May 11, 2026

Selling a House Without an Agent: A Live AI Experiment

TLDR

We’re documenting what happens when you sell a house without an agent, using AI to compress execution time from weeks to hours.

The house is in Maresias. R$ 4.950.000. Five suites, 345 square meters, 150 meters from the beach. It’s the kind of property that sits on the market for months if you do it the traditional way — list with an agent, wait, hope.

I don’t have months. I have a full-time job running marketing for ALTHERR and a personal brand to build. So I did what I always do: I threw AI at the problem and watched what happened.

This is a live lab. The house may sell. It may not. What I’m documenting is the process, the math, and what actually works when you strip away the agency layer.

The Math That Makes This Obvious

Before we get to the landing page, let’s talk about why this experiment is worth running at all.

A traditional real estate agent in Brazil charges 5–6% commission. On a R$ 4.95M property, that’s R$ 247,500 to R$ 297,000. Call it ~R$ 300,000.

What does that R$ 300,000 buy you? In theory: market knowledge, buyer network, negotiation skill, paperwork handling. In practice, for a property like this: a listing on the major portals, some photos, and a wait.

What are we spending instead?

Even if the house takes six months to sell and we burn through R$ 10,000 in ads, we’re still net positive by ~R$ 290,000. The ROI is obvious. The only question is whether we can match or beat the agent’s reach and conversion quality.

That’s what this experiment is designed to find out.

The Constraint That Shaped Everything

Brazilian buyers don’t fill out forms. They don’t answer calls from unknown numbers. They WhatsApp.

This isn’t a preference. It’s a behavioral fact. WhatsApp has 95% open rates in Brazil. It’s where deals actually happen. So I made a decision early: the entire conversion funnel would be one tap to WhatsApp. No lead capture form. No “schedule a call” button. No email field.

Just: see the house, feel the house, message us.

This constraint turned out to be liberating. Instead of designing a multi-step funnel, I designed a single action. Every element on the page either pushes toward that action or gets cut.

Building at the Speed of Thought

I started at 7 PM with a simple prompt: “Build me a landing page that sells a R$ 4.95M beach house. Conversion event is a WhatsApp click.”

What followed was 3 hours and 47 minutes of the most efficient creative work I’ve ever done. Not because I’m fast — because I had an AI operations partner that could execute as fast as I could think.

Here’s what that looked like in practice:

7:03 PM — First version live. Hero image, stats bar, property description, WhatsApp CTA.

7:15 PM — My wife reviews it. “Remove the bold from the roof text.” Done in 30 seconds. “The azulejos description is wrong.” Fixed. “Add a scarcity sticker.” Coral badge, rotated 3 degrees, appears top-right.

7:42 PM — Eighteen feedback items implemented. The page already looks better than 90% of real estate listings in Brazil.

8:15 PM — Google Ads conversion tracking installed. Event fires on every WhatsApp click. Value parameter set to R$ 4.95M so Google knows what a conversion is worth.

Here’s what struck me in that moment: implementing Google Ads tracking used to be scary. Now it’s copy-paste to my agent. The same agent that designed the page is also a Google Ads ninja. Magic.

8:33 PM — Slideshow added. Swipeable, auto-advancing, dot navigation. Because static images don’t sell houses — movement does.

9:47 PM — Deployed. Domain pointing live. Campaign published.

Total API cost for the evening: approximately $0.14. The Mac Studio running local models handled everything else.

The Real Lesson

Here’s what nobody tells you about AI: the bottleneck isn’t the technology. It’s your ability to know what you want.

My wife’s feedback was the critical input. She knew the house. She knew what buyers care about. She knew that “ar condicionado quente e frio” sells better than “AC” because Brazilians care about winter comfort. She knew the furniture was from October 2022, not 2024.

And the fixes happened in minutes. Mobile UX tweaks, picture swaps, wording changes — each round of feedback executed as fast as she could describe it. It felt like sitting in a shared office with an expert team.

But I need to be honest: this is not perfect. There were moments where I gave feedback — change this image tag, fix that specific wording — and the agent said it was done, but it wasn’t. I’d screenshot and point it out, and the agent would insist it had changed, as if the memory of the edit was more real than the pixels on the screen. You end up talking to someone who’s delusional about their own work.

I’m not a detail-oriented person. I don’t need 110% correctness. Mistakes happen, and I’m fine living with them as long as I know to look for them. The key is expectation management: this is fast, it’s powerful, but you still need a human eye checking the details. That gap will close over time. For now, it’s part of the deal.

AI didn’t know any of that. What AI did was compress the execution time to zero. In a traditional workflow, each round of feedback meant a Slack message to a designer, a wait, a revision, another wait. Instead, it was: say it, see it, ship it.

This is what I mean by agentic marketing. Not AI replacing humans. AI removing the latency between human judgment and market impact.

The Numbers I’m Watching

The campaign is live. Google Ads is learning. Here’s what matters:

Update — 12 hours later: First qualified lead via WhatsApp. One message, one interested buyer, zero friction. Total cost to acquire: $0.14 in API spend. The ad budget is separate — but the execution layer that captured this interest cost less than a cup of coffee.

I’ll keep reporting as the data comes in. Not projections. Not hope. What actually happens.

What This Means for You

If you’re still thinking about AI as a writing assistant or a chatbot, you’re missing the point. The operators who win in the next five years won’t be the ones with the best prompts. They’ll be the ones who can articulate what they want, judge whether they got it, and iterate faster than anyone else.

Will the house sell? I don’t know yet. The market decides that. What I do know is that the gap between idea and execution closed from weeks to minutes tonight — and that changes the math on every high-ticket sale, not just real estate.

That gap used to be weeks. Tonight it was minutes.

There’s a bigger thought I’m sitting with. What if the first real estate agent to seriously use AI agents in this state cleans up the market? It must be like the first realtors who discovered they could sell on the internet — an edge that a fragmented market will take years to adopt. How much economic value can be captured when the execution layer compresses to near-zero?

What a time to be alive.

Want to see the landing page? casa-maresias

Curious how agentic your marketing operation is? Take the AI Readiness Assessment — 15 questions, 2 minutes, zero fluff.

Want the system that built it? Subscribe to the newsletter — I document the tools, workflows, and decisions as they happen.

2-min check · free · instant result
AI ROI Curve Locator

Is your AI investment paying off already? 3 questions. 30 seconds. See where you sit on the ROTS curve.

1 / 3
/ 100

Enter a valid email address.

No spam. Unsubscribe anytime.

✓

Your full breakdown is on its way. Want the complete picture across all 3 dimensions?

Take the full assessment →