Most people are using AI agents the way they used ChatGPT in 2022.
They poke at it. They get a cool demo. They move on.
Meanwhile, a small group of people is quietly turning those same agents into actual businesses. One-person businesses. Businesses that cost almost nothing to start.
Brett Malinowski (@TheBrettWay) just published a breakdown of 7 AI agent businesses you can start right now, each one runnable by a single person. And unlike most "AI business ideas" videos, this one comes with the actual systems behind each model. He ran one of them himself as a test.
I pulled the whole thing apart. Here's the full list, with the real numbers.
1. An AI ad creative agency for local businesses
This is the one Brett actually tested. The target: car dealerships.
Running ads for a dealership normally means researching competitors, designing creatives, writing copy, managing campaigns, and generating leads. A full-time job. Probably three.
Brett's agent workflow replaces all of it:
→ Create a brand guideline for your agency (colors, typography, ad formats)
→ Have the agent analyze 50 competitor dealership ads and extract the tactics into a strategy document (he saved it as a Claude skill)
→ Feed it the dealership's website. The agent scrapes inventory and generates ad images, copy, headlines, and CTAs from real listing data
→ It uploads everything to Meta Ads Manager with geo-targeting set to a 25-mile radius around the dealership
You hit publish. That's your job now.
The part that got me: he set the agent to re-scan the dealership's site every hour. New listing? It generates the ad and runs it. Car sold and pulled from the site? It pauses the ad. 24/7, on autopilot.
Client acquisition is agent-run too. Scrape local dealerships, pull their Instagram, send personalized outreach. The agent books the calls.
Then you send a $2,000/month checkout link.
An agency with zero designers, zero media buyers, zero copywriters.
2. Agent-as-SaaS for freelancers
Think about Upwork. Thousands of freelancers spend hours every day searching listings, writing proposals, sending messages.
Now imagine an agent that does all of that automatically. Scans Upwork, finds relevant jobs, writes the proposal, sends it. Tools like Lancer AI are already doing exactly this.
The business model is the interesting part. You're not selling software. You're selling agent tasks:
→ $20/month for 200 automated proposals
→ $50/month for unlimited job replies
Brett's bet is that the trained agent becomes the asset. An agent that's genuinely good at one specific workflow will be worth a lot more than a generic tool.
I wrote about this exact shift in AI agents are the new SaaS. The pricing model changes from "access to features" to "work done for you". Freelancer proposals are just the first obvious wedge.
3. A B2B outbound sales agency
Companies spend absurd money on lead generation. And most outbound follows the same four steps: find prospects, get contact info, write outreach, follow up.
All four are now automatable. The agent scrapes LinkedIn for prospects, researches each company, writes the personalized message, sends the connection request, and follows up on its own.
The pricing flip is what makes this a business instead of a tool: charge per qualified lead. $50 to $200 per booked meeting.
You're not selling marketing services anymore. You're selling sales outcomes.
That's a much easier sale. Nobody argues with "you only pay when a meeting lands on your calendar."
4. An AI implementation agency
Brett thinks this will be the most popular model of the next year. I think he's right, and it's the least sexy one on the list.
The service: help normal businesses adopt AI. That's it.
Most companies are still confused about where to even start. So you install the tools (Claude, browser agents, Chrome extensions), you train the team, you write the documentation and workflow templates, and you copy the setup across every team member's computer.
Businesses will pay thousands of dollars for this. Not because it's hard. Because it saves their team hours every single week and they don't want to figure it out themselves.
Brett's comparison: when desktop computers showed up, people made a killing going business to business setting up machines and installing Microsoft Word. Same opportunity, new wave.
You just talk to the team, spot the workflows an agent can eat, and set it up.
5. A podcast sponsorship agency
This one hit home because I run a podcast, and Brett actually pays for this service himself.
Here's how his deal works. His agency negotiates sponsorship pricing and terms, brings the deal back for approval, and delivers the integration. Sponsor pays $10,000? Brett keeps $8,000, the agency keeps $2,000.
He calls the 20% worth it because he hates the mental load of the back-and-forth. I get it. Sponsor negotiation is exactly the kind of work that drains a creator.
