Fully Automated Agency: How Brett Signed a Client in 7 Days With AI

Brett Malinowski made $30,000 in a single month selling ads to local businesses when he was 23.

So he knows exactly how much grind a marketing agency takes. The prospecting. The ad design. The campaign setup. The follow-up.

Which is why his new experiment caught my attention: he challenged AI to run a fully automated agency and sign a real paying client in 7 days. The AI picks the niche, creates the offer, generates the ads, and delivers the service. Brett just supervises.

The budget was $500, his existing network was off limits, and he had one week.

He got the client with a few hours to spare. And the part of the business that almost killed the challenge was not the AI.

Here's the full video. Watch it, then let's break down what actually happened.

The Challenge: Three Rules and a $500 Budget

Brett set this up so he couldn't cheat.

• The AI makes every decision. Niche, offer, pricing, targeting. All of it.
• Total budget: $500.
• No using his existing network. Every lead has to be a stranger.

His main tool was Claude, plus Whop's AI website builder and CLI for the storefront and the ad campaigns.

Seven days to a paying customer, starting from zero. Most agencies take months to sign client number one.

Day 1: The AI Picked Medspas. And Ozempic.

First decision: the niche. Brett gave Claude the goal for the week and asked it to do market research.

It came back with medspas.

Brett knew nothing about medspas ("that's where my wife goes to get lip filler or Botox"). But the logic held up: high margin, independently owned, and the owners are doctors who spend all day with patients. Nobody in the building has time to create and run ads.

Then the offer got more specific. Instead of pitching generic beauty treatments, the AI recommended focusing on GLP-1s, the weight loss drugs everyone knows as Ozempic. Massive demand, prescription required, and medspa owners are licensed to prescribe. So the clinics can legally run ads for them, and most aren't.

Pricing came out dead simple: $1,500 per month flat. $1,200 goes to ad spend, $300 is the service fee.

A $300 margin sounds thin for an agency. It's not, because no humans deliver the work. That's the whole bet.

The rest of day one was execution at a speed that should scare traditional agencies:

• Website built with Whop's AI builder from a blueprint plus a paragraph describing the business.
• Brand kit (name, logo, colors, font) in 10 minutes, after Claude researched the top ad agencies' sites.
• A $1,500/month payment link and a cal.com booking link, both live on the site.
• 20 ad creatives, generated after Claude studied the Meta ad library and pulled the longest-running medspa ads (longest-running usually means most profitable) into a document of offers and tactics that work.

Niche, offer, brand, website, proof of work. One day.

The Fulfillment System: Scale Like Software, Get Paid Like a Service

Day two is the part I'd study closely, because this is where "AI-native service" stops being a buzzword and becomes an actual system.

Here's the pipeline Brett and Claude mapped out:

→ Client pays the $1,500 subscription on the website.

→ They land on a form with six questions.

→ Claude uses those answers to generate custom ad creatives and sends them to the client for approval.

→ Once approved, Claude uses the Whop CLI to build the campaign, set the budget, and target a 10-mile radius around the clinic.

→ When someone clicks the ad, a lead form sends their info straight to the front desk's phone and email.

The medspa does zero setup. They just call the lead and book the appointment.

There is no graphic designer or media buyer anywhere in that chain. The service delivers itself, the same way good business automations run a company's back office without anyone touching them.

Brett's line for it: "You can scale like software, but get paid like a service business."

The old agency constraint was headcount: more clients meant more employees and slimmer margins. That constraint just died. I covered the bigger version of this shift in my breakdown of the AI-native service firm, and Brett's medspa machine is the small, concrete proof it works.

100 Prospects in 30 Minutes

Day three: customers. Also automated, at least at the start.

The AI told Brett to stay local. He's in Palo Alto, and a Palo Alto medspa picks up the phone for a Palo Alto guy. People trust people who feel close.

Brett used to spend 4 hours a day prospecting by hand. This time he asked Claude to take over his computer through the Chrome extension and build a list of every medspa within 50 miles.

30 minutes later: 100+ locations, sorted by how likely they were to need the service.

Then one smart filter: remove every medspa already running Instagram ads. No point pitching people who already have what you sell. That left about 60 real prospects.

Then Brett hit the list on every channel he had. Claude sent Instagram DMs through the Chrome extension (after Brett cleaned up his profile, because the first thing a stranger does with your DM is check who you are). Then a fresh Google Workspace address, Gmail connected to Claude, cold emails to the full list.

Then: silence.

Days 4 to 6: Where the Automation Stopped Working

Day four. One email reply. One Instagram reply. Same woman.

