Why We Chose to Build an AI-First Construction and Development Company (A Pragmatic View)
It has been three and a half months since my last newsletter, and the reason is simple: I have been too busy actually building things to write about building things. This is, I've decided, the correct order of operations.
But writing matters. It forces me to clarify my own thinking, and sharing it lets other people poke holes in it. Both of those things make the thinking better. So here we are.
When I left off, I had just spent much of 2024 and early 2025 building getRealDeal.ai — an application that pulls in MLS data, crunches the numbers, and shows you what investment opportunities exist each day. One page. One purpose. It worked.
This was a hard-won lesson. My previous software project, FluidCM, had sprawled into about twelve apps on a single platform. Twelve apps is approximately eleven too many for one person to maintain, support, and market. RealDeal succeeded specifically because I gave it a single job and refused to let it grow legs.
So there I was with a validated deal-finding tool and a choice to make.
Three Paths
The backend systems with the biggest impact on our young development company came down to three options:
- Build a deal pipeline. Done. getRealDeal.ai handles this. Check.
- Build field efficiency tools. Software to make our construction crews faster on the jobsite.
- Build an AI decision-making system. Something that learns from our work and helps us think more clearly.
Path 2 is what a conventional builder would choose. It's tangible, it's measurable, and it's what every construction consultant would recommend. Bring in a specialist. Buy the software. Train the crews.
I chose Path 3.
Why Path 3
Three reasons, and they're all connected.
The asymmetric bet. In real estate, the money is made when you buy. If you don't buy a property at the right price, you can blow multiple months of work right out the window. A field efficiency tool might get you a 5-10% improvement on executed work. A better deal pipeline gives you more shots on goal. But avoiding one bad deal — one property purchased at the wrong price, with the wrong assumptions, in the wrong market — can outweigh all of that combined.
I wanted a system that could stress-test our thinking before we commit capital. Not a spreadsheet with hardcoded assumptions. Something that adapts to our specific way of evaluating deals and gets sharper over time.
The memory problem. Here is something nobody warns you about getting older in this industry: you start repeating mistakes you already made ten years ago, because you forgot you made them. I have watched this happen to myself, and I have watched it happen across companies — the same errors, the same budget overruns, the same "we should have known better" conversations, cycling through the industry like some kind of expensive folklore.
Since reading Ray Dalio's Principles, I have been looking for a way to codify the lessons and principles that guide our decisions — not in a binder that collects dust, but in a system with actual memory. A system that can say "you tried this in 2024 and here is what happened" before I cheerfully walk into the same wall again.
The learning window. Our company is small right now. I have a capable partner running the field operations, and I have time between deals to think about systems. This window will not stay open forever. Once we scale to ten, fifteen+ active projects, nobody will have time to build foundational infrastructure. It will be too late to start.
And AI itself is moving fast enough that learning it now — hands-on, building real tools, not just reading about it — matters. There is no owner's manual for this technology. Everyone is figuring it out at the same time. The best way I have found to understand what AI can actually do is to sit down with it every day and build something.
What This Looks Like in Practice
I installed Claude Code and Gemini — two AI tools that run on my local machine — and started building. Not a chatbot. Not a demo. An actual operating system for how we run the business.
Over the past few months, this has included setting up our accounting system, building memory structures so the AI retains context across sessions, and creating decision frameworks that encode the lessons we have learned the hard way. The interesting part has not been the technology itself. It has been the process of getting an AI system to respond consistently without drifting off course — which turns out to be a much harder problem than anyone selling AI tools would like you to believe.
I will be sharing the specifics of what we have built and what we have learned in upcoming newsletters. Including the parts that did not work, which are at least as instructive as the parts that did.
What's Next
Next time, I am going to walk through how we set up our accounting system using nothing but Claude Code and an open-source ERP platform — from zero to functioning books. It was not as smooth as that sentence makes it sound.
If you are building a construction or development company and thinking about how AI fits into your operations, I would genuinely like to hear what you are working on. Connect with me on LinkedIn — this is more useful as a conversation than a broadcast.
Originally published as a LinkedIn Newsletter.