saying services are the new software, with a stat that should make every agency-curious founder sit up:
→ For every $1 spent on a software tool, $6 are spent on services delivering the same outcome.
Brett calls the current moment an opportunity arbitrage. A new technology changed how the work gets done, but the market hasn't repriced it yet. Business owners still expect a website to cost $500 to $3,000 and take two weeks. You can now deliver it in minutes.
That gap is the business. And like every arbitrage, it closes. The people who move early keep the upside.
Why a restaurant owner pays $850 for something AI made in 2 minutes
The obvious objection: why wouldn't the business owner just do it themselves?
Because they're running a restaurant.
Hiring, kitchen, inventory, payroll, promotions, angry customers. The last thing that owner wants at 11pm is to learn a website builder, AI or not.
Owners pay for problems solved. They don't care how you solved it. I've said the same thing about AI services businesses before: nobody audits your fulfillment process, they judge the result.
The $850 barbershop site wasn't bought because of the tech. It was bought because a real person called, showed something concrete, and removed a chore from the owner's plate.
The full automation stack, step by step
Here's where the video gets interesting. Brett doesn't just pitch the model, he automates the entire pipeline using the Claude desktop app and its Chrome extension. No code.
The structure is three scheduled jobs, staggered an hour apart:
7:00 a.m. Prospecting runs itself
He picks a niche (local restaurants in Kansas City, because he grew up there and, his words, they're notorious for bad websites).
Then he creates a prospecting skill: a written instruction file that tells the AI exactly what a good lead looks like. Bad website or no website. Skip the big chains, they don't need you.
A scheduled task runs every morning and adds 25 new leads to a spreadsheet with their Instagram handle, phone, and email.
25 leads a day, every day, while he sleeps.
8:00 a.m. The websites build themselves
A second skill file defines how the sites get built: scrape the restaurant's existing site and socials, pull the real menu, real photos, real branding, and generate a custom site that feels made for them.
Then each site gets pushed to Vercel (free hosting), so every lead in the sheet gets a live shareable link. restaurantname.vercel.app.
His design hack is stupidly simple: grab a "brand guidelines" template from Canva that matches the client's style, save it as a PDF, and tell the AI to follow it. The before/after difference in the video is real. One looks AI-generated, the other looks like a designer charged $2,000 for it.
9:00 a.m. Outreach goes out
The third scheduled task reads the sheet, finds each lead on Instagram, and sends a personalized DM with the link to their already-built website. Then a follow-up with a calendar link.
So in theory: wake up, and there are sales calls on your calendar from owners who already saw their finished site.
This is the part that got me. Prospecting, fulfillment, and outreach, all running as business automations before he's had coffee. Brett says building lead lists and writing DMs by hand used to eat half his workday in college. That half-day is now a cron job.
The honest caveats (Brett includes them, so will I)
Two things he's upfront about, and they matter.
First: AI lies. His first generated website only included 5 menu items from a much longer menu. If he'd sent that to the owner, dead on arrival. You have to cross-check the first outputs against reality and tighten your instructions until shortcuts stop happening.
Second: automated DMs live in a gray zone. Platform rules on AI outreach are unclear and spamming hundreds of messages a day is a good way to lose your account. The TikTok kids doing this are mostly cold calling instead, which converts better anyway if you have the nerve for it.
And the bigger point he makes, which I'd underline three times: the number one skill in the AI era is taste.
AI removed the 3 years it took to learn WordPress. It did not remove the need to know what a good website looks like, or what makes one convert. If 100 people run this exact playbook, the one who cares most about quality wins.
The money: $500 to $3,000 per site, then the upsell ladder
Pricing in the video ranges from $500 to $3,000 for the initial website. The barbershop paid $850. His demo checkout was $1,000, set up in under 60 seconds with a payment link.
But the website is not the business. It's the foot in the door.
Once a local owner trusts you, the upsells stack:
→ Google reviews management
→ Local SEO
→ Online ordering setup (DoorDash and co)
→ A monthly retainer for all of the above
One happy restaurant owner refers you to three more. Local business owners all know each other. That's how a $850 one-off becomes a few thousand a month in recurring revenue, which is the same agency-to-recurring path I broke down in how founders are automating their agencies with AI.
My take: this is a cash engine, not an end game
I'll be straight with you. I wouldn't build this as my life's work.
It's an arbitrage. Margins this fat attract competition, and in 18 months "AI-built website" will be table stakes, not a pitch.
But as a way to go from $0 to your first $5K or $10K a month? It's one of the fastest paths I've seen. Real customers paying real money in week one, no fundraising, no 12-month build before you learn anything.
And here's the founder move: the service business funds and informs the product business. You do 30 websites for restaurants, you'll know exactly what software they'd pay monthly for. That's how some of the best bootstrapped SaaS companies started. Service first, software second.
The window is open right now. I doubt it still will be in 2028.
FAQ
How much can you charge for an AI-generated website?
In the video, the range is $500 to $3,000 per site, and a real barbershop paid $850 on a cold call. Price on the outcome (a professional site the owner didn't have to think about), not on your time. The owner doesn't know or care that AI built it in minutes.
What tools do you need to start this business?
Brett's stack is minimal: the Claude desktop app with its Chrome extension for building sites and automating outreach, Vercel for free hosting, Canva brand guideline PDFs for design quality, and a payment link tool for checkout. Total software cost is roughly one AI subscription.
Isn't it dishonest to sell something AI made so fast?
No, as long as the work is accurate and good. The client pays for a finished, correct website they didn't have to build. Speed of fulfillment isn't the product. But you do own quality control: verify menus, photos, and info are real before anything goes out, because AI will happily invent details.
Is cold DM automation safe on Instagram?
It's a gray zone. Rules on automated outreach vary by platform and mass-DMing can get your account banned. Keep volume low and personal, or do what the highest-converting people in this niche do: pick up the phone and cold call with the finished site ready to show.
Does this model scale into recurring revenue?
Yes, through upsells. The website is the entry point; the recurring money is Google reviews, local SEO, online ordering setup, and monthly retainers. One-off $850 projects become multi-thousand-dollar monthly clients, plus referrals from other local owners.
Steal playbooks like this every week
I break down how real bootstrapped founders actually make money, from $100K a year to $10M, every week on the Profitable Founder Podcast. Real numbers, no fluff, founder to founder.
