In 2024 I realized AI could do something more useful than answer trivia questions. It could run business operations.
Not “help with” operations. Run them. Research prospects. Compose emails. Generate content. Post to social media. Monitor its own health. Fix problems at 3am while I’m asleep.
I’m a licensed insurance agent. CRPC, CLTC. I spent a decade at Mutual of Omaha, AXA, and Ameriprise before going independent. I have a Series 7 and 66 from a past life. My technical background is “figured it out as I went.”
The laptop
I had an Eluktronics RP-17 gaming laptop. Bought it in 2021 for video editing and as a mobile office. RTX 3060, 32GB RAM, 1TB NVMe. Solid machine.
When the AI projects started, I wiped Windows, installed Ubuntu, and dedicated it to running infrastructure full-time. Named it LaptopLLM. It sat on my desk, lid closed, running Docker containers 24/7.
No dual boot. No VM. Full commitment.
The local LLM phase
First thing I tried was running models locally. The RTX 3060 could handle it. I set up CUDA drivers on Ubuntu, pulled open-source models, and started experimenting.
It worked for experiments. Not for production. The reasoning depth I needed for actual business operations, things like qualifying a prospect against three different ICP profiles or composing a personalized email that references specific regulatory context, that required more than consumer hardware could deliver.
The turning point
Claude Code changed everything. Not the API by itself. Claude Code specifically. The ability to sit in a terminal and build things conversationally.
I’d describe what I wanted. It would write the code. I’d test it. We’d iterate. Every agent, every skill, every dashboard feature, every cron job in the system was built this way.
I’m not a developer who picked up insurance. I’m an insurance agent who picked up enough code to wire AI systems together. There’s a difference. The code isn’t elegant. The systems work.
What happened next
Within weeks I had agents running on cron schedules. One researches prospects. One composes outreach emails. One generates blog content. One posts to X and LinkedIn in my voice. One compiles daily briefings. One audits the whole system’s health.
They coordinate through an event bus so they don’t step on each other. They write to a shared database. They read from each other’s outputs.
The gaming laptop that was supposed to edit YouTube videos ended up running a marketing department.
That was the beginning. The timeline has the rest.