The agent version: scan podcast platforms to spot which brands are sponsoring shows, which podcasts are growing, and which audiences match which brands. Then broker the match. Build the media kit, connect brand to show, take your percentage.
There are thousands of growing podcasts and no scalable way to match them with sponsors. An agent makes the matching nearly free.
6. An AI content repurposing agency
You've heard this one before. It's still on the list because the market is huge and the ceiling just moved.
A creator spends hours on one YouTube video. That video can become dozens of pieces: LinkedIn posts, X threads, newsletter sections, TikToks, Shorts, Instagram clips. Post once, appear everywhere. And with the right agent, it happens automatically every time a video goes live.
The obvious question: if it's so easy, why doesn't every creator just do it themselves?
Brett's answer is the most honest thing in the video. If 100 people build this agent, one of them will be best at matching the creator's actual voice and picking the clip that goes viral. Taste hasn't been automated yet. The person who encodes taste into their agent wins the whole market.
And he knows creators will pay, because he pays. He has a team member fully in charge of content redistribution who uses an AI agent to do the work, and Brett still happily employs him.
Two ways to package it:
→ Premium service run by your agent: $5,000/month
→ Agent-as-SaaS: $300/month, or charge per task
7. AI wholesale real estate
The weird one. Also maybe the biggest per-deal upside.
Wholesale real estate means being the middleman: you find someone who wants to sell their house, get the contract, then sell that opportunity to a real estate investor. You take around 10% of the deal's profit for brokering it.
The traditional grind is brutal. Hundreds of calls a day, property record searches, negotiations, buyer hunting.
Agents flip the economics. They can scan listings, search property records, flag properties that fit your buy-box criteria, and generate the outreach. Brett's bet is that agents will be making the actual phone calls very soon, if they can't already.
The real estate rule he quotes: find a good deal and you'll never struggle to find money. So the whole game is deal-finding speed. And an agent turns deal-finding from manual hunting into a system that runs while you sleep.
The actual lesson: systems, not tools
Look at all seven and one pattern jumps out.
The opportunity isn't "using AI". Everyone's using AI.
The people making money treat agents like employees, not tools. They design the workflow, build the process, and let the agent execute it. Brand guideline → research skill → generation → upload → monitoring. That's an employee's job description, held by software.
Most people are still experimenting. The ones building real businesses are doing systems thinking.
If you're going down this road, be careful with the hype side of it too. I wrote about the $1.3M token burn behind AI agent loops and which loop is actually worth stealing. An agent multiplies a working system. It doesn't fix a broken one.
Pick one model. Build the system. Charge for the outcome.
FAQ
Do I need to know how to code to start an AI agent business?
No. Every model on this list runs on existing tools (Claude, browser agents, ad managers, scrapers) glued together with workflows, not code. Brett's ad agency test was built from a brand guideline document, a research skill, and scheduled tasks. The skill that matters is designing the system: what the agent checks, when it acts, and what a finished output looks like.
How much does it cost to start one of these businesses?
Almost nothing. Brett's framing is that each business can be run by one person and costs virtually zero to launch. Your real costs are an AI subscription, maybe a scraping tool, and a payment link. Compare that to hiring the designers, media buyers, and SDRs these agents replace, and the margin story explains itself.
Which of the 7 should I start first?
The AI implementation agency is the lowest-risk entry. It needs no audience, the demand is obvious (most businesses are confused about AI), and every install teaches you workflows you can later productize into one of the other six. The outbound sales agency is the fastest to revenue if you can stomach performance pricing at $50 to $200 per booked meeting.
Aren't these ideas already saturated?
The tools are commoditized. The taste isn't. Brett's point on content repurposing applies everywhere: 100 people will build the same agent, and the one who best encodes judgment (which clip, which prospect, which deal) takes the market. Early advantage right now is real. Most people are still treating agents like toys.
Steal playbooks like this every week
This breakdown came from one YouTube video. On the Profitable Founder Podcast, I go deeper with the actual founders: bootstrapped operators making $100K to $10M a year, sharing the numbers most people keep private.
If you're building toward $100K MRR, it's the closest thing to sitting in on their war room.
