How to Build Systems That Actually Work: Two Years of Trying
For two years I tried to systematize how I worked. Not automate — systemize. There's a difference. One of those works. The other keeps you busy.
I had notebooks full of workflows. Notion databases. Trello boards. Zapier connections I'd set up and forgotten. I was managing a pile of processes that mostly existed in my head.
The turning point wasn't a new tool. It was asking one question every Monday morning: “What broke last week that I had to fix manually?”
That question led me to the actual bottlenecks. Not the processes I thought were problems — the ones that actually stopped me from shipping.
The question worth asking every week
What broke last week that I had to fix manually?
For me, it was client onboarding. Every new client took 3 hours of back-and-forth just to get their access credentials, their project context, their preferences. Three hours I couldn't bill. Three hours that taught me nothing about the client.
I systematized that. A simple intake form, a templated first-week checklist, a 15-minute kickoff call structure. The first time I ran it, it took 45 minutes of my time total — including the call.
The one question
Ask every Monday: “What broke last week that I had to fix manually?” The honest answer tells you which system to build next. Don't overthink it — just pick the thing that's been annoying you longest.
Systems only compound if they're simpler than what they replace
If your system requires more maintenance than the work it saves, you've built a job, not a tool.
The best system I built this year: a weekly review template that takes 20 minutes on Sunday. It's not automated. I fill it out manually. But it means I never lose track of what's actually happening versus what I think is happening.
The worst system I built: a full project management workflow with custom fields, status updates, and automated Slack notifications. I used it for two weeks before I was working on the system instead of working.
The simplicity test
Before you build any system, ask: Is maintaining this simpler than the problem it solves? If maintaining it takes more energy than just doing the work manually, abandon it.
Build systems slightly better than your current chaos
Not systems that are supposed to replace all chaos forever.
The discipline is consistency, not elegance. A five-step process you actually follow beats a fifty-step one you abandon by week three.
What would save you 30 minutes this week if it just... worked automatically? Not “what's the perfect workflow.” Just: what is actually slowing you down right now?
Fix that one thing. Use it for a month. Then fix the next one.
What I'd tell myself two years ago
Start with whatever you keep avoiding because it feels repetitive. That's usually the thing that's costing you the most energy without you noticing.
Build the simplest possible version — not the elegant one, the one you'll actually use — and run it for 30 days before you iterate.
The goal isn't a perfect system. It's a system slightly better than your current chaos, maintained consistently enough to compound over time.
The compounding rule
Every system you build and maintain for 30 days teaches you something you couldn't have known in advance. Build the first version bad. Use it. Learn. The second version will be dramatically better — and you'll actually build it this time.
Ready to systematize your AI business?
The AI Business Blueprint includes our exact systems and workflows — the operating system we use to run Regent with zero human employees. Templates, processes, and the 30-day roadmap.