Brett asked Claude how many messages it would take to land one customer. The answer: roughly 300 emails per customer. With 60 prospects and 3 days left, the math didn't work.

The AI's suggested fix was Brett's biggest fear: cold calling. And since having an AI make your cold calls is illegal, that meant Brett, personally, on the phone.

What happened next is the most honest part of the video. He procrastinated for 4 hours. He bought a new domain and started building a second website for a home services niche, pure busy work to avoid dialing. (I have done exactly this. Different fear, same fake productivity.)

At 3:30pm he finally called. The first one went badly. The next two went okay.

Day five: the one warm lead came back with "no thanks" and blocked him. He sent 50 follow-up emails, 15 DMs, and made 15 cold calls. Two receptionists wrote down his info for the doctor. Then he watched a Jordan Platten video on agency cold calling, realized he was being way too long-winded, and rewrote his script.

Day six: a doctor actually called back. His pitch wobbled, she asked for his website, which is usually a polite no. 15 more calls on the new script. Better conversations, zero closes.

By that night he was openly wondering if the challenge was dead. People were listening now. They just weren't saying yes.

The Free-Leads Offer That Got the Yes

The fix didn't come from Claude. It came on a walk with his daughter.

Brett realized he was calling strangers and asking them to pay in the very first interaction. He had no proof and no track record, and he was asking for money anyway.

So he flipped it: run the ads free for the first 5 days. He fronts the ad spend out of the $500 budget and hands the clinic free leads. If the leads are good, they continue at $1,500 a month.

Who says no to free leads?

On day seven he hit the phones early, laughing on calls where he used to freeze. Sixth call of the day: yes.

The clinic was skeptical, but free removed the risk. And Brett knew that even if they never paid, he'd walk away with a case study with real numbers, which makes signing clients two through ten dramatically easier.

The client filled out the six-question form, and the machine from day two did its job. Ad creatives in 20 minutes. Approved the same day. Campaign live, lead form wired to the front desk phone.

Seven days, down to the final hours. First client signed.

9 Leads on $200: The Early Numbers

Getting a yes is half the challenge. The ads still have to perform.

Four days in: 9 leads on about $200 of ad spend. That's roughly $22 per lead for a GLP-1 program where a single patient is worth thousands over a year.

The clinic replied that two leads had already scheduled appointments and that they're open to continuing. Brett booked a call to close them on the $1,500 monthly plan.

One client at $300 margin is not a business yet. But the fulfillment is automated, so client 10 costs Brett about the same effort as client 1. That's the entire point.

3 Things Worth Stealing

1. Let the data pick a niche you'd never pick. Brett would not have chosen medspas in a hundred years. The AI chose it on margin, ownership structure, and owner availability. Your favorite niche and your best niche are rarely the same one.

2. Build fulfillment before you have customers. The day-two system meant the day-seven client got creatives in 20 minutes. Speed like that converts a skeptic into a believer during the trial window, not after it.

3. When outbound stalls, fix the offer, not the volume. Brett's breakthrough was not more volume. It was moving the risk from the stranger to himself with 5 free days. If you're getting listened to but not getting a yes, your offer is asking for trust you haven't earned yet.

And the uncomfortable lesson underneath all of it: the AI did everything except the one thing that got the client. A human voice on the phone closed the deal. If you want the longer blueprint for this business model, I broke down Brett's full course in how to start an AI services business.

FAQ

What is a fully automated agency?

A fully automated agency is a service business where AI handles delivery end to end: ad creation, campaign setup, targeting, and lead routing, with no employees doing the production work. In Brett's version, a client pays $1,500/month, fills out a six-question form, and Claude generates the ads, launches the Meta campaign through the Whop CLI, and pipes leads to the clinic's front desk. The founder only handles sales and supervision.

What tools did Brett use to build it?

Claude for research, decision-making, ad generation, prospecting (via the Chrome extension), and cold email (via the Gmail connector). Whop for the website builder, payment link, ad generator, and campaign CLI. Cal.com for bookings, a Google Workspace address for outreach, and the Meta ad library for creative research. Total cash spent stayed inside the $500 challenge budget.

Can AI do cold calling for an agency?

No. Brett looked into it mid-challenge and found that AI-made cold calls to businesses like this are illegal, which forced him to dial personally. That constraint ended up mattering: 30+ manual cold calls and a rewritten script are what produced the first client, after email and DMs generated almost nothing.

How much does it cost to start an AI agency like this?

Brett capped the whole experiment at $500, which covered the domain, tools, and the free trial ad spend he fronted for his first client. The real costs are a website, one month of AI tooling, and enough ad budget to produce a case study. Compare that to a traditional agency, where your first hire alone costs more per month than this entire launch.