I Built a Self-Learning Website for My Own Company

We had four websites. None of them talked to each other. Our catering leads lived on sticky notes. So I killed all four and built one site with a brain.

We had four websites. None of them talked to each other.

One for Scooter's Coffee. One for Hawaiian Bros. One for hotels. One parent site for BraveHart Development that was supposed to tie it all together but mostly just existed. Four different sites, four different update cycles, four different places a customer or investor could land and get a completely fragmented picture of who we are.

And that was just the front door.

Behind the scenes, our catering operation at Hawaiian Bros was running on sticky notes, spreadsheets, and memory. Someone would call about catering for 200 people. The team would write it down. Maybe it got followed up on. Maybe it didn't. If the person who took the call was off the next day, good luck.

We were leaving money on the table every single week because the system was a human being's memory. And human memory is a terrible CRM.

I'd been staring at these problems for months. Four disconnected brands. A catering pipeline held together with duct tape. An investor relations process that involved emailing PDFs. And I kept thinking the same thing I always think: there has to be a better way to do this.

So I built one.

One Site, One System

The first decision was simple: kill the four sites. Build one. bravehartdevelopment.com, that's it. Every brand, every location, every piece of information about what BraveHart Development actually is, in one place.

Scooter's Coffee, 15 drive-thru locations across Wisconsin and South Dakota through Hart Family Coffee. Hawaiian Bros Iowa, two locations in Davenport and Waukee with a full catering operation, 8 more in development across Iowa. Two hotels BraveHart Development developed and Hart Family Hotels manages: the Sleep Inn & MainStay Suites in Franklin, Wisconsin, and the Holiday Inn Express in Marion, Iowa. The blog we'd been meaning to launch for two years. An invest page with a proper accredited investor gate for our Hawaiian Bros Iowa expansion, a $10 million raise across four tranches.

One site. One source of truth. No more brand confusion.

But consolidation was the easy part. The part that actually matters is what's running underneath.

The Chatbot That Gets Smarter Every Day

On the Hawaiian Bros catering page, there's a chatbot. It handles the stuff our team was drowning in: location routing between Davenport and Waukee, menu questions, pricing, how many servings you need for your group, and it links directly to online ordering when you're ready.

That alone would've been worth building. But that's not what makes it different.

The chatbot runs on The Phi Standard, φ, the same AI platform I built behind CropYield AI. And φ doesn't just answer questions. It learns from every single conversation.

Here's how it works in plain English: every conversation the chatbot has shows up in a monitoring dashboard. I review the responses, approve the ones that nailed it, and edit the ones that didn't. Those approvals and edits become training examples, weighted by quality and performance. The next time a similar question comes in, the chatbot pulls from a library of proven responses instead of guessing from scratch.

Over time, the confidence score climbs. When it's high enough, the chatbot earns autopilot, handling conversations without waiting for human review. But it has to earn that trust one conversation at a time. It literally gets smarter with every interaction because a human is teaching it what "right" looks like.

That's the self-learning loop. Not "machine learning" in the buzzword sense, a practical system where real conversations produce real improvements, automatically. There's a "Powered by φ The Phi Standard™" badge right on the chatbot because I believe in showing your work.

The system also scores its own confidence on every output. Above a certain threshold, it auto-approves. Below it, a human reviews. So it's not reckless, it's confidence-gated. The chatbot earns more autonomy as it proves it can handle more.

What's Next: The CRM Nobody Wants to Build

The chatbot handles the front door. But the other half of the catering problem is follow-up.

Someone inquires about catering. Great. Now what? In most QSR operations, the answer is: it depends on who's working that day. There's no pipeline. No automated follow-up. No way to see where every catering lead stands at any given moment.

🔜 Phase 2 is activating a custom catering CRM built directly into the site, powered by φ. Catering inquiries will flow in, get routed, get tracked, and get followed up on automatically. No more sticky notes. No more "I think someone called about a catering order last week." Every lead gets a status. Every status triggers a next action.

It's not Salesforce. It's not HubSpot. It's a catering CRM built for a QSR operator who needs exactly three things: know who called, know what they need, and make sure someone follows up. That's it. It's built. We're launching it soon.

🔜 Phase 3 is the full investor portal. The invest page is already live with an accredited investor gate for our Hawaiian Bros Iowa expansion. But right now, quarterly updates still go out as PDF attachments. The portal will give investors a real login: documents, updates, portfolio tracking. That's coming.

We're also expanding φ across more of the operation. The catering chatbot is the first live deployment, but the self-learning architecture works anywhere you have repeated customer interactions. Every conversation is training data. Every training cycle makes the next conversation better.

The Thread

This is what I do. I look at my own operations, find the places where we're burning time or losing money because the systems don't exist, and I build the systems.

The EHR company in Carroll, Iowa, same thing. The product existed, the operational infrastructure didn't. So I built it.

CropYield AI, same thing. My stepsons needed tools that didn't exist for managing the business side of farming. So I built them.

BraveHart Development, same thing. We had four websites, a manual catering pipeline, and no centralized way to manage investor relations. So I built it.

The playbook hasn't changed in 25 years. Find the problem. Build the solution. Make it work every single time.

The only thing that's changed is the tools. Now the solutions learn from themselves and get better without me touching them. That's The Phi Standard. Intelligence that compounds, every approved output becomes training data for the next one.

If you're running a business and your tech stack feels like it was assembled by committee over the last decade, four tools that don't talk to each other, three workflows held together by one person's institutional knowledge, and a follow-up process that lives in someone's head, I get it. I was living it.

The difference is I decided to fix it. And I built the platform to fix it for other operators too.

Check it out: bravehartdevelopment.com