And it gets smarter with every project I feed it
Three months ago, I was manually counting light fixtures on architectural plans like some kind of digital caveman. Last weekend, I created an AI agent that not only does my lighting takeoffs automatically but learns from my feedback and gets more accurate over time.
The kicker? I named him RZA—The Ruler Zig-Zag-Zig Allah—because if you're going to build an AI agent for construction work, you might as well give it some personality.
This isn't another "I automated my business and now I'm rich" post. This is about something way more interesting: building an AI that actually learns your specific work patterns and gets better at your job than you are at parts of it.
Two weeks ago, I got a quote for a 600-amp electrical panel. When I asked for "a basic 600A panel," the supplier came back with $16,000 in materials. My initial estimate using RZA? $7,550.
The difference? I was thinking basic service panel. They were quoting a complete Siemens PowerMod tenant metering system with four individual meter modules, surge protection, and test bypass equipment.
That moment made me realize: I don't always know how to ask for what I need. And if I'm not asking the right questions, my estimates are garbage.
So I taught RZA to ask better questions than I do.
Here's the workflow that blew my mind:
Instead of me saying: "I need 4-inch downlights for an office"
RZA asks:
Then RZA reads my plans, counts fixtures room-by-room, generates material lists, and creates professional supplier emails. But here's the genius part: it flags anything it's not sure about and gives me a confidence score.
The learning system is what makes this special. Every time I do a manual verification and tell RZA where it was wrong, it learns. It builds templates for office buildings vs. retail vs. warehouses. It learns that I typically use certain fixture types, certain material ratios, certain room layouts.