Sloane & Pearl

A real fashion brand where AI, not a human, pauses ad spend, rewrites strategy nightly, and writes product copy via Gemini. $18.6K in real sales, one real job created.

Comment

Inspiration

What inspired us? Honestly, we wanted to know if AI could actually run a real business, not just help one. Not a demo. A real store, real customers, real money. Sloane & Pearl (sloaneandpearl.com) has been open since June 9, 2026, and the decisions that matter, which ads get killed, what the ad strategy is, what the product actually says, those are made by AI. On its own. No one approving each one first.

What it does

At the end of the day, this is what it does: it runs the store. Importing products, pricing them, writing the descriptions, building the ads, managing the ad account, all of it. Two months in, real sales: $18,641.28 from 204 real customers. Every five minutes, AI checks every active ad and pauses the ones that aren't working. No human involved. 95 of 127 ads launched, stopped that way. Every night, a second AI reads 30 days of real results and writes a new strategy, and that goes live without anyone checking it first. All 2,529 products, AI-written descriptions, not copied supplier text. Since August 13, that runs on Google's Gemini.

How we built it

We didn't build an AI system from scratch for this. We took a platform we already had running and pointed it at a brand new business, and we're saying that straight up, not hiding it. The rules allow it, as long as the business itself is new, and this one is: its own store, its own products, its own ad account, its own customers, its own staff. What we actually built for this hackathon was one thing: connecting this store's product-writing to Google's Gemini. And we didn't just ship the code, we confirmed it with a real usage record from Google. Not code that could theoretically work. Code that actually ran.

Challenges we ran into

Real talk: this business isn't profitable yet. $18,641.28 in, $27,971.34 out, mostly on ads, and that's a loss of $9,330.06. The product margin is fine, over 67 percent. The ads cost more than they bring back right now, that's the actual problem, and we're not softening that number. We also had a real call to make on Gemini: the plan was customer service replies, but that meant touching a system already live for every store on the platform, too much risk for one requirement. So we moved it to product descriptions instead. That's a real change from the plan, and we're saying so.

Accomplishments that we're proud of

Here's the one that matters most: a real job. We hired a customer service person specifically for this business, July 16th, and she reads and sends every single reply herself. AI never messages a customer. Never. And this one hit us: a real customer wrote to us, unprompted, no one asked her to, "Wow..i didnt think I would hear from you. THANK YOU!! I love the sandal." She said yes to sharing it.

What we learned

What did we learn? Honestly, it wasn't really about the code. It was about being straight about what's new and what isn't. Most of what proves this works, the ad-kill engine, the nightly strategy, the automatic funding, all of that was already running before this hackathon even started. The real work was drawing that line clearly: here's the business, that's new, here's the platform, that's reused, and not blurring it. We also learned something harder: an automated system is only as good as what's feeding it. We built a pricing tool that adjusts prices on its own. We turned it off once we saw it was working off bad data. We could have said nothing. We didn't.

What's next

More people, not fewer. Next hire: a second customer service person. After that, someone for suppliers and quality. On the AI side: fix the pricing tool's data problem and turn it back on, and get ad efficiency to where this business is actually profitable, not just bigger.

Built With

Share this project:

Updates